Advancing AI: Apple & Microsoft Breakthroughs in Coding and Healthcare | TSG Ep. 875

Hey, everyone. Is Apple back in the AI game? You're watching Textron Game.

Hi everyone. Happy Tuesday. It's Alan Shimo for Techron Gang.

I hope your Monday went well after that long holiday weekend. You know, Mondays are always tough when you come in off a three day weekend, but now it's Tuesday and you really have no excuse. We're in the swing of things this week.

We actually have a tech field day, or networking field day, I guess is the proper term, right, Steven? That's right, that's right. Yep.

Coming up tomorrow and Thursday, which of course we'll be streaming live here on Tech Trunk tv. Might as well introduce our, our, uh, gang members today, and then Steven, I'll do you less and you can talk a little bit about networking Field day. But joining us is, uh, the one and only JP Morganthal, the Dean, Mike Ard, and of course, Mr.

Tech Field Day himself, Steven FoST. Steven, what's up with Networking Field Day this year? Yep.

So this is, uh, the second one of, uh, 2025. Uh, I think we might even have a third one. Um, yeah, we've got hedgehog coming back, uh, c packet of V and, uh, probably the biggest, uh, news maker of the, of the crew will be HPE networking coming in, because, you know, they just got approval to purchase Juniper.

So I can't wait to hear what HPE Aruba Juniper have to say at I, I think this is their first public appearance following the, uh, announcement that they got approval. Cool. Cool.

It'd be interesting, and as I said, we'll have it here live, uh, on text, on TV after the gang on Wednesday and Thursday. So let's turn to Tuesday's news. Mike Apple made news in ai, and it wasn't necessarily negative.

True, who would've thought? And this is our second day in a row talking about AI coding tools. But Apple is using diffusion techniques to write code differently and maybe more efficiently than humans do.

And it's an interesting, uh, tape because I gotta wonder if the way we write code and software's gonna fundamentally change in the age of ai, but we'll dive into that in a minute. Jp, what's your take on what's going on here? 'cause it doesn't seem like we're writing software the way our grandparents did.

For sure. There's a, there's a lot of things that I'm questioning about, uh, how we interact with these new creations. Uh, I actually wrote about one of these things recently with regard to resumes.

Why are we still using an 18 hundreds, you know, created construct to give to a machine to analyze whether the person is a good fit or not, right? It, it certainly can analyze individuals on so many dimensions. We need new inputs.

And the same thing with how we think about coding is, you know, as a person, I'm coding, I, I need to think a little structure. My brain needs the structure, the top down, right? And it, for me, you know, it starts out, I can, uh, create a, a, a, uh, a framework or a skeleton, and then I can put comments in, this is what I wanna do here.

This is what I wanna do. And then slowly and surely start to add more. And eventually my mind goes, oh, I've seen that before.

Let me refactor that so that I get reuse, right? So my brain's got background processes going on, analyzing as I'm coding, right? But the AI doesn't need to do that.

AI can kind of like take a high level snapshot and say, this is kind of what it looks like. And so, and it generate version one. And in the time it took me to write my code, which probably would be, you know, an hour to two hours, it, it, it can do hundreds of thousands of generations reiterating on that call code, making it better, right?

And so that's what's happening with the diffusion model versus the next token model, the diffusion model saying, yeah, let me start with, uh, you know, I, I think it looks like this. Nope, that's not right. Sh maybe it looks like this.

No, but each time it's getting clearer and clearer. Remember the old days when we had bandwidth issues and maps used to come in like that, right? You'd have first, you'd have the very, very coarse grain, low res, low number of data.

And then over time, depending upon, you know, the, you know, how much, how, how, uh, clear you needed the image to be the next you, you know, the next level down and the next level down. Each one being more and more data. That's exactly the diffusion model and how it works.

And to see it applied to code is interesting. I I, I do think that the other side of the story is, where did Apple come outta left feel with this? What made them, what made them move to this, right?

What do they, uh, so obviously Apple has been quiet. They're certainly not out in front making a lot of noise like some of the others. But, you know, they have done, you know, had Siri for years.

