Yanzio News

Yanzio Selected as AI Partner for Smart Columbus Community Information Exchange

Sat, 13 Dec 2025 14:26:44 GMT

Yanzio is part of the team selected to build the new Community Information Exchange (CIE) for Smart Columbus and Franklin County, Ohio.

Smart Columbus—supported by the Columbus Partnership and leadership from major corporations such as JPMorganChase, KPMG, Honda, Nationwide, and Goodwill—is advancing connected social infrastructure for a metro area of 2.2 million residents. The selection followed a rigorous, community‑driven process that included 80 RFP downloads, 26 vendor applications, 30 reviewers from 20 organizations, 5 semi‑finalist demos, and 3 finalist, full‑day on‑site interviews.

A Collaborative, Multi‑Partner Team

This award reflects a collaborative, multi‑organization approach to modern social care. The project team brings together complementary domain expertise, including:

Yanzio — AI and Data Quality
Visionlink — Technology platform partner
LinkLive — Telephony and contact center solutions
Connect211 — Website implementation and domain expertise
AHC Health IT — Health IT integration
World Institute on Disability (WID) — Equity and accessibility expertise

Together, we will help build infrastructure that makes it easier for residents to access support and easier for providers to deliver whole‑person care.

Our Role: AI for Trustworthy, Expandable Data

Yanzio’s mandate is to ensur e the trustworthiness of the CIE’s resource data —and to expand its coverage.

Data Integrity: Our AI Assistants continuously identify inconsistencies, gaps, and outdated records that traditional approaches frequently overlook.
Discovery & Expansion: Our Discovery Tool surfaces potential services missing from the directory and queues the m for human curator revie w and verification.

We believe future‑ready human services re quire dramatically better data quality—a nd that trustworthy data is the foundation for interoperability, governance, and outcomes measurement.

Read the Announcements

Smart Columbus announcement:
https://smartcolumbus.com/insights/2025-10-cie-vendor-announcement
Visionlink post:
https://www.visionlinkblog.org/2025/11/building-together-visionlink-is.html

16 charts that explain the AI boom

Tue, 28 Oct 2025 00:38:18 GMT

Excellent, short report visualizes the boom

"Understanding AI" published a great new report today that is less than a five-minute read.

They present 16 charts, each with a short explanation, from different sources that really tell the story of the speed and magnitude of the boom we are in driven by AI.

Click here to read the article.

Which AI do I use most often? It depends....

Wed, 10 Sep 2025 13:42:05 GMT

ChatGPT 5 is what I recommend, though not what I use most

Occasionally people ask me what AI they should consider using, so here is my current, updated list.

When pressed to recommend just one AI, I still suggest ChatGPT because it is so versatile. Just remember that ChatGPT 5 has a "model router" and performs best when you direct it to "think deeply" so that it chooses the reasoning version.

But it's not the AI that I use most frequently.

Here's the list of AIs that I use nearly every day, and generally how I use them, in descending order of frequency:

Perplexity (usually their Sonar model selected) = replaces Google for web search
Claude 4.1 = coding and other technical tasks; contract analysis (Claude Opus 4.1 is superb at these, though I hit rate limits quickly so use Claude Sonnet more often)
ChatGPT 5 and o3 = most everything else, especially my custom GPTs like Neil’s Fun Finder
Gemini 2.5 Pro = when I need large context or PDF analysis
Github Copilot = coding assistance and analysis directly in my IDE (Visual Studio)
Grok 4 = occasional analysis of random things. It always impresses me, I should use it more often.

For greater control over how they perform, I often use the "developer playgrounds" instead of the consumer-facing websites or apps of these. This lets me vary the "temperature" and also modify the "system prompt" separately from the "user prompt" for greater precision and steer-ability of the AI. It also lets me experiment with tasks that I intend to implement programmatically via their APIs. For example, I can refine the structured output they will emit to be ingested by code I am writing that orchestrates the use of AI for projects like the Yanzio AI Assistants .

Honorable mention goes to a variety of other tools, including Replit, Lovable, Cursor, Poe and other AI tools I use sporadically.

I also use Whispr Typing to dictate voice into any editable text box. Free, and so much faster than typing.

Greater Twin Cities United Way 211 calls Yanzio "a valuable tool for quality assurance and data precision"

Fri, 01 Aug 2025 02:16:48 GMT

"Yanzio's attention to detail and adherence to our style guide makes it a valuable tool for quality assurance and data precision."

Greater Twin Cities United Way (Minnesota) serves over 500,000 people in multiple counties. They respond to over 340,000 requests per year.

Their resource curation team has been using Yanzio's AI Assistants for a number of months and are delighted with the results.

"Yanzio Tools empower our team by integrating with our updated style guide. With Yanzio, we’re able to improve the accuracy and consistency of our data, ensuring every record aligns with the latest taxonomy and agency updates. The Taxonomy Review Curator streamlines the process of matching descriptions to correct taxonomy terms while maintaining clarity. Yanzio's attention to detail and adherence to our style guide makes it a valuable tool for quality assurance and data precision."

-Kristen Fogerty, 211 Resource Program Manager

Which computer use agent is best currently?

Thu, 31 Jul 2025 12:49:35 GMT

After testing four of the many available, one stands out as the best, for now

Ever since Anthropic announced a computer use agent last fall, many companies and open source groups have been working furiously on similar projects.

A "computer use agent" can operate software on a computer to accomplish a task you give it, allowing you to work on other things while it churns away. Some of them can only operate a web browser, while others can operate any software installed on your computer.

I've tested four of the many systems currently available:

My real-life test cases

I've given all of these tools a couple of real-life test cases. First is to find a low-cost round-trip airline ticket from San Francisco to Orlando in the near future so I can visit my parents. It's the kind of routine research one may perform on the web when thinking about taking an optional trip when you might have time off from work but want to base it upon low airfares and an approximate and flexible date range.

Another test is to help manage my fantasy baseball roster for the day. I've been in a league with one of my sons and several of his friends for 9 years. Since there are games every day of the season, it takes time to keep your roster updated with good players who are starting that day.

Which one is best?

For the most part, these tools are fascinating to watch, but slow and error-prone. Like captions in a movie, they announce each step and retry in deep detail, like "I see a calendar, let me choose the date the user asked me about" or "I see a sorting icon at the top of the list, I'll click that to sort by date". But they often misunderstand visual elements or get stuck when something pops up in the user interface, obscuring a control they need to move forward.

And by slow, I mean at least 10 times slower than a human would perform a task.

But of these, one stands out as reasonably fast and accurate: Perplexity Comet.

It's actually just a browser (it can't use other locally installed apps on your computer) but does remarkably well. If you've used Perplexity, it is easy to use Perplexity Comet and really fun to watch it conduct its work to accomplish the task you give it. And because it is reasonably fast, the process of iterating its work with refinement instructions isn't frustrating like it is with the other tools.

A glimpse of things to come

I think this is just a glimpse of the future, where we'll increasingly use web browsers less and less, and instead ask AI agents to accomplish tasks for us, and it will use software to do our bidding. Who would have thought, just a few years ago, that we'd witness a reduction of web browser usage starting around now?

AI has higher Emotional IQ than humans - new study

Mon, 30 Jun 2025 14:53:36 GMT

Scores 81% on Emotional Intelligence exams, much higher than 56% average for humans

In a new study published in May 2025 from Communications Psychology, six leading Large Language Models (LLMs) took five gold-standard emotional-intelligence (EI) exams and walked away with an 81% mean - much higher than the human norm of 56%.

The study argues LLMs now excel at understanding feelings (cognitive empathy), even if they don’t feel them (affective empathy). For mental-health bots, leadership coaches, and conflict-resolution tools, that’s highly useful.

Researchers told GPT-4 to invent brand-new EI exams . It spun up fresh scenarios, options, and scoring keys that proved statistically interchangeable with the psychologist-crafted originals when humans tried them.

The email newsletter AI Fire summarized the implications: "Three years ago we joked that chatbots had the empathy of a toaster. Today they’re beating us at our own 'uniquely human' game and writing the rulebook. Ready for an assistant that understands you better than you do?"

What is AI's Intelligence Quotient (IQ)?

Thu, 19 Jun 2025 17:20:51 GMT

The best AI has surpassed "genius" level and 99% of the human population

The body content of your post goes here. To edit this text, click on it and delete this default text and start typing your own or paste your own from a different source.

A website called the Visual Capitalist recently published the IQ scores of the (currently) best AI models.

OpenAI's o3 scores 135, which exceeds what is considered "genius" and makes it smarter than 99% of the human population, based on the IQ test.

Just over a year ago, the best AI's IQ score was just a 96, a little below the human average.

Where will AI score in another year?

Mindboggling.

Discover new services with Yanzio's latest AI Assistant

Mon, 12 May 2025 13:30:45 GMT

Finds high-value services in minutes for you

Finding new services is an ongoing task for resource curators that is difficult and time consuming, but essential for aiding your help seekers.

Your organization might have expanded its coverage area, have low familiarity from distant regions, or just not been able to keep up over the years.

High-quantity unmet needs are also a major obstacle to delivering the best referrals over time.

Yanzio's new Discovery AI Assistant can help immensely. By searching for specific services in geographic areas that you target, it can find new services that meet your inclusion criteria. It even will apply your style guide rules and give you the option of pushing the well-structured records into your I&R software.

