This Week Health

Don't forget to subscribe!

Some good points in this article. I like it when I can bring you both sides of a movement. Hope you enjoy.


  📍 Today in health, it we're going to take a look at a wall street journal article. The AI revolution is already losing steam. I'm going to agree with it. I'm going to disagree with it and we're going to examine it because there's some things in here that we should understand as we move forward. With some of these AI projects and some of the projections coming from companies. That are planning to use it. My name is bill Russell.

I'm a former CIO for a 16 hospital system and creator of this week health. So the channels and events dedicated to transform healthcare. One connection at a time. We want to thank our show sponsors who are investing in developing the next generation of health leaders, notable service now, enterprise health parlance, certified health and Panda health.

Check them out at this week. Hey, this story and all the new stories we talked about, you can find on our website this week. One last thing it's June and June is lemonade days for Alex's lemonade stand. And so we continue to partner with them to find cures for childhood cancer.

And we are doing a bunch of really fun things. If you had a chance to check out our website, click on the ribbon and give today to be a part of. Our lemonade days, we're going to try to raise, I don't know, 25 or $30,000. I'm going to have to listen more closely when my team. Is telling me about some of these things, but essentially I know it's, I know it's a lot of money and we need your help.

So love to have you be a part of it. One last thing, share this podcast with a friend or colleague use it as a foundation for daily or weekly discussions on the topics that are relevant to you and the industry. They can subscribe wherever you listen to podcasts. All right. Let's dive into this article.

The AI revolution is already losing steam. And they say three things. One, the pace of innovation for AI is slowing. Because, once you've consumed the knowledge of the internet and you've trained your model on the knowledge of the internet, there's no other internet to go out and swallow,

so it's the pace of AI slowing. That's their point. Its usefulness is limited, meaning it can't replace as many jobs as we thought it can replace. And it's not helping as many users as we thought it would help. And the cost of running it remains exorbitant, meaning. That we are creating these data centers and just building and building to run this.

And it's a little bit inefficient. So let me hit on some of their points here. The rate of improvement, and this is straight from the article. The rate of improvement for AI is slowing in there, appear to be fewer applications than originally imagined for even the most capable of them. It is widely expensive to build and run AI.

So that's what I just covered. Let's go to the pace of improvement in AI is slowing. Most of the measurable and qualitative improvements in today's large language models. AI is like open AI is chatty PT. Google's Gemini, including their talents for writing and analysis, come down to shoving even more data into them. These models work by Jed digesting, huge volumes of texts. It's undeniable that up to now simply adding more has led to better capabilities, but a major barrier to continuing down this path is that companies have already trained their AIS more or less on the entire internet and running out of additional data to Hoover up. To train next generation, AI engineers are turning to synthetic data and we know that this approach isn't the best. I that's paragraph, that's paraphrasing a paragraph. AIS. Like chatty beauty rapidly get better in their early days.

But what we've seen in the past 14 and a half months are only incremental gains. Says Marcus, who is Marcus? He's referenced earlier in the article, Carrie Marcus, cognitive scientist. Who sold an AI startup to Uber in 2016? The truth is the core capabilities of these systems have either reached the plateau or at least have slowed down in their improvement. Further evidence of the slow down and improvement of AIS can be found in research showing that the gaps between the performance of the various AI models are closing. And this is probably one of the more interesting points.

All of the best proprietary AI models. Are converging on about the same score on tests of their abilities. And even free open source models like those from Metta and Mistral. Are catching up. Okay. I think that's an important point for us, quite frankly. We get into this. We get into this mode in healthcare that we're not going to build anything. We're just going to buy. And the, and we go out and we sign up for open AIS chat GPT through our Microsoft agreement or whatever it happens to be. The reality is if this is true, if this statement is true and it's worth looking into if open source models like Metta and Mistral are catching up to chat. GPT and Gemini. And they're not utilizing advanced capabilities then. Very interestingly, it becomes a somewhat imperative upon us. To understand what it takes to stand those models up internally, because if you can stand those up. Internally the free open source models, and they're not going to improve much over time. It represents a significant savings. And ability to roll it out across your entire enterprise.