They have had, you know, an investment in, you know, integrating AI into their operating systems and phones, and uh, uh, and clearly they're on a path where they're doing research internally. I guess Apple just doesn't feel a need right now to be a noisemaker, right? To them, it's more about, it's a great tool.

We add it in for people, our users, we use it ourself to help be more productive. Right? Now, we don't need to demonstrate that we're, you know, that we're king of the hill.

We're not gonna fight the king of the hill battle with Anthropic and Microsoft and Google, right? We don't need to be there. That's not our game.

Our game is our devices. Our game is our operating system and machines and, and, and, and it's really user experience. Here's the thing that I kinda wonder about as I look at all of this.

So, we're finding a more efficient way to write code using AI agents. And I have to wonder, if I look back in time, we have all these programming languages that we created so humans could interact with machines, look at Java and everything else. They're higher levels of abstraction.

Steven, if I look at this, well, the AI agents at some point just decide that they're gonna create a more efficient programming language to write code. Maybe even, who knows, it'll be an assembly or something, and we're just gonna, you know, move to a whole different software era. And I may not even understand how the software is written.

Well, I certainly hope not, uh, because if we don't understand how the software is written, then it sounds like nobody does. Um, I think it's important to remember, as JP was mentioning here, that this is not some kind of, uh, I mean, this is fundamentally some new technology because they're using this diffusion concept where they essentially continually iterate on the entire section or the entire, uh, sub-routine, uh, rather than just predicting the next token, which of course was completely doomed to failure when it came to producing, uh, high quality code. But even so, um, I, I still don't know that I would trust, uh, blindly AI to spit out code that no one looked at and no one could review.

Um, and I hope that nobody else would either, though. I guess vibe coding is a thing. Um, it, it's maybe not a good thing if you want good code.

Um, you know, but, but you do bring up an interesting aspect here, and that's the different languages. I mean, we've heard of, for example, um, COBOL and FORTRAN code being improved dramatically by ai, uh, especially in terms of AI documentation and, uh, AI tuning. We've also heard of, um, you know, the, the questions about Apple's swift language.

There's been a lot of talk in the Apple community about whether Swift is really going to succeed because Apple just has not been able to put, you know, despite the might of the company, they haven't been able to put enough, uh, you know, development effort behind it. In fact, I wonder if perhaps the result of this paper is that it would be used, this technology could be used to improve swift. But one of the challenges there is that there just isn't a lot of examples.

And since ai, uh, as it exists, large language models are just basically, uh, they ingest and then disgorge, uh, tremendous, uh, amounts of, of, uh, text based on what they've seen. The fact that there isn't that much swift code to train a model on means that it's harder to get a model to spit out swift code, but maybe this diffusion technique works better. That, that was where my mind went when I saw this, uh, announcement.

Uh, what about you? Well, I, well, let me just answer Steven's point, because I thought about your issue with, with amounts to explainability of code written by an AI that we don't understand. But maybe I'll just go to a different AI agent and ask it to explain what the AI just did to me in a way that I can understand it and therefore solve that particular problem.

And then we can be more efficient. And if I need to check up on an AI agent, I'll just have another AI agent do it. That's the theory.

Just just a clarification of a point that was made earlier. They have come up with their own language. The only reason they continue to produce a program is structured programming language is for us, so that it's a means of communicating with us in a way that we can comprehend what they did.

And I can tell you, as somebody who's been working, you know, heavily over the past month with, uh, you know, AI generated code that, you know, there's a lot I need to go back and tell 'em, like, nah, nah, you didn't get this right. Go back and do this again. Right?

It really is like working with a junior programmer. So I've got a bunch of thoughts here. So first of all, this particular instance, jp, you're right, it is sort of how a map got used to get kinda filled in, but it, it's more akin if you're ever watched like how chat GPT draws a graphic.

It's layers and layers and layers and layers, and, and when it re and when you tell it to make a change, it can't just like, take a layer off, it really kinda almost starts over. So it's, it's a very, it, you know, it's not a linear way of, of drawing a picture, nor is it a linear way of writing a code. But that being said, you know, from an anthropo, anthropo, anthropological perspective, to your point about if we let you know, if you don't have a human do it, then you don't know.