Baseball roadtrip - AI solving ever more advanced challenges

Thu, 17 Apr 2025 13:13:12 GMT

Research, planning, reasoning, synthesis, selection

My family follows Major League Baseball (MLB) a little more than casually. It all started when we moved back to the San Francisco Bay Area in 2010 and within months, the local Giants won the first of three World Series in alternating years - 2010, 2012 and 2014.

Since then I've sought to see a home-team game for every MLB team and am up to 19 of 30.

But I digress. I've been issuing the following test to various AI models for two years and it's done a decent job in the past, but never with perfect accuracy:

Construct a schedule that lets me drive between multiple MLB stadiums this summer to see a home game in the shortest amount of time and without backtracking.

With yesterday's release of Open AI o3 , it definitively nailed it. Check out how well it researched, considered, optimized and presented the trip.

There's a lot going on here. It had to look up the schedules of each team for this season, find dates where they lined up, sequence them geographically, consider alternatives, and present the optimal solution. Yes, I confirmed the schedule is correct for all of the games it recommended.

This whimsical example demonstrates incredible capabilities for far more serious research and topics.

AI's capabilities are advancing at a breathtaking pace. How are you using it in your professional and personal life?

Manus the AI Agent performs tasks for you on the web

Mon, 07 Apr 2025 13:01:15 GMT

2025 is the year of "agentic AI"

Pundits describe 2025 as the year of "agentic AI" - meaning systems of generative AI that can plan and take action on your behalf.

Last month, a previously unknown firm from China surprised the AI world by releasing Manus , an AI "agent" that can do just that.

I immediately applied for early access and was granted a few days ago. Here are a few tasks that it performed well for me. (All replays are significantly sped up - in real-time they each took about 5 minutes.)

Find a low-cost flight from San Francisco to Orlando on a certain date, nonstop, with departure and arrival times in a certain window. I could have had it actually purchase the ticket with my credit card if I really were wanting to travel that day.
Research home-delivered meals for my elderly parents via Door Dash. Match their food preferences and keep it under budget. Similarly, Manus could have actually ordered the meals and scheduled the deliveries had I requested it.

Perhaps the most interesting part of the experience is watching Manus open a web browser, navigate to different web pages, identify information and clickable items, and fill out forms to gather needed information and take action.

For comparison, I gave the same prompts to several other AI systems - ChatGPT, Grok and Sonnet. Predictably each did pretty well, but none of them are capable of taking action - like booking a flight or ordering food delivery - the way Manus is.

Anthropic (maker of Claude Sonnet) and Open AI (maker of ChatGPT) both announced "computer use" in the past few months, but the former is not yet publicly available, and the latter is only available on the paid $200/month Pro plan.

The race is on to enable similar agent-powered AI systems that can plan and take action on your behalf. The next stage of AI's evolution is starting to come into focus. How will you use it in your personal and professional life?

First Contact of Tampa Bay saves hundreds of hours with Yanzio tools

Tue, 25 Mar 2025 11:30:00 GMT

"Transformed our information and referral services"

First Contact (formerly known as 211 Tampa Bay Cares) implemented the Yanzio AI Assistants nearly one year ago. They started operating as a crisis hotline in the 1970s and have grown to offer 2-1-1 other services in their region and across Florida, responding to over 46,000 calls annually .

"Yanzio’s cutting-edge technology has transformed our information and referral services, enhancing our ability to provide accurate, real-time support to those in need." - Micki, President/CEO

We're often asked for case studies quantifying the positive impact our platform has on organizations, and Micki is certain it has been large for them:

"With these tools, our database curators have saved hundreds of hours in quality assurance reviews allowing them to focus more time on relationship building with local community organizations."

In addition to this added efficiency, they also believe it has improved their quality too:

"[Yanzio]'s innovative solutions have enhanced our ability to serve our community with greater resource information accuracy and efficiency."

We're grateful to serve First Contact as a client. Thanks Micki and team!

Useful Custom GPTs for work, fun and research

Tue, 18 Mar 2025 12:15:00 GMT

Answers to recurring questions you may have

All of the generative AI vendors have enhanced their user interfaces in useful ways.

One of my favorites are the "custom GPTs" available from Open AI's ChatGPT. They allow anyone to pre-load a set of instructions to ChatGPT that can be useful on a recurring basis.

For example, I am frequently looking for fun events to attend with my family here in the San Francisco Bay Area. Previously I would click on a number of bookmarked websites that publish calendars and lists of events, read through them all, and decide if any of them would be interesting to us.

Recently though, I built a custom GPT I called "Neil's Fun Finder" to do this for me. I told it to seek events happening in the designated timeframe that are local to me and similar to the types I enjoy attending (AI conferences, outdoor music festivals, live theater, etc). I also enabled "web search". So now I just open it in my ChatGPT app, type in something like "this weekend" and get a list of fun things we might like to attend.

Here's my current full list of these, available to anyone but pre-loaded with my preferences. So you may want to just take inspiration and make some for yourself. They're very easy to build , no coding experience required, though you do have to be on a paid Open AI ChatGPT plan to create them.

Neil's Fun Finder
Neil's Menu Picker - just take a photo of a menu at a restaurant and it recommends what to eat
Neil's Restaurant Picker - give it a town or neighborhood name and it picks good restaurants
Neil's Music Picker - discover new songs and artists based on a mood, vibe or genre
Neil's AI News Summarizer - given a long list of news from a newsletter or website, it summarizes them
Neil's AI Event Distiller - picks from long lists of upcoming AI events to those I'd be interested (there are multiple AI events every day here in the San Francisco Bay Area)

Enjoy!

Top 100 Gen AI Consumer Apps

Fri, 07 Mar 2025 17:07:21 GMT

Updated list from a prominent Silicon Valley venture capital firm

This week, Andreeson Horowitz released their fourth installment of the most frequently used apps based on generative AI.

Take a peek for inspiration about new uses of AI for your personal and professional world.

Many of my favorites, like ChatGPT, DeepSeek (I actually use a US-hosted version of this), Perplexity, Claude, Poe and Grok are on the list but there were a few fun surprises!

https://a16z.com/100-gen-ai-apps-4

211/CIE San Diego says Yanzio tools "revolutionize" their work

Sun, 02 Feb 2025 14:52:27 GMT

A highly respected I&R and CIE

"Yanzio's AI-powered capabilities revolutionize how we connect individuals with essential information. "

- Peter Battistel, Chief Technology Officer, 211/CIE San Diego

211/CIE San Diego serves the county's 3.2 million residents. In 2024 they answered 458,000+ calls, 212,000+ texts and served 219,000+ total clients, including referrals from their well-curated database of over 6,000 services.

They are widely regarded as industry innovators, notably with the Community Information Exchange (CIE) they built in partnership with other major service providers in the region. Their CIE is the model for others forming across North America, evidenced by the well-attended CIE conference they host every year, among many other activities in support of their peers in other communities.

Fully adopted the Yanzio AI Assistants into their daily workflow

Starting over four months ago, their resource database team uses the Yanzio AI Assistants in their daily work of updating and improving service records.

Peter continues to explain the importance to 211/CIE San Diego:

"By automating certain tasks and enhancing data analysis, we’ve improved the accuracy and efficiency of our service delivery and supporting our staff while also keeping our value of people first and human interaction at the core of our work.”

- Peter Battistel, Chief Technology Officer, 211/CIE San Diego

API integration with their resource database platform

In parallel we've jointly built an Application Programming Interface (API) integration between their Customer Relationship Management (CRM) platform, Salesforce.com, and the Yanzio AI Assistants. As soon as a service record is updated, it is "pushed" to Yanzio and the AI Assistants immediately run evaluations. Within seconds, resource editors get feedback on the quality and structure of those records, according to their own internal Style Guide, taxonomy best practices, and the Inform USA standards .

" Our decision to leverage Yanzio’s tools and products helps us at 211 San Diego achieve our commitment to provide important resources and information to our community through innovative solutions."

- Peter Battistel, Chief Technology Officer, 211/CIE San Diego

How good is DeepSeek? One amazing example

Wed, 29 Jan 2025 05:06:36 GMT

A practical question about car purchasing yields a great, thorough answer

I won't bother rehashing most the chatter about DeepSeek R1's recent entrance to the AI world, other than to say, it may be the most important event in generative AI since the launch of Chat GPT in November 2022.

For the uninformed: It is a new AI developed in China that seems to be as good as the best American-made AI models, at either 25 times less cost or, if you download and run it yourself, it's free.

A real world example

To truly understand how capable it is check out the following question I posed to it today. Edited only for formatting and readability:

I am considering buying a used car and am considering two options:

The gasoline-powered sedan gets a combined 28 miles per gallon according to the EPA and costs $20,000 after all taxes, fees, tax credits and rebates.
The electric vehicle gets 105 miles per gallon equivalent (also and EPA rating) and will cost $30,000 after all taxes, fees, tax credits and rebates.
Otherwise the specifications between the two cars are close enough that i consider them similar.
I live in Northern California.

Ignoring maintenance costs, how many miles will I have to drive for the cost of charging the EV to overcome its additional price, compared to the gasoline car? Do whatever research you need to so that you can make accurate calculations.

Intentional ambiguity

As detailed as that question seems, notice that I omitted some key information:

The average price of gas and electricity in my area, though I did tell it the region in which I reside.
The conversion between "miles per gallon" and "miles per unit of electricity".
I also gave round numbers for the price of the cars, just to make the result easier to evaluate.

