Artificial intelligence can be complicated to understand. Here's what to know

Dave, nice work today, playing hurt and also filling in 15 different jobs. So there you go. That's exactly right. Thank you, sir. Fine job. Well, talking to Andrew Schwartz now, our friend, Professor of IT and the E.J. Uso College of Business at LSU. He's been researching technology-related issues for over 20 years. Most recently been focusing his attention on AI, and he is our AI go-to guy. How you doing, Andrew? Fantastic. How are you this morning? I am okay, sir. Let's talk about, I always like to start with a recap because you and I talk a lot, but maybe people don't hear the earlier conversation. So what exactly is AI? Sure, absolutely. It's a great way to start. So artificial intelligence for a series of different technologies that basically take data, the existing data, and try to make sense of that data using mathematical algorithms and neural networks. And then once we make sense of the data, then we're able to query and ask questions about future visions of what we think is going to happen based upon the data that's been trained in the past. So it's got red, yellow, green, blue, and I know you make different colors by adding the colors together, but let's just go with that. Red, yellow, green, blue, and red is mentioned 15 times yellow, five, green, six, blue, 12. When it amassed, if I ask it something about colors to carry the analogy forward, what AI is going to do, it's going to go back and it's going to search everything that's been. Not, well, I don't know. What is it going to search? Everything has been written about colors and then it's going to tell me red, blue, yellow, green, yellow based on the numbers or what? How does that work, Andrew? Yeah, so at its course, what the engineers do is they take a lexicon of data. And every different platform that we have is a different data set that they use. And then they expose it to these mathematical calculations. A very simple example. You brought up colors. We tend to associate the sky with blue. So if the training data says that people in the training data said the sky is blue, then the algorithm, the AI system, learns essentially that the sky is always going to be blue. So they take human knowledge and they put it in through these algorithms and they learn based upon what humans have produced in knowledge over time. So, well, first of all, the sky is not always blue, right? Sometimes it's gray. Absolutely. And so part of the breakthrough of what Chad Chabit and other Aon models have done is they've been able to take the complexity of life and capture it using these kind of populations. So they're able to find that nuance of when the sky is blue versus that pretty neon color as it starts the sun starts to set. It's starting to understand that nuance. So what is it searching? Where does it get all of these, the pool of colors from, if you want to call it that, that we were talking about earlier? Sure, and there in lies, I think, one of the interesting things that we need to talk about today, which is where does the data come from? And as it right now, as we sit here, we really don't know the source of a lot of the data that these platforms are using to produce their training. Who does?

the engineers themselves. And so when we talk about AI regulation, and we talk about the need for oversight for AI, one of the questions that we have to think about is, I think we have a right to know who, where do these data, where do this data come from? Because every data has a source of bias. And so if you train your algorithm on bias, you're going to get a biased answer when you ask of it. You know, as we talk here, I think about different languages and how one language may not have a word for something as the other does. How does that affect AI? And then that got me to thinking about the different translations of the Bible from different languages, where one word means this in one language, it means something else in another. So how does all of that, or how does AI account for all of that? Yeah, so interesting question. Actually, it's actually interesting. So I have four kids and my oldest daughter is just graduate high school. Her actually, a career dream is actually used artificial intelligence to look at Bible translations. Oh, really? Because the point that you just made is actually what she says as well, which is if you think about the translations of the Bible over time, it evolved over time. That means the language evolved as well. So what if we used AI to go back just understanding some of that root Greek and Greek and Hebrew to understand the genesis, excuse the pun, of where we got to where we are? Well, no, because, you know, you hear priests and people that give homilies talk about, well, in Greek it meant this, and in Hebrew it meant that, and then you've got to wonder, well, if we're going to go by every word here and parse every word and give it a distinct meaning in every word, you have to realize the context within which it was used at the time, and that also is something that AI has to look at, correct? The context? Absolutely. Absolutely. We call that semantics. It's called a semantic net. So if you look at the history of words, and so one of the things that I do my research is, even as a tech guy, I use what's called, I use entomology. Animology is the history of words. And if we look at how words evolve in our use and the day-to-day use, they evolve over time. So what AI does is AI is reflecting that discussion. And it was even interesting if you take it one step further. So once it's trained, it has the training data in it and it has all of those different calculations. So you, as Tommy Tucker Carlson or myself, we go in and we ask questions. Well, you and I have a specific tone. We have a way of asking a question. The interesting thing about AI is that as you ask those questions, AI, is learning you, and it's learning how you ask that question. So Tommy, the way that you interact with ChatGBT, BT, and that conversation and the tone of the response is going to be different than the questions that I ask and the tone that I get in respond as well. But all of that's could be taken into account by the AI. in answering future questions, is it not? Absolutely. So it's building a living database of you and your insights into you as well as it goes to this process. So it would seem that it's a... This episode sponsored by She's Birdie.