And we won't know, you know, if human history's full of this, there were probably good reasons. There was good reasons why Jews adopted kosher food, right? Don't eat shellfish, don't eat pork, because these were foods that actually went bad first, right?

They were the first foods to spoil. So there was probably a good common sense, real reason why shellfish was not kosher. But over time and refrigeration and everything else, that's no longer, you know, a viable reason not to eat shellfish.

But nevertheless, this is the way we do it. And so it's still not kosher. I suspect we may have a similar thing with coding, just because humans did it not, you know, and we're looking at what AI's doing.

We're saying, wait a second, it's not kosher, right? Well, because they didn't have, we didn't have refrigeration then, or we didn't have the ability to look at it in a non-linear way, in parallel process and do all these things. It doesn't make it wrong.

It just makes it different. And, and so over time, yes, it's a junior programmer today, but today's junior programmers are tomorrow's cracker jack programmers. And the same will be true here, and it'll do it its way, and we may just all go along for it because it's kosher, right?

In terms of new, in terms of new technology, though, let's not forget what Apple did here was evolutionary not revolutionary. They built on top of, I think it's an Alibaba Kwan, Yeah, it's a good library there. I like one.

And you know, I don't know how well that will sit with the rest of the world. People are liable to say Alibaba has some sort of backdoor into it, or can sabotage it, or who the heck knows? It's open source, It's open sourced on hugging face.

Anybody wants to review the model and review the model. So, absolutely. Well, but look what we saw with Deep Sea, right?

That was open source too. Uh, it's from China. It's tainted at some level with a certain segment of our industry and population.

But give Apple credit, you know, a lot of people don't realize the real strength of Microsoft is their channel and specifically their developer channel, right? It's huge. Apple has an okay developer channel.

You're right. Swift hasn't become the, the standard that everyone thought it was. It would become, but look, maybe this is a way for them to piggyback and, and, you know, leapfrog over in, into something here.

I it Is actually kind of more revolutionary than you're giving it credit for sure. They're using an existing model, which actually, as JP mentioned at the top, I think that's a great idea. And it shows that Apple is not trying to build their own foundational models.

I think that's a wise move for them. Uh, the last thing we need is yet another fin foundational model. I think the, the revolutionary thing is using this diffusion technique instead of just predicting the next token.

And, and that could result in a really novel way of using this technology and, and may produce better results. Time will tell. Time will tell.

Alright? Yeah. And, and I think the applications will be better because one of the issues that humans have is thinking about processing things in parallel is difficult for a human.

And you have to kind of structure that maybe easier for machines to build applications where I've got 10, maybe hundreds or thousands of processes that are going along simultaneously, which would be much more interesting. There are, there are considerable, uh, performance, uh, improvements that you can see when using diffusion over linear. Uh, it's quite noticeable.

You know, you can really see it. Like if you were to use something like a VS code and hooked into, you know, one of those coding agents, if, you know, you can watch it build the serial code, and it's, it's kind of slow. And, you know, this thing, this is just in the background, re doing iteration every iteration of mutations of, you know, what was it, what we, we learned during COVID, right?

The thing how viruses, how many generations viruses can create in a short period of time, right? It's, it, it, it's at that level of mutation. My question is it, is it more or less energy efficient to do this?

Because by doing things in parallel, as things change over here, they gotta change over here. And that means you gotta redo stuff. I wonder, I, and I don't know the answer, I'm just speculating.

I wonder if it's less energy efficient to do that. That might be something we look into. Anyway, we're, we're outta time on this segment.

Let's take a quick break and come back and, and talk about some more here on Textron Gang. Discover Textron Group, the epicenter of tech innovation. We are your go-to for reaching IT leaders and practitioners worldwide.

Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more. Join our satisfied clients.

Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Textron Group. Hey, folks, we're back and we're talking about AI a little bit more, but this time it's a use case.

'cause everybody's arguing about what's the return on investment on ai? Well, how about just living life longer or maybe even saving a life? Microsoft is talking about how they've come up with a way to use AI that will do diagnosis and other tasks better, more efficiently than a human physician would.

And as we all know, you know, doctors need all the help they can get. Alan, what's your take here on what's going on? Is this a, an an early primary example where the ROI is just too big to ignore it?