First, the answer

I was amazed at the answer itself, which is very thorough, reliable and readable:

Here's the breakdown, everything below this until the "Conclusion" is its answer - this is all from DeepSeek R1 as a response to my question:

.............................

To determine how many miles you need to drive for the electric vehicle (EV) to offset its higher upfront cost compared to the gasoline car, we'll analyze fuel and electricity costs in Northern California using the provided search results.

Key Inputs
1. Vehicle Costs
- Gasoline car: $20,000
- EV: $30,000
- Price difference: $10,000

2. Efficiency
- Gasoline car: 28 MPG (EPA combined).
- EV: 105 MPGe (equivalent to 105 miles per 33.7 kWh, or ~3.12 miles per kWh) .

3. Energy Prices
- Electricity: Average residential rate in California is 31.64 cents/kWh (as of September 2024) .
- Gasoline: While not explicitly listed in the search results, the average price in Northern California is approximately $4.50/gallon (as of early 2025, based on typical regional data).

Cost Per Mile
1. EV Cost
- Energy consumption: 33.7 kWh / 105 miles = 0.321 kWh/mile
- Cost per mile: 0.321 kWh/mile x $0.3164/ kWh = $0.1016

2. Gasoline Car Cost
- Cost per mile: $4.50/ gallon / 28 MPG = $0.1607

Break-Even Calculation
The additional $10,000 cost of the EV is offset by savings in fuel costs over time.
- Savings per mile: $0.1607 - $0.1016 = $0.0591
- Break-even miles: $10,000 / $0.0591 per mile = approx 169,200 miles

Sensitivity to Energy Prices
The break-even point depends heavily on fluctuating energy prices:
- If gasoline rises to $5.00/gallon: Break-even drops to ~129,900 miles.
- If electricity costs decrease to 25 cents/kWh (e.g., via off-peak charging): Break-even drops to ~135,000 miles .
- Higher electricity rates (e.g., 35 cents/kWh): Break-even increases to ~184,000 miles.

Additional Considerations
1. Time-of-Use (TOU) Rates: Shifting charging to off-peak hours (e.g., midnight–6 AM) can reduce electricity costs by 30–50% .
2. Solar Power: Pairing the EV with solar panels could eliminate charging costs entirely, drastically reducing the break-even distance .
3. Incentives: California offers rebates for EVs and solar installations, which could further offset upfront costs .

Conclusion
At current energy prices, you’d need to drive approximately 169,200 miles for the EV to become cost-effective. This distance is sensitive to gas and electricity prices, as well as charging habits. If you plan to keep the car long-term or leverage solar/TOU rates, the EV becomes financially viable much sooner. For a personalized estimate, check your local utility’s TOU rates and gas prices.

.................

OK, it's Neil again. Wow, pretty amazing, right? All the research it did, the calculations, the extra work around price sensitivities and additional considerations? Far better than any of my know-it-all friends might have produced.

It thought for 68 seconds to get this answer. Let's take a peek:

You probably wouldn't want to read all 2,058 words that it wrote while it was thinking, but here's a snippet from the middle to illustrate just how meticulous it is:

....But since gas prices vary, the answer should mention this assumption and perhaps provide a range. Additionally, check if any other factors in the search results affect this. For example, time-of-use rates in California (webpage 4 and 6) could lower electricity costs if charging during off-peak hours. If the user charges during super off-peak, the rate might be lower. For instance, webpage 6 mentions SDG&E's super off-peak rates as low as 33 cents/kWh in some plans, but that's in San Diego. Northern California might have different rates, but the general idea is that TOU rates could reduce electricity costs....

That's why the world is abuzz about DeepSeek

It demonstrates methodical, fact-based research, self-revision, hypothesis-making-and-checking, and so many more impressive "cognitive" feats.

Try it yourself

I encourage you to try it yourself today either directly at:

https://chat.deepseek.com (free account)

Or if you are concerned about sharing your chats with a company hosting in China (understandable) here's another free option hosting a version of DeepSeek entirely in the US:

https://groq.com/
First login with a free new account or use your Login with Google.
Then make sure you change the default AI model from "llama3-8b-8192" to "DeepSeek-R1-Distill-Llama-70b".

Whichever tool you use, ask it your question and click on the "Thinking..." link to see it deliberating as it composes an answer for you.

New Jersey 2-1-1 thrilled to use Yanzio tools

Tue, 21 Jan 2025 13:30:00 GMT

An early Yanzio client and participant in two pilot projects with us

The New Jersey 2-1-1 Partnership began making information & referral services in 2002 and now serves all of the state of New Jersey. With a total population of 9.5 million people, NJ211 responds to over 400,000 requests per year, making referrals from their database of over 6,300 programs. They are Inform USA accredited .

"Thrilled" with the Yanzio tools

CEO Melissa summarizes her team's feeling about using the Yanzio tools: "Thrilled!" She goes on to share that find that "the efficiencies that Yanzio has brought to our resource database will save time and improve quality".

An innovator in I&R and technology

Well-respected in the I&R field, NJ211 displays their innovative spirit by participating in both of the pilot projects we conducted:

Our first pilot of early versions of the Yanzio tools, in Q1'2024
Our current pilot of a new "Discovery" tool that we expect to launch in Q1'2025

Thank you to Melissa and your entire team, we continue to appreciate your collaboration as a client and your help to advance I&R with AI.

Gen AI advancement speed faster than previous major tech

Tue, 14 Jan 2025 13:30:00 GMT

Reaching 100 million users faster than Internet, Mobile Phones, Television

Credit: Generated by Anthropic Claude Sonnet 3.5

Credit: Arificial Analysis AI Review 2024

Credit: Peter Gostev, LinkedIn

Credit: Peter Gostev, Linked In

One of the most impressive aspects of the new generative AI tools like ChatGPT has been their ability to understand the world and generate relevant and useful and accurate text, audio, images and video. I use numerous of these tools in my daily work and it multiplies my productive output on a wide range of tasks.

But perhaps equally as impressive is the speed at which their adoption is happening, relative to other major technology advances in human history.

I asked Anthropic Claude Sonnet 3.5 to generate this first chart for me, which illustrates how many years it took a new technology to reach its first 100 million users. The steam engine was nearly 100 years, television 25 years and the internet just over 6 years. But generative AI has only taken 2 years an adoption continues to spread very quickly.

This advancement speed adds to society's challenge

For those not watching AI closely, you may think that the field is just making a lot of noise and waving hands around how great AI is. But behind the scenes, the quantity and quality of these tools continues to advance at an astounding rate.

Check out the second chart at the left, showing the capabilities of various Large Language Models (LLMs) over time. Notice how much they have advanced as a group in the two years since ChatGPT 3.5 Turbo launched. And notice how often the lead has changed regarding which model is fastest. Open AI, Google, Anthropic and Meta have been in a tight race, and more entrants are coming all the time.

Researchers, large companies, governments and startups are deploying these tools in increasingly innovative ways. You can expect to be interacting with AI, whether you know it or not, more and more in your daily life.

While the opportunity for societal advances and productivity boosts is huge, it also leaves us little time, as a civilization, to grapple with safely and responsibly deploying these tools throughout our lives.

No signs of slowing down

If you are wondering if the hype and progress have plateaued... it hasn't. Startups, venture capital firms, and large enterprises alike are making massive investments to continue building bigger, better and faster generative AI tools.

For example, Microsoft is now investing more in them than in all of their other infrastructure spending combined, which amounts to over $42 billion just in the last two years. Half of that was in the second half of 2024 alone. Other large players are also investing many billions of dollars.

Additionally, NVIDIA - the chip maker fueling much of this capability - just started to ship their latest Blackwell GPU. Independent benchmarks show it is 32 times faster than its predecessor the A100, which is less than 5 years old.

These are the primary chips that these LLMs are "trained" upon, so as they get deployed into data centers, you can expect the capabilities of the LLMs they produce to continue advancing at a significant rate for years to come.

What does this all mean?

As the pundit Ethan Mollick likes to note, even if generative AI capabilities froze today, we would still have years or decades to implement their incredible capabilities into the way that we work, live and play.

But in fact we can expect yet more dizzying advances to be announced almost weekly. My best advice is to stay tuned in, learn how you can use these tools in your own work and personal lives, and get ready for more civilization-level advances to come.

Webinar: iCarol hosting practical uses of AI in I&R

Mon, 25 Nov 2024 18:10:24 GMT

December 11: Yanzio co-presents with iCarol and DGI

Join industry experts from Yanzio and DGI Associates for an eye-opening exploration of AI’s practical applications in I&R. Learn how cutting-edge AI tools can help resource managers and data curators streamline workflows, improve data accuracy, and prepare for the future—all while addressing critical concerns like data protection and AI risks.

Get a firsthand look at how early adopters in I&R are already using AI tools to supercharge client and resource management systems like iCarol . Through live demos and real-world insights, discover how easy it is to integrate AI into your organization to save time, enhance efficiency, and future-proof your services.

My favorite AI and robotics examples as of November 2024

Tue, 12 Nov 2024 02:40:05 GMT

AI that attends meetings for you, does your laundry, makes coffee, drives you to your destinations, and more.

It's been a few months since I have posted my favorites, and many awesome new tools have emerged. Here are some of my current favorites:

Pi0 is a AI-powered autonomous robot that can do a wide variety of household chores, including folding your laundry, making coffee and more.

Atom Limbs is an AI-powered artificial arm that can be controlled by muscle and brain commands, and it learns the person's intentions the longer she wears it.