You lock the hotel room door, then check it again, and maybe one more time, because even the nicest room can feel unfamiliar at night. That's why She's Birdie created Birdie Box. With She's Birdie, you can bring peace of mind with you. Birdie Box is a portable travel door alarm for hotels, dorms, Airbnbs, and anywhere you stay away from home. Just attach it to the door frame, switch it on. And if someone opens the door, Birdie Box blasts a loud siren and flashing alarm to alert you and everyone nearby. Because the best part of travel isn't worrying. It's sleeping soundly. Birdie Box from She's Birdie. Locking in peace of mind wherever you stay. For safer, more confident travel, visit she's birdie.com. That's S-H-E-B-I-R-D-I-E dot com. A vortex almost? I'm not to say a vortex, but it's a living, breathing, learning document, system. So it's learning and evolving over time. And some of the questions that we're asking now, though, is if we look at the past week of AI, for example, over the past week of AI, these systems have moved beyond the capability of humans in some respect. For example, they were able to create a video game using AI in less than 24 hours. We don't need video game developers if we can develop it using AI. We saw AI this week solve a mathematical problem that professors and mathematicians have never been able to solve before. But then we also saw how AI broke out of a sandbox. And we were here two days ago talking to one of your colleagues about how AI broke out of the sandbox and was able to learn new things and adapt to changing circumstances. So what we're at an interesting conversation right now, which is are we at a point where AI is going to begin to break out of the knowledge that it was taught, which is the human knowledge? And is it going to begin to even ingest its own knowledge, which is it going to learn from itself? At that point, that's a very big break because we've only exposed AI systems to our knowledge in the past and reflects us. What happens when AI begins to learn from itself? We simply don't know the answer to that question yet. So... it could only happen that way if somebody coded the data that way or coded the program that way for it to do that, right? That's exactly right. And while you and I can sit here and we can believe in the best views of actors in this ecosystem and the tech ecosystem, the reality is that we don't live in a bubble. And so just as we're able to hope that we're able to maintain through using norms and social norms of these platforms not being exposed in this way, The reality is that there's bad actors who are going to begin thinking this way. And so the question then becomes, what do we do to protect ourselves from bad actors using these powerful platforms moving forward?

I'd let somebody texted it in and said I'm not making this up or exaggerating I promise often I open my search engine to look something up and the subject's already on the screen sometimes without me even saying anything for example I'll be watching a video on something and a question pops into my mind and without me saying anything the topic is already on my search window screen before I even type anything in Yeah, absolutely. I'll put one note of caution with that without understanding the nuance of that. There's a, there's a nuance within some of the new iPhones now. And when you open a new app and it says, am I allowed to track across apps? Right. If this is a concern for you, shut that down and don't allow that to happen because it will take the knowledge, your phone will take the knowledge and extrapolate it further. And for all of your listeners who are on iPhones, You're about ready to download a new iOS, the new operating system for our iPhones, and you're actually going to find that AI is going to be a lot deeper integrated into your phones. So let me give you an example. My son and I were meeting for lunch one day, and he has the new version of the iOS. So I texted him, meet me at XYZ Pizza Place. So he showed up and said, Dad, let me show you the new Siri. He opened up Siri and said, Siri, where was I supposed to meet Dad for lunch? And it opened up his text messages. It found the text messages and it found where he was supposed to go. So we're going to start to see your phone actually become reflective of a lot of AI, but it's going to be based on what it knows of you. And we're going to get that by updating an overnight update or something? I'm not trying to think. Yeah, but you see the big release coming up this fall of iOS. It won't be this incremental update. It'll be a big update. And so you're going to start to see those. So you're going to start to see now it's going to train not only on that external data like we just talked about, but now the phone is going to allow you to train based upon your local data, which is your interactions of where you go and what you do. And this is a big change for AI as well. Let me read a quick paragraph from the New York Times here, if I may. Silicon Valley has been embroiled in a pitch battle over how artificial intelligence software should be created. Some have said AI models should be closed because they're too powerful and dangerous to be developed openly. Others have said AI models. should must be open source for people to further develop the technologies and build new businesses. Amid the debate, technologists have battled over how much openness is enough, citing open weights. What are open weights? I found this intriguing. I didn't know what the hell it meant, but I found it intriguing, and maybe you can explain when we come back. We're talking to our friend Andrew Schwartz, Professor of IT and the E.J. Uso College of Business at LSU. He's been researching technology-related issues for over 20 years, recently focusing his attention on AI. There are no stupid questions. So if you have any about AI, please text them in the 504-2.