I don't know if doctors would agree with you, Mike, but, and, and look, it, it's refreshing to talk about this when it seems, you know, all we talk about with AI is its ability to generate code, right? Or help make code better. But sometimes we can't lose, not sometimes we can never lose sight of the fact, really how disruptive AI can be in so many different verticals, not just it related.

This, this one, here's a perfect example. You know, it, we, we've sort of had it without the ai, but the AI is supercharged it. You know, you have a computer with a collective medical knowledge of humanity over thousands of years, right?

And the ability to analyze and sift through many, many more cases, many, much more information than even the human brain can. It was only a matter of time until someone puts this out and says, Hey, when it comes to at least diagnosing, I'm not saying doing surgery. I'm not, you know, talking about that, but diagnosing based upon symptoms and information, this is, you know, this is a no brainer in my mind.

This, this, this, yeah, this, this was gonna happen. And, and let's not, I don't wanna pick on doctors, it's not their fault, but how can you compete? And, and quite frankly, it's gonna be the same thing with lawyers and, you know, accountants, anything that, you know, the, the body of knowledge an AI could get its head around and tabulate better than a human, quicker than a human, it's going to happen.

So, you know, I say bring it on. I mean, as Mike, as you said, I think makes, it, makes for better diagnosises will save people's lives. I think also in the long term, with the shortage of qualified doctors, I mean, for all of us, right?

When you go to the doctor's office now, how many of you actually see the doctor versus the registered or, you know, the, the, No. Yeah, somebody comes in, the doctor comes in at the end just to sign off. Yeah, because it's their team on the, uh, on the paperwork, nurse Practitioner and everything.

I mean, this is, thank God could, couldn't have happened, you know, not soon enough. But think about it's, uh, uh, the, the input to this process, right? I mean, uh, that overcome silos within the medical profession, whereas, you know, I, I'm gonna give it your x-rays, so you have the radiologist, I'm gonna give you the symptoms.

So I have the, you know, the general practitioner. Um, and then I'm going to say, you know, based upon, uh, the symptoms I provided and the radiology, you know, uh, images that I provided, you know, give me the top five likely causes for this person's condition. And, uh, and the, and right now that would require a collaboration between multiple doctors, which are very hard to schedule.

Anybody who's ever gone through any kind of, you know, operation or any kind of medical examination across multiple practitioners knows how long they sit and just wait around waiting for these doctors to be able to schedule their time together to get a, uh, you know, a, a conference on, on the individual's state, that all of that can kind of be mitigated to some degree. And you, you know, the AI can push back out to each of these individuals and say, you agree with this or not, right? And then it can be kind of a consensus vote, uh, or vote voting type process that allows the person to, you know, the procedures to move ahead or hold up.

Mm-hmm. You know, this is a survey of one, I gotta say. So it's usually just me, but, and maybe it's my perception, but you know, I've reached a certain age where, you know, I'm losing friends as we all are.

I imagine, and it seems like there's a current theme that keeps coming up, is that a lot of these folks, it just took a long time to get diagnosed and for people to understand what they had, and then they didn't get the treatment going in time as a result. Yep. Now that's, right.

Now I'm seeing that. I don't know if everybody else is, but I feel like, um, if AI makes a difference on that, you know, that's, that's probably, you know, no vote prize material in my mind. And it, it, it's important to note Microsoft isn't announcing that they're using AI to provide diagnosis here.

I mean, that's been done for a long time. What they're announcing is that they've come up with a system that simulates the way the doctors diagnose medical conditions through sequential analysis of symptoms, uh, testing, and importantly, by getting together a group of, I guess you could call them independent minded, um, uh, diagnostics to determine what the most likely cause is. I mean, anybody who's watched a medical program on TV has seen the scene where the chief attending says to all of those young, handsome, probably doing naughty things, doctors around the patient and says, okay, what do you think is wrong?

Now? What do you think is wrong? How about you, you know, how do you build on this?

That's what Microsoft is announcing here. That this, uh, MAI diagnostic orchestrator is essentially that chief resident and the people that he's talking to, people in quotes, isn't, uh, handsome residents. It is in fact, uh, existing ais.