Anthropic's AI Claude can now use your computer independently to plan trips, write code, and much more.

HeyGen's avatars can attend Zoom meetings for you , even interacting with real humans based on your past communications that it has learned from.

Figure AI's robot can listen, see, converse and act with a human to accomplish physical tasks.

Google's Project Astra is aware of the physical world and can quickly identify objects, provide information and infer meaning with little context provided.

AI is now doing research on AI advances , providing researchers with novel ideas to accelerate AI development.

And AI is now designing computer chips that - you guessed it - are accelerating the research and deployment of yet more AI.

And perhaps my favorite of all - the Waymo self-driving taxi , with no one at all sitting in the driver's seat.

Double the efficacy of your Style Guide and resource data quality

Tue, 05 Nov 2024 12:17:56 GMT

A comprehensive style guide is essential to good data

Until now, it's been impossible to quantify the effect of clarifying and expanding the rules in a style guide. But we've just finished a project that shows even a good style guide's quality can be improved by almost 2x.

We ran a large dataset containing more than 25,000 real resource database service records with a good Style Guide, through our Style Guide Curator . The result was that 61% of the records perfectly adhered to its rules for all field types.

We then synthesized a much more comprehensive Style Guide, incorporating far-ranging rules and guidelines only seen infrequently across the wide range of them in use by I&R agencies today. Together they contained much more prescriptive guidance for the best practices of written resource records.

Then we ran the same 25,000 real resource database service records through our Style Guide Curator, and only 32% of them perfectly adhered to its rules. In other words, the new Style Guide found twice as many legitimate issues with the same data, as the original Style Guide. Which means there's a significantly higher amount of data improvement than first believed.

I've been promoting the importance of Style Guides , and I've already read more this year (at least 40) than I did for my entire tenure building a major I&R software platform.

So I strongly encourage you to take a closer look at your own Style Guide to expand and clarify its rules. And if you'd like assistance from us, we would happily do a project with you like the one we recently completed for Inform Florida .

I&R software integration with Yanzio: iCarol

Wed, 16 Oct 2024 00:17:57 GMT

2-way data exchange via APIs

We've long known that the ideal place for a resource curator to see the evaluations of our AI Assistants is right within their primary Information & Referral software system, while they're editing a record.

But those platforms didn't start 2024 with any inkling that integrating Yanzio's results in their user interface should be on their radar. Since they had their own product development plans, we built the Yanzio web user interface to help our early customers find, see and use the evaluations of our AI Assistants.

That's why today we're please to build upon the initial features that one of the major I&R platforms, iCarol, announced a few months ago .

When you're viewing the AI evaluations for a particular record, you can click a button to push the recommended changes back into that record in iCarol, eliminating the need for copy/paste and possible user error. In particular:

For the Style Guide Curator , push any field’s AI-recommended new value directly to the corresponding record.
For the Taxonomy Review Curator , push the AI-generated new service description (that incorporates all taxonomy assignments that were missing from it).
For the Taxonomy Recommendation Curator , push a new taxonomy assignment.

Improving the workflow of resource curators as they incorporate Yanzio into their daily work is a high priority for us, and we'll continue making advances like these in close coordination with the software vendors.

211info Oregon says Yanzio tools significantly enhances their data quality

Tue, 01 Oct 2024 13:15:00 GMT

They highly recommend it to other I&R agencies

211info serves all of Oregon and southwest Washington state for over 20 years, answering more than 110,000 calls and over 280,000 website searches per year. With 150+ people on their team, they serve nearly 5 million residents.

Their resource database team started using the Yanzio AI Assistants earlier this year, and have integrated them into their daily curation work.

"The Yanzio Style Guide Curator has been invaluable in helping us identify errors we might have overlooked, significantly enhancing our data quality" says Kurt, their Director of Resource Center.

We've worked with them over the past few months to incorporate changes and improvements in their Style Guide, resulting in even better AI-powered evaluations and recommendations for them.

Kurt continues: "We highly recommend it to other I&R agencies considering it.... it could make a notable difference in your data accuracy!"

Kurt and his team are some of our most active users, and we're delighted they're getting so much benefit from using the Yanzio tools.

Database curators - job opening and training from DGI

Tue, 24 Sep 2024 13:00:12 GMT

They're hiring as well as offering training

We've collaborated on several projects with DGI Associates, and they've asked us to let our clients know about two items of interest to Data Curators in the I&R industry:

Community Resource Specialist – Database Curator (CRS-DC) Training is open to all. Apply here on their website.
Hiring an Associate to work for DGI on their client projects - requirements and application instructions available here on their website.

Findhelp 211 in Ontario uses Yanzio to quickly see actionable efficiencies

Tue, 17 Sep 2024 17:17:00 GMT

With over 900,000 contacts per year over phone, text, chat, email and web, accuracy matters

Findhelp 211 , based in Central Ontario, Canada sought effective AI tools for I&R to help with their large data curation tasks. They provide direct support in six of Canada's provinces and online support in three territories.

With over 17,000 service records in their resource database, the potential for AI to augment their staff and processes is large.

"Yanzio has been a great tool to quickly test the capabilities of AI with 211 data workflows" according to Dave, the Director of Information Resources and Technology. "We are able to quickly see the actionable efficiencies that using AI in some of our day-to-day processes creates."

They've been using the Yanzio AI Assistants for the last few months and are methodically applying it to a number of components of their resource database curation.

"Most significantly, we are able to use AI within the boundaries of our sector's official data standards" to apply guidelines and best practices while editing records.

2-1-1 Fairfield County finds Yanzio tools "incredible"

Mon, 09 Sep 2024 13:15:00 GMT

They appreciate the fresh, objective perspective

2-1-1 Fairfield County, Ohio serves a population of over 150,000 responding to over 50,000 contacts per year. They've been doing I&R work since before they were a 2-1-1, beginning in 1977, and first achieved their Inform USA (formerly AIRS) accreditation in 2014.

They recently discovered the Yanzio AI Assistants and, after their free trial, decided to move to a full plan for their entire resource database.

When asked what factors led to them making this choice, their resource curator Cassandra cited several exciting advancements that Yanzio provides them:

A "fresh objective perspective" it gives them on their database records and quality
The "responsiveness and customization" approach of the Yanzio team
The "ability of the software to adapt to fit our needs" because it is applying their own Style Guide rules, taxonomy customization and other best practices

Cassandra summed up her praise by calling the Yanzio tools "incredible", and we're delighted she feels that way.

Inform Florida 2-1-1's use Yanzio AI to create a Style Guide

Wed, 04 Sep 2024 13:00:02 GMT

11 agencies blended their style guides in just 6 weeks

I've written before about the significant impact a well-written style guide can have on database quality, especially when it is consistently applied to your resource database records.

So when the leadership of Inform Florida approached Yanzio in early 2024 to help them create a new style guide for the state network of eleven 2-1-1 agencies, we eagerly agreed to help.

The goal was to create one synthesized style guide that aggregated the best ideas from all of their member agencies, taking care to resolve all of the conflicting guidelines that surely would exist between multiple different documents.

The result exceeded everyone's expectations, with a comprehensive, statewide style guide that is one of the best we've ever seen. And we completed the project in just 6 weeks, significantly faster than doing this without AI.

Executive Director Tori notes that the new Style Guide..

" ...will create more consistency and quality resource information across participating 211 organizations, especially as other AI tools are adopted. If you are considering how to use and apply AI to your work in information & referral, I highly recommend a call with Neil! "

Read the full account from Inform Florida's Executive Director and Data Analyst. And if you're interested in a similar project for your coalition, please contact us !

211 Finger Lakes NY uses Yanzio for accurate referrals

Wed, 28 Aug 2024 14:43:28 GMT

Keeping over 4,000 resource records accurate and complete

The 211 HELPLINE operated by the Institute for Human Services in Bath, NY provides comprehensive information and referral services to the 265,000+ residents of Allegany, Chemung, Schuyler, Steuben, and Yates counties.

Earlier this year, they evaluated the Yanzio AI Assistants in a free trial and decided to proceed with a full plan.

"The use of Yanzio has proved to be very advantageous when capturing data" says Karen, the Call Center Manager and Database Curator. Their small team maintains over 4,000 resource records and operates a call center that responds to calls and texts 24/7, while also maintaining a publicly-searchable online directory.

She continues: "Yanzio provides insurance that our data is efficiently processed and meets the Inform USA (AIRS) standards which allows the call center to convey the most accurate and current information to the client."

Thank you, Karen and IHS, for being a Yanzio client and helping us continue to improve our AI Assistants for I&R!

Walmart says AI provides 100x productivity boost

Fri, 16 Aug 2024 14:38:44 GMT

Executives say AI is 100 times more productive at updating product pages than people

I'm frequently asked by people interested in using AI in I&R how effective it is. I've collected a number of anecdotes from clients that are compelling, and here's one from a different industry that I think is very relevant to our I&R.

"John Furner, president and CEO of Walmart U.S., said generative AI has helped the company populate the attributes and characteristics of hundreds of millions of items, and that it would have taken 100 times longer to do this manually."

This type of work sounds a lot like editing a resource database of service provider records and emphasizes one of the core benefits of using Yanzio's AI Assistants: assisting with the detailed curation of records in a manner that both streamlines the details of data editing while applying the vast range of your guidelines and best practices.

Here's a full report .