So that's an interesting phrase. I would dig into that more. I'd also dig into this whole concept of pace of AI. Getting smarter faster, better. I'm not sure I agree with it because what we're seeing right now is this push towards AI models that can see and hear. And the reality is that when they start to see things, they will process things differently than if they just read things.

It's true of us when we see things. We process things differently. There's. There's just more nuance to everything that's going on. And so computer vision and the the inclusion of that into these AI models, especially right now, we're primarily talking about the generative AI models, but bringing computer vision into the generative AI models. Creates a different level of learning.

And so I'm not sure. I agree with the pace, not increasing. Definitely the usefulness is increasing as we can interact with it. In a more natural way. You get to this situation where we're interacting with our computers in the way we were designed, we don't even think twice about the fact that we were not designed to sit in front of a keyboard and mouse all day and type things in. We were designed to interact.

When we interact with people, We speak. We talk, we make gestures. We we interact with. We interact with data differently in the real world. But we think this is the way we were designed to operate as human beings. And we think this is the way. That the world must operate because we've been doing this for the last 30 years. The reality is I think the interaction. We'll change the way we. We go about. I don't know, utilizing the world's knowledge, if you will.

This gets back to, gosh, it was a long time ago. Microsoft had information at your fingertips. And the idea was information at your fingertips meant it was at the end of your keyboard. But what if it was just information? At your, not fingertips, but at your thoughts. So that we're essentially thinking some, oh, we already experienced this. When we sit. At a table now there's no.

Hey, I wonder how much Scottie Scheffler has made on the tour this year. I wonder how much has caddies made on the tour this year? It used to be, these would end up being 15, 20 minute back and forth with people. And now it's just look it up. He's made. $12.5 million or whatever happens to be.

I'm sure it's more than that actually. But we now answer those questions almost immediately with our phones. This is not something that was even imagined 30 years ago. And I would guess that the way we interact with computers is going to fundamentally change and I would love to see the keyboard and the mouse go away. And we have a situation like this, where I have a microphone in front of me. And I essentially sit in front of my computer and I interact with it.

Like I do a person. I ask a questions. I tell it, Hey, pull up a spreadsheet. I'd like to I'd like to graphically represent. The growth of our website based on blah, blah, blah, whatever the factors happen to be. And it puts some things together. And I say here's some of the numbers that I have, blah, blah, blah, blah, blah.

And it, it builds that stuff for me. And we're seeing that stuff come through on these gender VI models. Let's go on to their next point in this wall street journal article, by the way, the AI revolution is all. Oh already losing steam wall street journal. You can also hit it on our website as well. So it says AI could become a commodity.

A mature technology is one where everyone knows how to build it. Absent, profound breakthroughs. Which becomes exceedingly rare. No one has an edge in performance at the same time. Companies look for efficiencies and whoever is winning. She has from who is in the lead to who can kick cost to the boat. Okay.

I think they already recognize this, that this is happening. You have open AI saying, Hey, we're going to scrap everything and start from scratch. So instead of seeing maybe that you probably will see a chat GPT five. But whatever the next generation of this is, they're already retraining a model in a different way because they believe they can create something that has a differentiating factor. In fact, what they're saying is that gets close as close to anything we've seen with artificial general intelligence, which is essentially thought generating computer intelligence.

And they're even saying, Hey, you know what, the current models and how we're throwing things at them and training them. If there's a different way to do that, we're going to come up with a better model. And so they are exploring those kinds of things and it's improving on the original concepts, the the transformer concepts and whatnot. That are the foundation for all of these things. Let's see. I skip that paragraph. And they say this is happening already.