And so that's really what they're talking about here, is that they're, they're using all of these off the shelf ais, they're feeding them, uh, the known the facts as of now about the patient, about the patient's condition, about test results and that sort of thing. Having them all respond with what they think is the problem, and then iterating on that. And I think that's a really interesting idea, because like I said, they're, they're duplicating what doctors actually do, rather than just throwing it at chat GPT and saying, what do you think chat GPT?

That being said, none of these models are actually medical trained models in any way. They're just large language models. They're just spewing out the next token that statistically comes up based on the inputs that they get.

And so, you know, I mean, they're not doctors. Maybe they can diagnose things better, and that's certainly better. But once again, we're using AI in a cool novel way, but in a way in which maybe it's not the best for.

So, I, I, Steven, I think it's just a matter of time until you have specialized medical LLMs. I mean, that, that's not gonna be a great leap for me. The great leap is when do you let them prescribe remedies, whether it be, you know, uh, medicine or, or, or therapy or whatever, right?

That there, because it's almost like that's the point where human has to look at the code. Well, before I give someone some prescription that may or may not help them or hurt them, you know, I probably, I personally, and maybe it's just 'cause I'm old, would feel comfortable having a human look at that. But again, it could become kosher in another couple years, right?

That, hey, just the AI makes the diagnosis and gives you the, the, the, the prescription based on it. People are typing in their symptoms into various apps these days. Then a, some sort of doctor is allegedly reviewing that, and then they're getting a prescription and it pops up at CVS.

So, uh, for all I know that AI agent one day is gonna review the symptoms because the doctor doing the review of the symptoms is just statistically make it a guess. I mean, how many of you guys have done I'm on telehealth, I what? Done to telehealth, telehealth?

Oh, I hate telehealth. Yeah, But you know what? I, when I'm on the road and I get a, a bad cold or a sore throat or something, I use telehealth.

I've done it in Europe, I've done it at, you know, on the road here in the us. And I'll tell you the truth, I don't know if that's a doctor, a, a, a, a, a practical nurse, you know, or, you know, whatever the term is, I apologize or fraud. I know it is an ai, right?

But the last time I did it, the, the, the, the person on the other end said, well, let me see your throat. Can you turn your camera on your iPad into your mouth? And I did.

Ah, he said, oh, yeah, I see it's ratted. You know, they prescribed something. Who knows?

I, I, I know that there's a test going on within a chain, uh, pharmacy chain, the u the uk. Now the UK's a little different 'cause they have the NHS, which is the socialized medicine, right? And as an attempt to offload some of the traffic that's been going to NHS centers, they allow now these pharmacies, uh, and for particularly the pharmacists to diagnose a subset of conditions and give a certain, uh, um, medications based on those conditions.

And it's like seven. And it's really lightweight, nothing life threatening or anything like that. But what happened was all this traffic started now piling up, like, you know, waiting for the pharmacist to get the diagnosis.

And that's holding up things in the pharmacy from running. So, uh, uh, this particular test that they're running in one of these chains is that the workers in the pharmacy where, you know, get a, uh, a, a handheld assistant, and it, it's listening to the person talk. It's listening to the conversation, and the person reads the prompt to the, to the individual and says, you know about symptoms, do you have this?

And then the person responds, and then the AI tells it, okay, the next question should be this. And it listens. And, and it gets to a point where it makes a recommendation for the person.

And if it's within a yes, you're in the subset, we can recommend this. This is what the recommendation should be. And the pharmacist just has to sign off on it, and then it, it's done.

So it, it's being used to offload traffic from, you know, critical, uh, patient care. My father used to say, whatever you do, don't go to the hospital. 'cause that's where all the germs are.

That's where the same people are. You know what though? But let me, so my my oldest son graduated law school in May, and he worked in the legal technology lab there at Suffolk University Law School.

He worked on chatbots, two kinds, chatbots for lawyers, chatbots for civilians. The chatbots for lawyers basically is you uploaded a fact pattern and it spit out your pleadings for you, your, your interrogatories, your bill of particulars, your complaints, everything just crazy. Boom.

And that's great. It's gonna put lawyers outta work, but it's great. But for civilians, it was really, and the one, the one my son worked on was for landlord tenant court, 75% of tenants can't afford a lawyer.