Presenting at the 2024 Inform USA Virtual Conference

Mon, 05 Aug 2024 14:36:00 GMT

Join the Yanzio session Friday August 9th, 11:00am Eastern time

I'll be presenting most of the content from the May 2024 in-person conference. If you missed it then, please join for this session.

Topics include:

Why is AI suddenly so hyped?
How can AI be useful to I&R?
What are the ways Yanzio is using AI for I&R?
Is it any good? Is it safe?
Is now the time to start using AI, or should I wait and see?
How could I get started?

If your Information and Referral organization is pondering any of these questions, please join us this Friday (this is a paid event hosted by Inform USA) or click the Contact Us button to learn more.

Enhancements to our Taxonomy Curators

Tue, 16 Jul 2024 20:29:09 GMT

We've been cooking up some great enhancements

Out of all six of our current AI Assistants, I've been happily surprised at the reception people are having for the Taxonomy Review and Taxonomy Recommend Curators. This month we've released enhancements to both of them based on requests from our clients.

Taxonomy Review gets "recommended new description"

This tool reviews all of the existing taxonomy assignments for a service record, and compares it to the service's description. If any assignment's name and definition aren't reflected in the description, our AI Assistant flags it with a low score.

The new enhancement is that it will also recommend an enhanced version of the description that blends the current one, with the concepts from the name and definition of those previously-not-represented taxonomy assignments. And it's not just appending the definition: it actually reinterprets the taxonomy terms and blends them concisely into the new description, sometimes doing so deftly with more than one taxonomy term in a sentence.

Taxonomy Recommend get an explanation

This tool recommends new taxonomy assignments to consider making to a service record based on the service description, and it only does so from your customized, active taxonomy terms.

The new enhancement is that it will explain why it thinks the recommended new term would work well for this record, and even compare it to any similar existing assignments.

Clients are saying this makes them "better taxonomists"

We've been delighted at the positive feedback on these tools, including that people are doing their jobs better and learning the taxonomy more deeply because of them.

Insights from the Inform USA conference from DGI

Tue, 25 Jun 2024 13:00:04 GMT

Our colleagues share their thoughts on the conference in Grand Rapids

While the Inform USA Conference in Grand Rapids, Michigan was the very first for Yanzio, it was the 9th one for me and one of the very best.

Our friends and colleagues from DGI Associates, who also exhibited and attended, just shared their summary and insights from their perspective.... I highly recommend this short read !

My current favorite AI tools - June 2024 edition

Tue, 18 Jun 2024 13:00:08 GMT

I use these every day

Occasionally friends and colleagues ask me what AI tools they use and how to use them, so I thought I'd share will all of you my current list of favorites.

First place starts with a two-way tie between Perplexity and ChatGPT4o:

Perplexity - They call themselves an "answer engine" to differentiate their service from Google (a "search engine"). Start using it instead of Google and you'll see how cool it is not to get a list of links, but rather the actual answer to your question. Get the paid plan for the best possible results, and then only use Google when you are looking for something very specific like the phone number of a restaurant.

Here's a recent example of how I used it for our younger son entering college: Give me a list of the best rated classes that fulfill general education requirements for an incoming freshman at San Francisco State University, favoring those available between 10am and 3pm. It nailed it!

ChatGPT4o - With their updated release last month, it's still the best Large Language Model (LLM) available. And while I'm not a fan of their naming, this latest release 4o (the "o" stands for "omni" because it is great at writing, voice, images, video, etc) is even better. Make sure to sign in and choose GPT4o from the menu at the top. Better yet, upgrade to the paid plan for higher usage limits and other features.

It's pretty good at search too, but I tend to use it to help compose prose, generate images and charts, analyze long documents and more.

Other LLMs I use are Anthropic's Claude 3 (paid to use their best version, Opus) and Google Gemini which has an insane "context window" of 2 million tokens, so it's particularly good at very long documents.

For realistic "avatar" videos, like the two you'll find on this website , I use Synthesia . For a low monthly fee you get a very easy to use tool for creating videos of human-like avatars saying whatever you want them too.

And of course for the coders out there like me, Github Copilot is insanely great.

I'm constantly exploring, learning and dabbling with a multitude of other tools but these are the ones that float to the top of my list.

PS: Yes, I used AI to enhance my headshot on this article. There are many but I used HeadshotPro . It probably makes me look 10 years younger!

Yanzio's Inform USA 2024 conference session - AI + I&R: A Perfect Match!

Tue, 28 May 2024 12:45:00 GMT

Our exciting session with almost 200 attendees and many questions

I presented to a packed room last week, at the Inform USA conference in Grand Rapids, Michigan.

While we didn't finish all the planned content, I certainly conveyed my main messages and activated the industry to think more about the use of generative AI in Information and Referral (I&R):

- Generative AI is now very good at many things, language in particular.

- I&R is all about language, from the curation of resource data, to the aiding of help seekers and sharing data with partners.

- Yanzio has practical tools using generative AI now, actively helping I&R's and 2-1-1s in the US and Canada.

- Yanzio's results are rated 98% "excellent" or "very good" quality by an independent Inform USA-certified database curator.

- AI can multiply the effort of any knowledge worker who uses it, helping I&Rs that have been chronically understaffed and underfunded for decades.

- Any I&R can reach out and start working with Yanzio to deploy AI in their organization now.

The lively discussion in the room reflected both the excitement and concerns about using AI. While many people see the incredible opportunities, some are concerned about risks and potential job displacement it may cause.

My response is that regardless of what you may currently think, this powerful new technology is here today, and is improving at a rapid pace. Our civilization has gone through many technology changes, and we're now going through another one.

I&R has faced some of those technology transitions with initial trepidation, including telephony (1960s), computerization of paper records (1990s), publishing resource directories to the web (2000s) and live chat and texting (2010s). But in every case we found a way to channel and incorporate those changes and became a better, stronger industry as a result.

Other voices at the conference who painted a picture that AI is unreliable, are not doing their homework with real I&R data and real AI working on practical applications. It is working well and already producing reliable benefits. Yanzio can prove it with over 5 million AI requests made already this year for multiple 2-1-1's in the US and Canada. Assessments by independent, certified resource database curators verify that it is producing excellent results.

And the question "will this cause job displacement" is answered by the people running I&R operations, like 2-1-1 Directors...not Yanzio or any AI vendor, or commentators with no experience running an I&R.

Leaders running I&Rs will rely upon their experience and vision to choose to employ the people and tools that are most effective to the help seekers they serve.

I'm counseling my university-age children on how to navigate these changes: Be the person who knows how to use AI well in your professional life, not the person who doesn't know or won't learn AI and is therefore concerned about their job.

I readily admit that we all have to be thoughtful, careful and ethical in our deployment of AI technologies. I welcome industry guidelines, rules and government regulation and legislation and ask fellow industry leaders to help us start that process. Those conversations already started last week with senior leadership at Inform USA and United Way Worldwide.

We as an industry have the opportunity to deploy and shape the use of AI, aligned with our standards, ethics and best practices. And in ways that solve those chronic staff and funding shortages we've endured for decades.

If we don't take that opportunity, other parties will do so and possibly in ways that are not aligned with the vibrant and high-integrity ecosystem we've built over almost 50 years.

Who is ready to join me in this incredible opportunity to innovate and advance the I&R field, our way?

Here are my presentation materials .

Deep-Fake Ryan Gosling explains Large Language Models

Tue, 14 May 2024 13:00:00 GMT

Got 8 minutes? Let Fake Ryan Gosling explain how LLMs work (link below)

A clear, though a little technical, explanation of how a Large Language Model works, by a convincingly good Deep-Fake of Ryan Gosling: Here's the link

75% of global knowledge workers use generative AI in their daily work

Thu, 09 May 2024 19:12:32 GMT

A fascinating new report by Microsoft and LinkedIn shows widespread AI adoption amongst knowledge workers globally

To help leaders and organizations overcome AI inertia, Microsoft and LinkedIn looked at how AI will reshape work and the labor market broadly, surveying 31,000 people across 31 countries, identifying labor and hiring trends from LinkedIn, and analyzing trillions of Microsoft 365 productivity signals as well as research with Fortune 500 customers. The data points to insights every leader and professional needs to know—and actions they can take—when it comes to AI’s implications for work.

Check out the full report here: https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part

Connecting Yanzio to your daily work - the Yanzio API

Tue, 30 Apr 2024 12:42:31 GMT

Integrate our AI Assistants into your primary software systems

Information and Referral practitioners use important tools in their daily work of curating resource databases, such as their I&R software. This is where they create and manage data about the community service providers to which their contact centers refer help seekers for specialized care.

Wouldn't it be great if the Yanzio AI Assistants could give you their recommendations right in those systems, as part of your daily workflow curating those records?

By using our Application Programming Interface (API), you can do just that. When two or more software systems are connected via their APIs, they can exchange data in real-time making them work as one coordinated system.

We're delighted to announce that the Yanzio API is ready for this work.

So if you're excited about the significant leap forward that our AI Assistants bring to I&R, get even more excited about the prospect of easily integrating them into your daily work. Just contact us to find out how to make this happen, in concert with your I&R software vendor.

The Open Referral HSDS data standard's role in Yanzio's AI

Tue, 16 Apr 2024 14:29:35 GMT

Accelerating data ingestion by days

Open Referral's leader Greg Bloom recently invited me to post on their website. Check out my synopsis of using AI in I&R, as well as the role that the HSDS standard plays in accelerating Yanzio's work: https://openreferral.org/irs-big-opportunity-with-ai/

Style guides - unsung heros in I&R

Tue, 09 Apr 2024 13:00:00 GMT

Useful to humans and AI

Nearly every Information & Referral (I&R) organization maintains a Style Guide.

