AI Skeptics, Software Intelligence, Deepfakes and More – Techstrong AI Podcast EP27

Hello and welcome to the Techstrong AI Podcast. I'm Amanda Ani and with me today is Mike Ard. How are you?

I'm well. But it seemed like the world is a vulnerable place this week That it does. And with that we have lots of articles to share with you in regard to ai.

So starting off, we're gonna talk about, um, AI skepticism. So we have a few different articles I'm gonna read from, um, the first one, some 29% of organizations participating in a study consider themselves doers or leaders in adopting AI for infrastructure app delivery, security data automation. Last year that figure was just 4%, at least 90% of US business leaders are encouraging AI use or are confident in their ability to effectively harness its benefits, putting the country ahead of other leading nations in AI transparency and investment.

5 million in gen in generative AI this year. However, after all that we have another article that says consumers and financial analysts have some cautionary words for gen AI's plans for world domination. Not so fast a whopping 72% of consumers say they don't use gen AI services according to a fresh survey.

So what are your thoughts? Well, there's a lot to unpack here. As they say, first and foremost, like the investors are kind of outta their minds because they thought that maybe all this AI stuff was gonna drive up revenue immediately.

And it takes time for people to kind of figure out how to use this stuff because um, yeah, it's available to consumers and they might be using it to write a better email or something of that elk. But that's not really the point, right? The point is, is to figure out ways to apply this to processes so that we all become more efficient and um, a lot of the toil goes away, but it takes a while for people to reverse engineer the processes and figure out how to use these tools to embed them in those processes.

And by the way, the gen AI is what they call probabilistic and the processes are what they call deterministic. So you got an AI platform that is giving a best guess to something that's gotta be a hundred percent right a hundred percent of the time. These things are not just gonna, you know, slam dump their way into existence.

So I think the Wall Street side of the equation's a little crazy and consumers are um, you know, they, they toy with it and they're happy to have a summary of say a Google search result rather than having to flip through 22 ads before they get to what they're looking for. And that sounds good too, but does that drive additional revenue? I don't know.

And uh, the other side of this coin is to be a little deeper into one of those surveys. It will explain that a lot of the funding for these projects is money that's allocated to other things that's just being repurposed for gen ai. So it may not be net new spending.

So that's another thing to take into account if you are a um, wall Street type. But I think it was Bill Gates who said we frequently overestimate the impact of innovation in the short term and underestimated its impact in the long term. And it's pretty clear that AI is gonna transform all kinds of things.

We just need to have a little more patience. Yeah, absolutely. And I think what we've heard a lot and we've spoken to in other podcasts is that business leaders tend to put a whole lot of faith in AI for everything, but maybe they have a little too much faith in it.

It's still just a tool to augment processes at this point. Yeah, business leaders, uh, the minute they hear anything that's gonna, you know, reduce the cost of labor and increase profits, you know, everybody's bell starts to ring. But uh, let's be honest, not a lot of business leaders really understand how the processes in their organization really run.

So I think that the people who are involved in this kind of look at it and go, well that's an interesting uh, tool at the moment, but we're not really using it for every little thing that we imagine maybe immediately because, well, it's not consistent enough. It's that it's basically that problem or we don't have data that's applicable that's been trained by an AI model for a given process that will come in time. We will see um, more domain specific language models show up that have been trained for specific tasks and then we'll get to the level of increases in productivity that everybody's holding for.

So, um, stay tuned. Yeah, absolutely. And I'll point out that we have another podcast, the digital CXO podcast.

And if you listen to that one, you'll hear about a survey in regard to AI that shows that even though it may increase efficiency, there's only so much that humans can do and there's a point where they're burnt out and they're saying the use of AI is not helping them. Yeah, I know, certainly whipsaw on the other side of it, once we get it going, people will be like, well wait a minute, can I really uh, manage 12 things? Do I have the cognitive load to manage 12 things simultaneously?

