AI Lawsuits, Use Cases and More – Techstrong AI Podcast EP7

Hello and welcome to the Techstrong AI Podcast. I'm Amanda Razani, and with me today of course, is Mike Baard. How are you doing today?

I'm great as always. Good to see you. Good to see you as well.

Well, we have a lot of topics to cover, so let's dive in and I'll let you start, Mike. All right. Well, we're starting off with, apple is apparently bought a company called Darwin AI outta Canada.

And it seems like what the folks at Darwin AI are focused on is applying, uh, algorithms to computer vision, which kind of seems to dovetail nicely with, you know, what Apple's been talking about with these fancy vision glasses of theirs that they've been talking kind of getting people excited about. They're still super expensive, so I don't think they're gonna go mainstream, but maybe a lot of the use cases will be industrial types of applications, and Darwin AI will probably help with some of that. So folks are getting excited that, you know, Apple's figured out how to spell ai, at least because they've been getting beat up pretty heavily about, uh, not being in the game per se.

I'm not sure that buying one small company in Canada is, uh, gonna alleviate those criticisms, but it seems like it's a step in the right direction. Well, and I also think that, um, apple is always known for kind of working backwards, uh, and they're more the, the, the innovators that set their why first before they try to produce a new type of technology. So I think maybe they're not jumping on it as quick because they have to have that why and reason first, and, uh, maybe this is gonna be the direction they go.

Well, you make a good point about the fact that a lot of times it seems like we have like AI for the sake of AI going on, and there's a lot of what I would call computer science projects, but I'm not quite sure that they're actually gonna meaningfully result in some sort of business process. In fact, we see this all the time, right? Every day there's a story about somebody launched some sort of AI initiative that didn't quite live up to expectations, and so people had a pullback.

And I think, um, this is the year of trial and error, shall we say. Yeah, absolutely. And it would be really cool to see what they do with these glasses, because I imagine a future where everybody's just sitting at their desk with these glasses and they're pulling up all this information with a, like, touching the side of their glasses.

I don't know. And they can look into all these spreadsheets on their glasses. I don't know.

I don't know. We're all gonna look like Jordy on Star Trek. Right.

Alright, speaking of trials, speaking of trials, let's just jump real quick into this whole thing that's going on with Elon Musk and uh, uh, Sam Altman and OpenAI. It seems like it's getting a little outta hand, but the root of the issue is that, um, Elon Musk is accusing, um, Altman and OpenAI of violating the founding principles of the company where they're supposed to be for the good of humanity versus a commercial entity. Um, I'm not quite clear of how much of this is for real and how much of this is a PR stunt, but, um, Elon Musk, of course, there's emails where he's basically saying, I too want to make a commercial entity out of this thing.

I'm just mad at you because you did it without me. Um, and then there's also some sense of, you know, how much of this is just Elon saying, I too have an AI platform over here, and maybe this is an interesting PR stunt because the guy's good at pr. You gotta hand it to him.

Um, oh, absolutely. But Yeah, I dunno, I think it might be a mix of both. I, I, you know, he does kind of march the beat of his own drum, and I, I know he has strong opinions about a lot of things, and he did put a lot of money into OpenAI, um, apparently with the idea that it was going to be open.

Um, I guess he thought always open, but as we know, it cost a lot of money to keep this technology going. So it's not really a surprise. I mean, I think we were just talking about that on our TV show the other day.

Eventually, it's kind of hard to sustain without, um, you know, commercializing it a little bit. So, um, you know, maybe he is a, a little bit shocked by that because he did, um, introduce Agro and he, and it's open right now. So, um, is he planning to maintain it being open?

I mean, he kind of has the pockets to do so, so who knows? I don't know how this is all gonna play out because, um, I know this, all this focus on these really large language models, but in my mind, I'm kind of starting to think about it this way. It's like there will be interesting foundational models that people will build, and then we'll take subsets of that to create smaller tasks, specific LLMs, and then we'll orchestrate all these LLMs together to drive a workflow or process.