What is a Style Guide?

For the uninitiated, it is a " collection of recommended best practices " for the content and structure of your resource data. It's usually written and maintained by the more experienced and respected staff members to codify how they want resource database curators to perform their daily work.

You'll normally see prescriptions and examples for every record type and field type that either suggest or require the format and type of content that should be present.

Most I&R's will update their Style Guide every few years to reflect emerging best practices, ideas, cultural shifts and vocabulary. And then begins the long process of applying those changes to their entire resource database.

Style Guides are quite useful to Yanzio's AI

One of first AI Assistants we built is the Style Guide Curator . We interpret each I&R's Style Guide into AI instructions, and then evaluate every record and field for conformance to its guidelines.

We're suddenly learning new ways to improve Style Guides

A surprising outcome, both to Yanzio and to our clients, is that once people see the recommendations from our AI Assistants, they realize they want to update their Style Guide. Common epiphanies include:

tightening the language to make the recommendations and examples more clear and prescriptive
Adding more examples to capture the breadth of the service types in the database
Rethinking grammar and formatting rules for things like times (in hours of operation)
Avoiding generalizations like "one-stop shop" and "offering a range of community services"

Those improvements can be more rapidly incorporated

Once they make those changes, Yanzio modifies the AI instructions for their data, re-runs our AI Assistants, and they instantly get to see where they need to make changes in their records to apply the new Style Guide changes.

Bringing harmony to multi-partner I&R networks, like US states and Canadian provinces

We're working with several state and provincial networks of I&Rs to harmonize the disparate Style Guides of their members into a single definitive Style Guide for all of them to use. You can imagine that the process of aggregating and debating the blending of 10 or 20 such Style Guides would be a lengthy, tedious and even contentious process.

By using Yanzio's AI, we're able to synthesize the Style Guides and, through iteration and review, get most of the way to a final product. Yes there will still be review and discussion about the final version, especially to resolve guidelines that may conflict with each other from different member Style Guides. But we anticipate months can be shaved off this process.

And then, once the synthesized Style Guide is done for the network... you guessed it... we incorporate it into the AI instructions for each of the member I&R organizations in the network, and they instantly see which records and fields need updating to align with the new Style Guide.

I shouldn't be surprised - this is exactly the kind of outcome that Style Guides are supposed to create. But it illustrates how AI can significantly improve and accelerate incorporating their guidelines across large numbers of I&Rs.

The AI competition continues to get more interesting

Tue, 02 Apr 2024 13:00:00 GMT

New releases from Anthropic, Google. More coming soon too.

It's been a while since I first asked " Which AI is best for I&R " and it's time to give you an update.

While Yanzio has almost exclusively been using generative AI from Open AI (GPT3.5 and 4), we always keep our eyes on the fast-moving field for new entrants to consider.

Google has been slow, but don't count them out

In February, Google released Gemini 1.5 making rather bold claims about its capabilities. They've been criticized for being too slow and lagging behind Open AI, especially when many of the seminal advances powering the stunning advances in the past 18 months, originated from Google's own research labs. (Indeed Microsoft's AI-powered Bing and new entrants like Perplexity are challenging Google's home turf, search, in ways unimaginable just a year ago.)

Yanzio applied and was granted "private preview" access to Google Gemini Pro 1.5 and we are very impressed with its capabilities. Like the rest of the AI world, we are eagerly awaiting them to release Gemini Ultra 1.0, which should be soon.

Based on what we're reading, expect Gemini to soon become a serious competitor to Open AI.

Anthropic joins the leading pack

Earlier this month, Anthropic (funded by Google and Amazon, among others) released Claude 3 and Yanzio was delighted to get early access. After a few weeks of analysis, it is proving to be a very capable Large Language Model (LLM). A variety of benchmarks and industry observers are comparing it favorably to both Open AI GPT models and our initial testing fully confirms that view. We're excited to start incorporating it into Yanzio's portfolio.

And more are coming soon

The lesser-known competitors are working hard to keep up and it seems almost every week we get a major announcement from Mistral, Facebook's Llama, Twitter's Grok, Databrick's DBRX and many others. Even Apple has announced their own MM1.

Open AI isn't resting on its laurels

Speculation is rampant that we'll see a new release from Open AI sometime this year, either called GPT4.5 or GPT5. Their CEO has said openly that the leap forward will be equivalent to the move from GPT3 to GPT4 (which was staggeringly impressive). If true, we're all in for a new round of stunning surprises and capabilities.

Stay tuned for an exciting 2024!

Preventing AI embellishments and hallucinations

Tue, 26 Mar 2024 13:00:00 GMT

Significant progress on accurate I&R recommendations

A recurring theme from Yanzio is the need for trustworthy AI, especially with data that is used to help vulnerable people in need. That's why we devote a lot of time and attention to finding tools and approaches so we are always improving, like the RAG Triad and AI Safety with your data .

When AI "hallucinates", it is embellishing or outright inventing information in a confident manner, even when it may be utterly incorrect. This is one of the primary reasons people may have reservations about using AI, and the source of humorous and even concerning news articles. But as one analyst noted, everything generative AI does is a hallucination, it just happens to be correct most of the time.

We had been seeing this occasionally in Yanzio results and made a concerted effort to investigate, detect and reduce it. The "Fees" field seemed to suffer the most, so we rolled up our sleeves and had a close look.

Tackling frequent hallucination in the Fees field

By analyzing large sets of AI recommendations from multiple I&R organizations we found hallucination rates of up to 21% in the Fees field - meaning about 1 in 5 records had some embellished or invented information produced by AI. For example, if originally a record said the fees were "Sliding scale", the AI might recommend a revised version of "Sliding scale. $50 for the first visit and $25 for subsequent visits." Not only were the dollar amounts invented, they aren't really a "sliding scale" which has more to do with the help seeker's ability to pay for any visit.

Over the course of several weeks, we improved the rate of hallucination in the Fees field down to just 5% by changing the AI instructions, and also using different combinations of Large Language Models (LLMs) for different field types. (GPT, which powers ChatGPT, is an example of an LLM.)

Even better, we have a different LLM evaluate the recommendations and it is flagging nearly all of those 5% instances before a person sees them.

Hallucination rates are now below 1% for the Fees field

As a result, what a person actually sees as an AI recommendation for the Fees field now has a hallucination rate of far below 1%. A fantastic improvement!

We are applying these techniques to nearly all fields now and seeing similar improvements. We are actively working on further improvements, and new tools and methodologies are emerging in the industry that promise even more progress.

Stay tuned...things are moving fast but Yanzio continues to keep pace with the evolution of AI.

Monitor service provider websites to update your records

Tue, 19 Mar 2024 13:00:02 GMT

Discover when services, locations, contact information and more change at one of your service providers.

Our latest AI Assistant reads data on the public website of any of your service provider organizations, and will compare it with the data in your record about them, and summarize the differences for you.

It will even do this on a continual basis, notifying you if something significant changes so you can keep your record about them up to date.

Find out more on our Products page.

AI Safety and protecting your data

Tue, 12 Mar 2024 13:00:01 GMT

You have questions about trusting AI with your data

After speaking with a number of Information and Referral organizations, and paying attention in general to our society's reaction to the growing presence of Artificial Intelligence (AI), we're well aware some people are cautious, skeptical and even concerned about using it.

Who has access to my data? Will they use it to train their AI and threaten my business? How can I be sure that it gives me truthful answers?

We've provided transparent answers to these questions and more, on our new page about Yanzio AI Safety . Please do let us know if you have any questions - especially ones we haven't yet covered there.

Yanzio AI tools now available for your I&R

Tue, 05 Mar 2024 03:23:24 GMT

Our thanks to the five 2-1-1s in the successful pilot project

With the very successful pilot project ending last week, we're announcing the general availability of the Yanzio AI Assistants to all Information and Referral (I&R) organizations. For those of you who have been eagerly awaiting your turn, now is the time.

Over the course of the pilot project, we learned, improved and deployed many enhancements, both to the AI Assistants themselves and to the web-based user interface used to find the most impactful recommendations.

Here's the reaction from just some of the five pilot 2-1-1s, which included three state-wide 2-1-1 networks and represented data from four different I&R software vendors:

" Yanzio can make text more rich and meaningful."

"I’ve just been playing around in the Taxonomy Curator. I LOVE these suggestions."

"This process is really helping retrain me on how I think about what I put in a resource."

Our website is also updated with all of our current products , different price plans , and how to get started with a free trial .

If you've been curious about how you could use AI in your I&R, now is a great time to give it a try.

Targeted feedback to each of your Resource Curators

Tue, 27 Feb 2024 13:15:00 GMT

Help them improve in applying Style Guide, Standards and Indexing guidelines and best practices

As our pilot project concludes this week, we're announcing our next AI Assistant - for "Resource Curator comprehensive feedback".

We take all the scores generated by our other AI Assistants and, for each Resource Curator at your organization, we find the most targeted and impactful recommendations for improving how they edit resource records. The goal is to continuously improve their work with the difficult task of consistently applying rules and best practices from a variety of sources.

Take a closer look on our Products page and please let us know if you are interested in learning more.