We'll see. Alright, so moving forward, this is the rise of software intelligence. So we have an article on Techstrong AI that I'm gonna read from when it comes to generating new software code AI can be extremely useful, but typically over 75% of large organization software engineering resources focus on maintaining, enhancing and modernizing existing applications.

So an answer to this, we have some other articles that I'm gonna share. BMC as part of a broader commitment to building generative artificial intelligence assistance to simplify the management of mainframes has made available in beta a tool that explains code functionality. Also darpa, the defense department r and d agency, will lean on emergency AI capabilities in a new program to deal with the costly and time consuming challenges of rewriting c and c plus plus code to rest in a move designed to meet the push of federal agencies and private orgs to adopt memory safe programming languages.

So can you speak to that? This is all gonna be a killer use case for ai. I mean one of the things that plagues us and nobody talks enough about is this thing called technical debt.

And we have all these applications out there that are written in older programming languages that we continue to support. And that ranges from all kinds of stuff from old SAP code and COBOL code to c and c plus plus stuff that has been shown to be, um, vulnerable as cybersecurity attacks more times than anybody cares to admit. And rewriting all that stuff has been nearly impossible.

Now is it gonna be a flick of a button and it's all gonna magically go away? No, but um, generative AI does make it possible to reverse engineer that code a lot more easily and then I can tweak and tailor it. So things that might have taken a year might be done in weeks and months as we go forward.

And that's all positive because a lot of this stuff we need to get rid of, let's be honest. It's just that we can't right now 'cause it's too hard to modernize it. But I think we're getting to the point now where modernizing all that stuff, sometimes replacing it is gonna be a whole lot more feasible and these articles are all pointing in that direction And it also helps with that, a lack of skill that's in demand, it's gonna help, uh, lower the bar for that as well.

Yeah, I may not have to be a cold ball rocket scientist to rewrite something into Java or uh, any other particular language that I can run on that particular mainframe platform, so, you know, cross your fingers. But um, I also think that competition will increase as well because so many things that we have today that we run, we just continue to run from that existing vendor because well, it's too hard to switch. Well if I can go in and reverse engineer a lot of the software that that vendor embedded in there, well the cost of switching just went down dramatically.

So I think great things are coming here and this is gonna be one of the best use cases of AI ever. Absolutely. Alright, so moving on.

Uh, open AI is working with the federal government on early access testing for its next major gen AI model, a nod to growing concerns over the safety of its products. What are your thoughts? I like the idea, I'm not quite clear the federal government's in any better position than anybody else to assess the safety of these things.

I mean, you know, they might be able to figure out the, uh, safety protocols for food in our food chain, but AI is a whole other conversation. So I might feel better about this if I had a, a better understanding of what it was that the government was gonna bring to bear here to assess the safety. And two is what exactly is open AI sharing with them Precisely because it may just be, you know, a bunch of schematics and that doesn't tell me a whole lot about what the safety protocols here.

So I'm kind of hoping that there's gonna be some sort of independent consortium made up of actual AI experts that will assess the safety of these things and where all those people are gonna come from and who's gonna pay them for their time as anybody's guess. Yeah, that was my thought too though, is that it shouldn't just be open a ai it needs to be all the, the major players and then a team of experts working with the the federal government. Yeah, well cross your fingers.

I think that eventually we'll get there. Uh, let's just hope we don't have some sort of catastrophic event before that. Let's hope so.

Next Gemini, everywhere. Google as part of a raft of updates to its cloud services made available a preview of an update to its looker business intelligence applications that leverages Gemini generative AI intelligence capabilities to enable end users to employ natural language to query data and generate reports with much fewer constraints. Yeah, if you've ever used a BI app, you've always had this feeling of being contained as it were.

And the reason for that is they were all canned, right? They're graphics and things that I can ask questions to. Usually I needed to know SQL or something that was equally powerful programming language and then to interrogate the data, but it was always within the bounds of some assumption that somebody put the right data in the right place and so I could actually query it.