And that's important because the smaller LLMs will be trained on a narrow set of data, so they'll be more accurate and there won't be as many hallucinations and weird things happening. I mean, I don't know if you've noticed this, but um, you know, if you're writing a longer form story and you're using OpenAI about three quarters of the way through, it starts to lose the thread a little bit. And you gotta kind of go back and, and sort that.

And that's fine if I'm writing an email or something, you know, basic. 9997% of confidence that that thing's gonna work. So I think, um, a lot of this stuff is gonna be shown to be, you know, an interesting means to a larger end down the road, but we'll see how it plays out.

Yeah, absolutely. You know, this could be a a, a free open software he provides, but I'm sure there's some reason behind it, uh, you know, uh, to, to pull more interest into his company. Yep.

Um, maybe it, maybe, um, it, it's gonna, maybe it is a big marketing stunt, who knows? We'll see, You don't think that, that they're doing this outta the kindness of their heart? I'm surprised.

I I, I would doubt it. I don't think so, but who knows? All right.

Well, the nice thing is if this goes to court and actually like, you know, goes to trial, the what will be revealed will be interesting and it might just be the next great soap opera of our times. Yes, for sure. And everybody likes the drama, so.

All right, so moving on. Um, next up is the EU Act. So we finally have some, a little bit of guidance when it comes to the use of ai.

So, um, I think this is important. This is a wonderful, um, global, uh, move that's, uh, we're all gonna have to try to adhere to now. So what are your thoughts on it?

I think the operative phrase there is a little bit, um, if you read the document, I'm like, okay, it's awesome. We're saying we are going to enact regulations based on the risk level associated with whatever AI activity that we're doing. So if it's AI for, um, you know, some household chores or something, the level of regulations will be this low.

Um, if it involves, you know, our children or cars or whatever else, the regulations will be this much more stringent. Okay, that's great. Um, I would say that the road to hell was paved with good intention.

Um, this stuff is basically, you know, a bunch of lawmakers and politicians that don't really understand how AI is constructed and being implemented and at the pace it's being implemented. So I think by the time they got around, they were agreeing what the level of risk is and actually came up with a regulation to apply to that level of risk, whatever the thing was, is probably gonna be running for the better part of a year or more, and much harm will ens stew. So I just don't see this as a practical way forward and we just gotta find another methodology here, but at least, hey, we're talking about it.

So that's something. Yeah, absolutely. Um, and in, in the US government as well, so, so maybe we're gonna soon follow suit, right?

Yeah, well, you know, what will happen in the US is the standard operating procedure. Each state will therefore copy some subset of the EU framework and then decide that they wanna enact that law. Somebody will get mad and follow a lawsuit saying, I'm sick and tired of dealing with these laws, laws from 50 states.

So will the federal government finally act? And then finally it will, uh, work its way through Congress, which, you know, agrees to something these days about once every, what, six weeks that is their current record of process. So let's see, it's 2024, so sometime in 2026 we should have some watered down version of the EU Act that will be equally probably ineffective.

So other than that, it's great. Alright, well we'll have to see if you're, if you're a fortune telling and predictions come, come true. All right, so next we're gonna move on to, um, AI being used.

This is a interesting new scenario that we haven't really heard much about. So the global phishing watch did, um, some research using AI to see how many dark fleets are out there, and this is ships that, um, don't have their transponders on and, um, aren't being tracked. And apparently there's 75% of, uh, fishing ships that aren't being publicly tracked, um, which could be harmful to the environment.

Um, uh, there could be a lot of illegal fishing going on, and there was 25% of say, energy boats and transportation boats that also, uh, were undercover not being tracked, um, which is a little disturbing, uh, uh, imposes some risk because they could be working for countries like Russia, North Korea, um, Iran and such. So, um, what are your thoughts? This is a huge issue and it's been around forever.

There's, um, fleets, quite literally multiple fleets in any given time traveling around that are not having their transponders on. We don't know what's in those ships. We can assume that they are part of some sort of effort to get around some sort of sanction somewhere.

Um, this is how, you know, funds find their way into North Korea, wherever else that they may be that's solicit. Um, and so, you know, we kind of know they're floating around out there. We can guess that they're out there, but I think the algorithms will make it, um, easier to more narrowly focus our efforts to find those ships.