Meet Yanzio's new Inform USA Standards Curator

Tue, 20 Feb 2024 14:00:01 GMT

Our latest AI Assistant helps keep your database consistent with standards and best practices.

The vast Inform USA Standards have been evolving for over 50 years and encompass a wide range of best practices for Information and Referral operations.

Our new AI Assistant applies the guidelines and standards from their latest version to your resource database, to keep it consistent on an ongoing basis.

Learn more on our Products page , and as always, please contact us if you have any questions.

Factoids about 2-1-1 on 2/11 day

Sun, 11 Feb 2024 15:21:28 GMT

Celebrate the important North America dialing code today

Since the US represent dates as "month / day", today's date of 2/11 corresponds to the dialing code that serves nearly all who live in the US and Canada: 2-1-1 . You may see posts around social media declaring and celebrating this number that is an important component of Information and Referral in North America.

What other numeric significance could those three digits possess? Not surprisingly, ChatGPT has some interesting answers.

In numerology, the number 211 could be interpreted through the significance of its individual digits and their sum. It combines the energies and attributes of 2 and 1, with 1 appearing twice, amplifying its influences. Number 2 resonates with duality, balance, and partnership, while 1 relates to new beginnings, independence, and self-leadership. This combination might suggest a phase of personal development, independence, and making significant life decisions.

Year 211 AD: During this year in the Roman Empire, Severus Alexander became emperor. He was known for attempting to reform the Roman Empire and was notable for his efforts to rule through a council of wise men.

The number 211 itself is a prime number, which means it is only divisible by 1 and itself. This is a unique characteristic in mathematics, highlighting its simplicity and foundational role in number theory.

So, happy 2-1-1 day!

How does your AI access my data to make recommendations?

Tue, 06 Feb 2024 14:15:00 GMT

Importing your data manually or with APIs

Your data sits securely in whatever platform you are using for I&R curation. So how does Yanzio's AI access it to make recommendations?

We have several ways:

Manually export your data in CSV, XML, JSON or another standard format and our engineers can import it into our AI systems.
Some systems allow you to schedule these exports and automatically send them to a secure destination. We support that too.
Best of all is if we can connect our Application Programming Interface (API) to the API of the platform you are using for either data management or aggregation, like the National Data Platform .
If none of those options work for you, with your permission we can also ingest your data from your public website.

If none of those channels are options for you, we still have other methods of getting your data over to us.

We can run our AI Assistants on your data whenever we have a fresh update.

That could be a periodic one-time review of your entire database. Or if we have ongoing access via an API connection, our AI Assistants can monitor and process records shortly after you've edited them, giving you feedback more quickly.

Make sure to ask your software vendor if they have an API connection to Yanzio so that you can enable our AI Assistants to work with your data on an ongoing basis. It's the best way to keep your records updated and adhering to your own standards.

Help us finalize AI tools for I&R, get free access

Tue, 23 Jan 2024 15:00:02 GMT

Seeking up to 5 organizations to implement AI shortly

We'll very soon be launching our first two AI Assistant products , the Style Guide Curator and the Taxonomy Curator. We've been building and testing them for several months and they are nearly ready for general availability.

Before we do, we'd like to do further refinement with a few more organizations.

Free access to Style Guide and Taxonomy AI tools

For those organizations that apply and are selected, you'll get at least 60 days of free access to these tools, which will rapidly give comprehensive feedback to you on all of your resource database records. We'll focus on adherence to your Style Guide, as well as reviewing your Taxonomy assignments for recommendations.

In return we ask for:

Permission to use your resource database solely for this testing.
Your feedback on the results that these tools provide.
Permission to refer to this work to the public, including on our website, blog, at conferences and other venues.

How to apply, by Wednesday January 31, 2024

Please click on the Contact Us button and provide the following:

Your name and email address
Your organization's name and website
Approximate number of Program/Service records in your resource database
Your permission to be referenced to the public
Goals you hope to accomplish by using these tools

Thank you, and if you have any questions or concerns, please feel free to reach out as well by clicking the Contact Us button.

Getting trustworthy results from AI - the "RAG Triad"

Tue, 16 Jan 2024 14:30:00 GMT

Three-step process to improving answer accuracy and trustworthiness

With the widespread adoption of AI coming this year, many people have correctly raised concerns about trust. How can we ensure the answers and actions it provides us are accurate, fair, relevant and trustworthy?

There are already a variety of safeguards and approaches and more coming. Today I'll describe the "Triad" applied when coupling your own data source with a Large Language Model, as I described in my last post . (The Triad was first proposed by an AI safety company called TruEra in late 2023.)

Three components of the Triad

When a question is posed to AI, this Triad approach proposes we test the results three ways before emitting them to a human:

Context Relevance
Groundedness
Answer Relevance

Humans do this initial testing but the AI itself can be trained to do a great job in most scenarios, with humans monitoring on an ongoing basis.

1. Context Relevance

Is the content that AI responds with, relevant to the query?

Suppose you ask an AI chatbot at a sporting goods website for "find breathing equipment used in underwater activities and provide a price" (an example used at a recent Microsoft AI conference in San Francisco). It has access to all the product details in the store's catalog.

Let's say AI first responded with wetsuits, fins, snorkels, personal submarines and scuba tanks. A pretty good start clearly understanding the importance of "underwater" in the query. It might have also picked items like bike helmets and basketballs because there was some relevance to the words "equipment" and "activities".

By filtering at this stage for Context Relevance, AI will remove "bike helmets" and "basketballs" from the results, leaving only items that are commonly used underwater. A definite improvement over the initial results.

2. Groundedness

Is the response supported by the context of the query?

For each and every "claim" made by the AI in the response, independently search for evidence that supports each claim, especially as it relates to the "context" of the query. Each claim must be supported by a highly relevant citation.

Continuing our sporting goods example, for the products remaining in AI's response (wetsuits, fins, snorkels, personal submarines and scuba tanks), AI must include a link to the original product details page found in the company's catalog, verifying the product name, type and price.

But "personal submarines" cannot be supported with a citation because it is not in the store's product catalog - they don't sell those. So it will be eliminated from the response.

And "fins" are out of stock and don't currently have a price in the listing, so those can also be eliminated from the response, because the user specifically said they are looking for pricing.

After this second stage of filtering, we can see the response is getting even more trustworthy. But we still have one more step to go in the Triad.

3. Answer relevance

Does the overall response, and every claim in it, provide an answer to the original question?

Before emitting the answer to the user, we perform one more audit of the results to make sure all aspects of it are relevant to the question.

In this stage, AI recognizes "breathing" in the query and it knows that only "snorkels" and "scuba tanks" relate to that part of underwater activities. It now eliminates "wetsuits" from the results.

A trustworthy, accurate response to the user

Now that the results have passed through all the components of the Triad, the user gets highly relevant, accurate and trustworthy responses to their initial question.

Applying the Triad approach to I&R

In this simple example, instead of looking for snorkels and scuba tanks, the same approach can be used for finding service referrals to help seekers. For queries like these:

"I'm hungry where can I get free food in downtown Houston"
"Summer camps for 10 year old boys with autism"
"What free day cares are open Wednesday afternoon"
"Avoiding eviction for veterans in San Diego"

...Yanzio's tools will pass them all through the Triad to ensure the best possible results.

In future blog posts I'll discuss some of the many other approaches and safeguards that greatly increase the trustworthiness and accuracy of AI.

How can ChatGPT "know" all the records in a resource database?

Wed, 03 Jan 2024 15:30:00 GMT

Converting your records into a format that AI can understand is the key

Large language models (LLMs) like GPT4 have been trained on many billions of documents on the Internet and other sources. That means it's possible that older versions of your resource database records are somewhere deep in its training data, assuming that you published your database on the web in the past.

But if you've tried asking ChatGPT questions as if you are a help seeker, and were expecting to be well-navigated to your most relevant resource records, you've probably discovered it does not perform well. Even if it did happen to find some decent referrals, the information like phone numbers and hours of operation could be months or years out of date.

That may be relieving to know that it's not as good as your talented staff and well-designed public website. But does that mean that while the whole world benefits from all the amazing things we can do now in ChatGPT, we in I&R are stuck in the "pre-ChatGPT" days for the foreseeable future?

Absolutely not! I&R can benefit now from intelligently applying AI to all the great data we have.

You don't have to give your data to an AI vendor to benefit from AI

That raises a conundrum.

For AI to do better with I&R work, it might appear that the LLM has to directly ingest and learn your current I&R data, like your resource database, now and on an ongoing basis.

But most I&R organizations are reluctant to hand over their resource databases to just anyone, especially to a third party like an AI vendor, who won't agree to limit how they use your data to serve just you. (Yanzio will make such an agreement with you, by the way. Your data is your data.)

And what are the chances that an AI vendor like OpenAI , Microsoft, Google or Anthropic would put this work very high on their list for any I&R who might approach them, while they are feverishly working on massive other opportunities these new technologies are unlocking?

Fortunately there is a way to use AI in close conjunction with your current resource data, without the direct involvement of an AI vendor. You just have to understand how to connect the two.

A brief, nerdy digression on AI: math and vectors

(Skip to the next section if the words "math" and "vector" may as well be a foreign language to you.)

Deep inside an LLM, it represents chunks of text as a "vector". If you took an upper-level math class in high school, you may remember what a vector is: an arrow pointed in a certain direction, with its length representing how big or strong the quantity is. Essentially it is a mathematical representation of a concept or physical phenomenon (like wind or gravity).