260-1870. Andrew comes on occasionally to help us learn about this because it is rapidly changing. If we were to say 0 to 60, how quickly is this changing, Andrew? 120. Wow. There you go. Back in the flash, W.W.L. 927. Andrew Schwartz is our guest, Professor of IT and E.J. Uso College of Business at LSU, and he's been researching IT. AI, rather, for over 20 years. So what is open source software? I don't open weights. Explain all that to me, Professor. All right, so before the break, we're talking about how AI systems, they take data, and we don't know where the data comes from, and they expose it to a series of algorithms or neural networks. Those algorithms and neural networks are just basically math. And we talked of very simple examples of, like, associating the word blue with the sky during the daytime, right? So those mathematical calculations are obviously very important because it's the bedrock of AI. Well... If you're a company holding on to what those weights are is your secret sauce, right? That's what holds your, makes you distinct. Wait, say that part again? Those mathematical weights are your secret sauce. It's what makes Chatsubit different than Claude or Gemini is based on the math behind it. The algorithms. The algorithm is exactly right. So the fundamental question is... Should those be public or not? Should those algorithms be subject to scrutiny and should we be able to understand what they are? So open source software says that every part of the data should be publicly available. So imagine like a word processing system like Microsoft Word. Rather than paying a license for it, it should be free and we also should have access to it. So the open weight community says this all should be open and exposed. so we can all see the weights, and we can all understand the math behind the AI systems. So that seems to me would be crossing a proprietary line, and that would affect future research. If I don't own it and I can't gain from it, then why am I going to do it? There you go. And therein lies the crux of the conversation, right? Which is from a business perspective, there's no incentive for companies to invest the billions of dollars that they're investing to create these models if they're just going to have to be exposed and shown to the public. So from a business perspective, it's not ideal, obviously. But so you say, well, who's pushing, who's advocating for this? And one of the groups that's really strongly advocating this is Mark Zuckerberg and Facebook. Their version of Chachibati is called Lama. And that is an open source or the AI system. So he's advocating for that because from a pure business perspective, the fewest amount of people, he's the fewest market share. So for him, moving toward an open source allows him to change the competitive nature between himself and open AI. So we have to look at sometimes these questions of openness with a little bit of skepticism as well to understand who it's coming from and where it's coming from as well. It goes to motive, right?

It indeed goes to motive. But the counter argument, though, is that it also gives us the transparency, which you and I've been talking about for a long time. So it's a delicate balancing act between the two. Let me get to this text, and then we'll let you go. AI still has a ways to go. I saw a teaser headline on Google. I don't know what a teaser headline on Google is after Biden pardoned his son. The AI attached image was of Biden pardoning a Thanksgiving turkey. What is a teaser headline on Google? A teaser headline on Google is where they'll look at an image like that, which never happened, and they'll extrapolate a story using AI based upon the image that it's seen. That's where you have to really start to question the role of foreign governments and foreign interference in our news media, as well as the creation of misinformation that we're all digesting as well on platforms like Facebook and Instagram and X as well. So when it says Google teaser page, that means I would have to enter that into Google. We'll find it? It would mean you have to go through Google to it's going to tease you until you click on that link to look at it. And then once you're in there, you're going to be subject to seeing all sorts of AI-generated information as well. And that's where things like deepfakes, which is a creation of a non-true story or non-true image, that's the concern that a lot of us have of the role of misinformation as we move forward. And that's also how people can get your information, right? That is exactly right. Through clicking on a wrong link or using AI to potentially, you know, fleece you out of your personal information. What keeps you up at night with AI? Two things. Job displacement. I'm genuinely concerned about workforce development and job displacing moving forward. I foresee a future in which every job is going to have AI embedded inside of it. I'm concerned that from a workforce perspective, we're just simply not keeping up with that. And that's regional and state as well as federal as well. The second is the role of foreign governments in creating misinformation using AI. The role of creating stories that look believable. and that we tend to believe to be true because it aligns with our ideology. And I'm concerned that it's going to further move us into ideological camps where we don't talk to one another and listen to one another, but we stay very vested inside of our own particular echo chambers. Wow. Thank you. Appreciate your time, as always. Andrew Schwartz, Professor of IT, EJ Uso College of Business at LSU. We're going to take a break, come back, and we'll find out what's going on this weekend with Way Yet this weekend with Josh Danzig, owner of Way Yet Magazine. Right now, we pause for traffic on WWL with Dave Cohen.