And so they go in the landlord tenant court unrepresented. And as one might expect they lose, well, most of them haven't paid their rent, so they, they kind of, you know, it's a for, but there's things they could do to delay being evicted too. Sometimes the landlords have an agenda why they wanna evict him, and they don't accept payment or, or whatever.

And this chat bot really, I mean, it did yeoman's work. It was, it was actually adopted by the state of Massachusetts. They're using it out of the lab at his law school there to really help people navigate the legal system, put pleadings in, put in extensions of time, work out re uh, uh, litig mitigation, you know, arbitration kind of things, rather than getting put in the street.

And so, again, guys, I'm telling you it's novel now, but today's novel ideas are gonna be common stance a lot sooner than we think. I I think this thing with, with the doctor, uh, with diagnoses is gonna be standard within two to three years. I'm looking forward to that new metric that comes out.

It's called AI kosher, AI Kosher for ai. Now, that's a, maybe I could be like the council that gives it its stamp. Uh, all right, let's take a break.

We'll be back here on textron act. com is the leading resource for news analysis and education on challenges facing the cybersecurity industry. com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats, and more.

com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more. com.

Home of Security bloggers network. Hey everybody, it's Tuesday. And well, we always talk about the environment in some form or another.

So let's not make this one an exception. Steven, it's interesting times. There's the big, beautiful Bill has kinda reduced their thinking about at least reducing some of the credits for clean energy.

It's not going into an immediate effect, but as I turn around, I also noticed that we seem to be making a lot of progress on clean energy lately, and there's about a million announcements, including ones from just about every one of the big cloud service providers. But what's your take on what's going on here? Are we kind of solving a problem even though the politicians don't even know what the problem is?

Well, I, I think there's a couple angles here. I mean, as you mentioned, I guess we could start with the, uh, the bill. Um, it has indeed changed some things related to clean energy.

Uh, I should note that some of the worst parts of the bill that were proposed by the House and the Senate have actually been removed. Uh, there was a huge tax on wind and solar that was injected in there in the name of National Security. Uh, essentially they were going to put, uh, massive, you know, 30 40% tax on wind and solar projects that relied on, um, products that, uh, uh, physical components that came out of China, which would be, you know, all of them.

And so that would've really caused a problem. Uh, that's gone, uh, that didn't get through. But what did get through is a phase out of credits for, uh, tax credits for wind and solar projects starting at the end of 2027.

Many of, uh, the provisions of that bill, by the way, um, don't happen this year or even next year, and happen, uh, later, uh, for political reasons. Uh, the phase out of the wind and solar though, um, gives us, you know, 12 months to start these projects and another 24 months to actually reap the credits for them, which frankly is, um, surprisingly good news for the industry because it means that they've got some time to still get some of these credits. And frankly, by that point, uh, I've said many times on the show that wind and solar is already so much cheaper than other forms of energy that the horse is kind of out of the barn in terms of, of tax credits.

Maybe we don't even need tax credits to, uh, push these projects since the projects are so financially viable already. Uh, we certainly don't need a 40%, uh, excise tax on top of them that, you know, would cause problems. But, uh, frankly, just letting them run on in a free market is probably going to, uh, cause them to grow, uh, even without that.

Um, there are other elements, by the way, of the, of the bill in terms of, um, renewable, uh, fuels and hydrogen and things like that. But, but let's focus on wind and solar. Uh, as you mentioned, there's other news here as well.

Um, once again, we have, um, major hyperscalers talking about adding wind and solar energy for data center. Uh, I have questions about that. I did some research into the background of some of these things, and, and it, and it all kind of, um, comes out.

I, is this all just sort of propaganda from the industry, or is this actually news? Because the fact that meta is gonna be buying all this wind and solar energy, the projects that they're buying from aren't located anywhere near the data centers that they're going to be actually deploying this power in, which means, I mean, it's fungible, it's a grid. The, the, the electrons don't have to come from this particular solar panel.