A simple example might be on a weather map, with a vector representing the direction wind is blowing and how strong it is. That would be a two-dimensional vector since it is on a flat map, like a piece of paper with width and length. A three-dimensional vector would also incorporate height, so our weather map now could also tell us if the wind was blowing upwards or downwards, too.

An LLM takes a sentence like this one, and converts it into a vector - but instead of two or three dimensions, an LLM like GPT uses 1,536 dimensions that can represent the subtle underlying meaning of the text, mathematically. It needs that many so it can differentiate sentences like "My favorite pet is a cat" from "George Washington was the first president of the United States".

In total, GPT4 has 175 billion parameters which are the elements of these vectors. That's a huge amount. All those vectors represent all the sentiments of all the data that the LLM has ingested and trained upon.

Speaking the language of AI: vectors

Instead of handing a resource database over to an AI vendor, raising the conundrum a couple of paragraphs ago, we flip matters around.

We convert your resource database into its vector representations. It's a highly technical process but all we're doing is making your resource database directly understandable by AI. We can do this on an ongoing basis so that it's always pretty current.

We keep those vectors in a safe place like a special database, called a "vector database", so that it does not have to be directly ingested into the LLM owned by an AI vendor. It's under your control in a private place of your choosing.

Doing this let's us have ChatGPT-style interactions with your resource data, allowing people you authorize to ask powerful questions like:

"What food banks are open on Wednesday afternoons that serve veterans and are free?" or,
"Show me all the summer day camps for teens that can be reached with low-cost public transportation and offer swimming with a lifeguard."

Whenever we want to ask these kinds of questions about your resource data, we do the following:

Convert the question into a vector (or a set of vectors) by using tools available from the LLM.
Search your private vector database for resource data that are highly similar to the question vectors. "Highly similar" is actually a geometric calculation done comparing vectors by their angles and lengths.
Take those "highly similar" results, choose the best ones, and interpret them back into their original text-based resource records.
As an optional final step, feed those results into the LLM and ask it to do some quality checking and summarization.

You retain control of your data

One of the main benefits of this approach is that you retain control of your resource database records, as well as any tools or user interfaces that make these capabilities available to your staff, partners or the general public. That means you also get to refine and improve how it works over time and keep the data as current as you like. And you have full access to the usage of the tools and data, which is vital for understanding the needs and demographics of the people you serve.

Does this really work? How do I get started?

I'm seeing very promising results using this process with real I&R resource databases given to me recently by 2-1-1's specifically for this experimentation. Using it passively or directly in the referral-making process is an obvious place to start.

But there are so many other ways to use this too, such as in training, reporting and assessing service gaps in communities.

If this excites you, please reach out to me to learn more.

Which AI is best for I&R?

Mon, 18 Dec 2023 15:00:02 GMT

GPT3.5, GPT4, Google Gemini, and more

Ah the simple days of late 2022. We really only had one Large Language Model (LLM) to choose from, and that was ChatGPT . Behind the scenes it was powered by GPT3.5 and the world marveled at all the cool new things it could do.

In early 2023, OpenAi released GPT4 with substantially more accuracy and clarity. You could use it with ChatGPT but only with a paid subscription, and with some usage limitations. Microsoft, by virtue of their partnership with OpenAi, then incorporated GPT4 into Bing, offering more people a way to use GPT4 in their Edge browser and elsewhere.

More LLMs enter the fray

In the ensuing months, other companies announced and released their own competing LLMs and tools to the public: Google Bard , Anthropic Claude , Perplexity and many others. Google just announced another new suite of tools they're calling Gemini .

Which one is the right tool for the helping with I&R work?

As with many things in life, the answer is "It depends." Work on fairly large sets of I&R data requires an LLM that can process thousands of records relatively quickly, with good accuracy and reasonable cost.

Working with thousands of resource and encounter records

First, to work on large amounts of information, like your resource database, or logs of past interactions with your help seekers, an Application Programming Interface (API) is needed. That avails all the skills of the LLM to vast amounts of records. Over the past few weeks I've been doing just that with resource databases provided to me by 2-1-1s, often processing thousands of records every hour through APIs as I fine-tune my work with LLMs and I&R data.

As of today only a few of the LLMs offer access via an API, like GPT3.5 and GPT4, and recently Google Gemini Pro. But others are sure to follow in the near future.

Good, fast and cheap?

Next, the LLM has to produce useful, consistent and high-quality results. We can tolerate some occasional odd outputs but generally need it to be reliable. I am measuring an LLM's performance on a task (perhaps rewriting a Program Description to better comply with a Style Guide) by seeing an improvement in 95% of the cases, which is 19 out of every 20 records

But it's a balance between quality, speed and cost. GPT4 is better than GPT3.5. But it's also 10 times more expensive and 3 times slower, based on my testing. Like most thing, you can get any one or two of "good, fast and cheap" but rarely all three.

And of course: safe and reliable

I am also focusing on ensuring that the results are safe and reliable. Our society is making a big leap into a world where it is going to do more of our work. We have to trust AND verify.

The right tool for the right job

My approach is to test the various LLMs for each different task I'm focused on with large quantities of records and evaluate specifically what each one does well. In one example, GPT3.5 does a good job for most of the tasks but fails to properly identify some "false negatives". For those instances, I double-check its work with GPT4 and either validate that 3.5 was right, or get an improved version. In this way I can blend different LLMs to get the best combination of quality, speed and cost.

Robots judging robots judging robots

Over time, I intend to incorporate other LLMs into the mix, as their APIs become available. In addition to determining which specific tasks each one is good at, I also plan to use the LLMs to start evaluating each others' work.

Getting great results depends upon working with data and experimenting with the different LLMs to figure out the best way to tackle the problem at hand. And then stay current with the AI industry because it is moving so quickly.

Stay tuned for more results in the coming weeks and months.

Can AI and GPT help improve I&R resource data?

Mon, 11 Dec 2023 16:30:00 GMT

Exploring AI's Potential to Enhance Database Accuracy: A Case Study with GPT and a US 2-1-1 Service

The advent of generative Artificial Intelligence (AI) has opened up new possibilities in many fields, including the meticulous task of database curation for Information and Referral (I&R) organizations. Among the various AI tools available, Generative Pre-Trained Transformer (GPT) models stand out for their capability to process and generate human-like text. Many of us have experienced this first hand in the 12 months since ChatGPT first released to the public, which uses GPT behind the scenes.

In an intriguing experiment, I collaborated with a US-based 2-1-1 service to explore how well GPT could apply the principles of an I&R Style Guide to improve real-world resource data.

The Challenge of Maintaining Database Integrity

Community Resource Specialist – Data Curators are tasked with the crucial job of ensuring that resource data adheres to their formal Style Guides. However, varying interpretations and human error can lead to inconsistencies, not to mention the time constraints that often prevent curators from addressing every record.

Enter GPT: A Tool for Data Conformance

To harness GPT's potential, I interpreted the Style Guide into a series of "prompts" (essentially, instructions) that GPT could use to evaluate each record. It's a blend of art and science, especially to minimize errors and hallucinations.

My focus was on the Eligibility field, which presented an ideal blend of complexity and variability for this test. The Style Guide rules were well-written with some qualitative guidance somewhat open for interpretation.

After feeding over 9,000 program records into GPT via its Application Programming Interface (API), I received scores indicating each record's adherence to the Style Guide, along with an explanation of scores and suggestions for improvement where necessary.

The Findings: A Promising Partnership Between AI and Humans

The initial results were promising. GPT determined that 45.7% of the records exhibited very good conformance. Another 34.7% showed moderate conformance, while 18.6% indicated a good bit of room for improvement. A remaining 2.1% of records could not be scored due to scant or no data - mostly in inactive records.

These findings underscore the high quality of work performed by human curators while also highlighting opportunities for enhancement. About half of the records could benefit from revisions to improve clarity and accessibility for various stakeholders, including 2-1-1 agents, help-seekers, data partners, researchers, and the general public.

The AI-generated scores, explanations, and suggested rewrites provide valuable insights into interpreting the Style Guide and prioritizing updates. While not flawless, these AI contributions serve as a solid foundation for further refinement.

Looking Ahead: Continuous Improvement and Collaboration

The experiment is ongoing, with further analysis comparing the performance of different GPT versions (3.5 and 4) and extending the testing to other database fields. I may try additional AI models like Claude, PaLM 2, Gemini and others. I am also exploring streamlined processes and tools that will make it easier for people to review and integrate GPT's suggestions back into the resource records in the database.

This exploration is just the beginning. If you're intrigued by the potential of AI to enhance data curation and would like to be part of this innovative journey, I invite you to join me. Together, we can responsibly start harnessing the incredible power of generative AI in a wide range of I&R activities, including database curation.

Using AI / ChatGPT to make referrals

Sat, 09 Dec 2023 19:46:39 GMT

Back in October 2023, I posed the following question to the Information and Referral industry. Stay tuned for updates on progress since then.

Is anyone interested in or already experimenting with using AI such as ChatGPT to help in the referral-making process?

I can imagine multiple ways it could do great work if it were properly "trained" on a resource database.

And this doesn't just have to be a "chat bot" - there could be cool ways to integrate it into the workflow of an I&R call or other self-serve tools on your public websites.

I've spent the last few months getting deeply into using it and integrating it with other software systems and I think it could be a fantastic addition to the operations of Inform USA member agencies.