And um, you that's limited. So when Google's talking about Hereward Looker and other people are talking about this in a similar fashion is I can use the basic prompt to present it by gen AI to have a much more, uh, iterative conversation with the data as I interrogating. And I'm not constrained to whatever is on that screen at the moment or was ever in the visualization presented.

I can query something or ask through a prompt something and then keep that conversation going. There might be some memory limitations about what I can remember and there might be some, uh, hallucinations along the way, but it allows people to actually kinda experience what, when they hear the term business intelligence, this is kind of what they always thought they were getting versus what they actually got. And I'm kind of hopeful that this is gonna change our relationship with BI applications and make them more accessible to a broader number of people and we will make more informed decisions and hopefully we'll make them faster.

Um, today sometimes we all rely a little bit too much on our gut and sometimes that gut is correct and it's experience talking to us. And other times it's just what we had for lunch didn't say, well, So you think we'll see this integrated in a into a lot more places. Yeah, I think this, this'll be pervasive everywhere.

I think, um, every bi app is gonna move in this direction to one degree or another. And who knows, maybe one day I won't type in anything, it'll just be a voice and me talking to the machine going, Hey, what do you think about this? Or if I change this parameter on this kind of what if scenario, what would my results look like?

And then, and then it will generate some sort of graphic that will show you that data in a way you can visually understand it or, uh, and maybe even further down the road, maybe I'll just walk into work and I'll say, Hey, tell me the three things that are likely to get me fired and it'll just tell you, Well, that would certainly simplify things. I would love it to just be able to just ask. I mean I imagine that would change everything if they get to that point.

Yeah, sometimes you know, they say that the sign of true intelligence is knowing what questions to ask. It's not knowing the answers to everything and the machine might know the answers to everything, but you still gotta figure out what questions to ask. True.

All right, so last in our list lineup today is, uh, pin drop the tech company that tracked down the sources of AI generated voice cloning incidents, targeted targeting President Biden and Elon Musk has now traced the tech used in a more recent case involving Vice President Kamala Harris using the company's pin drop s technology. The researchers said the unknown individuals behind the video that made the rounds on social media used Tor Tortoise, an open source text to speech system that can be found on code and project repositories, GitHub and hugging face as well as in voice cloning frameworks. So what does this mean to everyone else?

This means that all these deep fakes that are starting to emerge have digital footprints that can be traced back to, um, their source. Then we may not know exactly who created them, but the fact that we know where they were created and how they created it suggests motives and also suggests how, uh, shall we say, uh, invalid and untrue they are. So if we can trace that back to their source, even if we don't know who created it, we can at the very least start to suppress them because we know that they're made with malicious intent.

Uh, and so this is a positive development in my mind. It's probably not from the perspective of the our uh, presumptive democratic nominee there who's probably not enjoying these kind of deep fakes, but they're everywhere and they soon will be. And I think the question now is how do we identify them and trace 'em back to their source and ultimately suppress them if they are malicious?

Yeah, absolutely. And I know, um, we've talked about some other technologies that could help potentially with this somehow harnessing blockchain technology, somehow using watermark technology and things like that because it certainly is, um, confusing many people. Many people are falling for it and do not realize that these are fakes and they're getting shared everywhere.

And um, so any way that we can prove what's fake and what's real would be helpful. There will always be gullible people. They have been around since time began and there's not much to be done about that.

But between certifying content that we know to be true and suppressing content that we know to be false, we could at least get to something that is eh, reason. Absolutely. Well that brings us to the end of our list today.

So Mike, do you have any final thoughts? Um, don't believe the hype and don't believe the skeptics. So somewhere between those two extremes is a middle and that's where we're all gonna land.

Alright, well thank you for tuning in and if you miss last week, it's up. Go check it out and go to Textron ai, let us know what you think of the articles you find there, or send us an email if you're interested in hearing about certain topics. And have a great week.

Alright, we'll see you.