But I will say that, you know, in looking into the background of these things, these projects that, that, that meta is gonna be leveraging, for example, they were built in places like Ohio, Arkansas, and Texas with promises to power homes. In fact, the Ohio one specifically says that it's going to power 46,000 American homes. Well, that energy ain't gonna be power in no homes.

It's gonna be powering AI data centers. And I'm curious what people will think once they see these big contracts come through for these big projects that were supposed to benefit Well, states and people, rather than ai, Maybe the metas or the world will pick up people's consumer bills as part of this. What do you think?

Snowballs chance, You know, interesting use of ai. What, what, what, what are, what are these, when, when a bill comes to the floor, especially these big ones, what's the one complaint representatives always say, who has the time to read 2000 pages in order to, you know, affirm what I'm voting for? Right?

I have a limited amount of time to, to get my, you know, to, to the vote. And I gotta understand all this. Well, you kind of don't anymore, right?

Just hand the thing off to, to Gemini, throw it in a notebook, lm, and say, I'm a representative for this state. What do I need to be wary of? Right?

And get your list back instantaneously. Now you know what you're voting on. Now, you know where you know what you're looking at.

Uh, I think it, i I I keep saying I'm ready to hand this government over to Claude because I think it could do a better job, but either side right now, I, I, I think a bunch of monkeys on a board might be able To do. Yeah, I, Steven, I agree with you, but, but a couple things. So Steven, you're right.

Thank God they, they didn't go ahead with the, the Chinese tariffs that would just kill the industry. And I, I think even as we sit here today, wind and solar could go toe to toe. If you took off some of the inherent bias towards fossil fuels in this country, especially, right?

I, I think we're getting to that parody level right now in a couple years more. But in a couple years you might have a new administration. And you know, the great thing about legislation like this is what's done can be undone.

The Supreme Court can change who, you know, things. This isn't necessarily forever, but here's the thing. These hyperscalers say what you want about them, but they're smart money.

They're smart money, and they're making their bets. They're talking, they're betting with their money, right? They're talking with their pocket books, and they're betting on renewable clean energy because they know, they know that that's the future for them.

They know that's the only way they're gonna do it. Is it gonna be at the expense of consumers? Steven, perhaps you're right.

And that's a shame, Jp, let's bring this full circle and close it out before we run outta time. Um, might AI agents not write more efficient code that would result in less energy being consumed by infrastructure at least more efficiently? And maybe we might solve our own problems?

'cause I like it when technology solves its own problems. I, I think it has a great, tremendous power to optimize its own, you know, uh, approach to the way it's used, right? You we're getting to the point where, you know, with the right observation tools, right?

Everything is data. With an, with an LLM, if, if it hasn't seen it, it doesn't know it, right? So it needs a feed, it needs a feed of data given to it that says, here's your power consumption model for, you know, the past six months, you know, given the body of work that you've been working on, you know, how, how would you optimize this?

You know, where do you see opportunity for improvement? And I think it can analyze it and give you back a, you know, a rudimentary set of steps to say, well, you know, I think if you know this, this, this, and this happened, you know, I would've used this many less megawatts. Okay, but who's feeding that?

Nobody's feeding that data. Nobody's capturing, you know, megawatt usage yet. And you would think that an anthropic or, or somebody like that would be like, you know, now we're talking almost like manufacturing levels.

If I save a penny, creating a car, it, it works out to, you know, millions of dollars over the long run, right? You're getting to that level where these guys are, are, are adding so much volume and, you know, there's only so much compute available for them that it, it's gonna be become a requirement to say, well, we need to do more with what we have. We can't just keep adding endlessly.

Agreed. Guys, I need to end this one right here. Jp, Steven, Mike, thanks for joining.

As usual, we have a full text on TV schedule immediately following today's event, and a reminder again tomorrow after. Actually, Steven, what time does that start tomorrow? Yeah, Alan, the tech field day presentations start at 9:00 AM Pacific time, uh, Wednesday and 8:00 AM Pacific time on Thursday.

And, uh, basically continue throughout the day. So they'll be on, well, uh, on Wednesday. It'll be about an hour after we end up here.

And on, uh, Thursday, it'll be right after the gang, so there you go. Alrighty, until then, everyone, this is Alan Shimmel for Textron Gang. Thanks for joining in.

Take care.