Some AI startups have already run into turmoil, including a inflection, AI, its co-founder and other employees decamped for Microsoft in March, the CEO of stability, AI, which built a pop popular image generation AI. Tool stable diffusion left abruptly and March many other AI startups, even well-funded ones are apparently in talks to sell themselves.

And what they're saying essentially here. Is that the large players are gonna have enough money to keep playing in this arena. But the small players, if it gets commoditized too quickly their investments are going to get lost. You're going to see significant downturns in their evaluations. And the small players will just. Cease to exist. And they go on their last thing here is, and by the way, I don't disagree with that.

Whenever you have a fast moving market, like I can go through the PC industry. And I can start naming like 15 companies that just don't exist anymore. Like back in the day. You would get a page in PC magazine, you list your computers and you had the gateway 2000. I think it was called gateway T.

It was definitely called gateway computers and they sold a ton of computers through PC magazine. And Dell was one of those original ones that was there as well. You had gosh, a zoo. Zeus you had I can't even remember all of them, but a Compaq for that matter compacts gone. You have a whole bunch of companies that used to sell computers. Direct to consumers, to businesses and whatnot.

They just don't exist anymore. That's what happens in a fast paced market, even a maturing market. The thing with this AI market is maturing so rapidly that some of these companies are going to appear like they're doing something. That is very interesting and. The next day, you're going to see somebody else doing the same exact thing. For less money.

And so if it gets commoditized too quickly, they lose their R D budgets. It doesn't mean that we don't benefit because it that rapid movement to commoditization. Means that we won't go. Bankrupt. Implementing AI technology. All right. Finally, Hey, I could become a commodity in mature technologies, one where everyone knows how to build it so forth and so on.

And we've looked at that the final one is today's AI remains ruinously expensive to run, and they just talk about the fact that. You have the quote from Sequoyah. Capital Silicon valley venture capital forms. Coya that the industry spent 50 billion on chips from Nvidia to train AI in 2023. But only brought in 3 billion in revenue. Again, not uncommon early on in the evolution of a new platform or new capability within technology. And that make no mistake about it.

This is a new capability. It's very interesting. And but it's not sustainable. And if it commoditizes too quickly, that money goes away. And then finally they talk about the fact that there's a lot of research around the fact that we are not seeing the kind of productivity gains there's narrow use cases for this slow adoption. Again, I don't think this is a out of the ordinary for newer technologies.

It does take a little time to figure out, Hey, where are we going to use this? And and what is the the value? And then. Propagating that across the entire organization. Let me tell you where I think the magic is of the the AI models that we see. I think there is magic in and layering on the ability to have natural language and there's. There's magic in the ability to layer on computer vision. I think the other thing is we are underestimating the combination of technology.

So the combination of automation. And AI, I think represents a significant benefit. I'm talking about non programming automation. So a no code automation kind of tools and AI. I think we'll create all sorts of workflows. Now it's challenging to manage those workflows, but I think that is going to be. The area where we see a significant move and opportunity.

And so if I were a CIO right now I would still be, I'd still be bullish. I'd still be moving forward very quickly. As I've said before, I'd be standing up. My governance team, I would be educating my governance team. I would share an article like this one. Not all the articles you share with them should be bullish.

Some should be. Bearish on the on the potential of AI. And paint as, as accurate as picture as possible. But then I would explore the different use cases and the different layering of technologies. That can lead to gains. All right. We're at 15 minutes, I'm going to end that's all for today.

Don't forget to share this podcast with a friend or colleague. And use it as a foundation for mentoring. We want to thank our channel sponsors who are investing in our mission to develop the next generation of health leaders. Notable service now, enterprise health parlance certified health. And 📍 Panda health. Check them slash today. Thanks for listening. That's all for now.

Thank You to Our Show Sponsors

Our Shows

Today In Health IT with Bill Russell

Related Content

1 2 3 267
Transform Healthcare - One Connection at a Time

© Copyright 2024 Health Lyrics All rights reserved