And then of course, once you find 'em, you know, you can actually send somebody who looks like law enforcement to board them and inspect them because, well, they're violating international treaties and laws and all kinds of stuff. So, um, generally speaking, they're up to no good. So, um, I think, you know, this is gonna be a great use case for AI and we'll solve the problem entirely and probably not, but, um, will it reduce it considerably?

I sure hope so. Yeah. Interesting how they were using the AI to, to sort through like billions of, of data and, um, and then using some satellites like the signals bounced off satellites within the last five years, and then using the AI to go through all that data to find this out.

So a great, a great way to utilize AI for, um, hopefully to crack down on this issue. There you go. Everything has a digital trail, you just have to know how to follow it.

Mm-Hmm. Um, and so next I think we're gonna talk about, um, another interesting, uh, use for ai, which we have been hearing more about using AI and technology for military training. Uh, so we're seeing, uh, some synthetic training with the use of AI to reproduce environments around the world, uh, the topography and, um, the vehicles that are being driven so that, um, they can get practice, um, uh, simulation practice.

Mm-Hmm. Yeah, I like this application a lot just because, you know, we don't really talk about it all that much, but the number of soldiers that wind up getting maimed and injured and sometimes kill during training exercises is substantial. And if we can reduce or at least make that, um, less onerous, then, you know, hopefully, uh, more of them we'll survive that training encounter.

Of course, you know, I don't think we're gonna be fighting the war itself in any synthetic form anytime soon, but who knows? I seem to remember some episode of Star Trek many years ago that dealt with that particular outcome or issue. But, um, ultimately I think that we are getting to the point where, to me, this is just another example of where we're kinda creating simulations of all types and, um, we're able to think about this stuff much more broadly.

Whether it's an aircraft engine that we're working on or some sort of, uh, you know, vacation that we want to take. At some point you can actually like, plan that thing out with a digital twin and see what it might look like or feel like the real experience might not be exactly the same, but you know, you, your expectations will be set better and that's generally a good thing. Yeah, absolutely.

Uh, when you were talking about your scenario earlier, I was thinking of that movie Ready Player One. I don't know if you've ever seen that. Mm-Hmm.

But, uh, yeah, I think, uh, we're nowhere near that yet. All right, cool. And there is one other thing I wanted to talk about, um, and it was this stuff that, uh, Atlassian is doing where they have, um, figured out how to kind of embed videos into our asynchronous communications using AI to make it easier for mere mortals to actually create these things.

And I think what they're getting at is that, let's say the example that was brought to mind was if I wanted to, uh, send you an email that said, we have an issue with this, that, and the other, I can now also attach a video that I created easily. I don't have to be a video expert, and then the Atlassian software will help me edit that and shape that and take all the ums and ahs out and kinda all the flubs that go on. And, um, so I look reasonably professional when I send you this video that says, see here, this thing in particular is driving me crazy, and how do we fix this?

And you can see it, rather than me trying to describe in text what it is that you know, is never gonna fully, uh, give you the, the level of, uh, insight required, understand what it is I'm trying to get at here. So I don't know, what's your sense of, um, do you think video will come and, uh, an increasingly important part of our asynchronous email communications? Oh, it certainly is.

I think more and more we're seeing people turn to video. They, they wanna just see and, and get the information quickly. Um, and, and so it's becoming a, a popular method is video.

And I for one, would love if AI would take out all my ums and ahs because I know I struggle with this. Well, hopefully we can all be your own little movie directors in the future, right? So you can create a video.

I mean, we're talking about using Gen AI for text, but, um, there's no reason why wouldn't apply to video. You. The latest versions of the LMS from OpenAI certainly are down that path.

And I think we're going to get to the point real soon where we're all gonna be, you know, here's my, uh, latest documentary for Oscar consideration. And I don't know, we'll see what happens, but, um, mm-hmm, uh, can we all wind up competing with each other for the quality of our videos and our peers? Interesting times.

Yes, it is fun. All right, well, I believe we've covered all the topics of today. So that's it.

And thank you for tuning in and stay tuned for next week. Thanks everyone, and thanks Mike. All right, guys, take care.