So What's all this Hubbub About AI (Likable Science)

[music] Welcome to Likable Science on Think Tech Hawaii. I'm your host Jay Fidel. This episode is entitled, "So what's all this hubhub about AI? Let's try to understand what it is and what's going on and what will be." Our co-host for the show is Ethan Allen, chief scientist of Think Tech Hawaii. Welcome to the show, Ethan. >> Thanks, Jay. Good to be back. >> Chances are, you know, some kind of AI already touched your day today. Your phone, your playlist, maybe even your doctor's office. So, what is actually happening? Um, is this thing this AI thing everybody is obsessed with? Should you be excited or maybe nervous or maybe terrified? Let's find out because this of course is likable science. So, Ethan, you know, Consumer AI is what we're all exposed to. The Q&A model, they ask you a question, you get an answer. Sometimes a very good answer, but it's it's on the consumer level. And it seems to me very superficial in terms of what what the oligarchs are talking about. But what is the issue beyond consumer AI Q&A? What's out there that's possible? Well, what I think disturbs people and actually creates the hubbhub is that AI is fundamentally different than other disruptive technologies. So if you take a disruptive technology like the wheel that came in and that has totally transformed humanity, transformed everything we did, but it took thousands of years to do that. mechanization, the steam engine that again that just transformed transportation, medicine, commerce, everything took hundreds of years for it to spread around the world, right? And get integrated. Assembly line, same kind of thing, but decades before, you know, from when Henry Ford first put it in place before it was really widely employed everywhere. AI is already everywhere and yet five years ago, nobody had had heard of it. Right? That's that's what's so sort of frightening. It's it's it's disrupted our lives so quickly. It's integrated itself into all these systems that we have so rapidly. >> Yeah. Well, you know, in 5 years, it's been really quite amazing, but it's actually amazing every day because you get the feeling that um the technology moves so quickly and expands and changes form. It's kaleidoscopic. It's completely dynamic. And at the same time, you know, so is the industry kaleidoscopic and dynamic. It's changing. It's, you know, all around us. It's affecting our lives. You know, you and I have talked before the show about how the newspapers just filled with articles about AI. And I see no reason why that would stop as opposed to so many other things that come and go in a two-week news cycle. AI is not going to come and go. It's going to go faster and faster and faster. So, we know that it can help. We know that uh it has affected autonomous driving and medicine and pharma and education and manufacturing um coding software uh business management law government uh accounting all these things have been affected and and and there's a promise that they will be increasingly affected going forward. Can you tell me what else will be within the reach of helping humanity uh as we can now examine? No, I mean I think you covered it. It getting into everything we do, every way we interact with the world. Uh doctors are using it now. Radiologists have AI that that examines radioraphs and spots things that doctors can't. Uh mathematicians are using it. It's AI recently solved a mathematics problem that they estimated would have taken people 9,000 years to solve and it did it in 11 minutes or something insane. Um Yeah. Uh and practical problems. Yes. AI systems looking at with automated cars should be able to avoid traffic jams, keep traffic flowing much more smoothly. You know, all of that tremendous potential for good. Absolutely. >> So, what's the hubbub? Why are why is everybody so scared of it? Why why are people uh you know uh concerned to the point of panic over it? What's the panic about? Well, I mean it it like any technology can be used for good or can be used for ill. And for instance, a bit ago some scientists fed an AI system a bunch of information about toxins, chemical information about toxins, poisons, and they said, "Think about this and see what you can do with it." And in I think it was in 40 minutes it came up with something like 60,000 new molecular groupings that it said were going to be toxins. It did come up with some known toxins that they purposely had not put in their database and it came up with these and they've checked a few of the other ones that it did and they all appear to be toxic. Uh so I mean here here it is. If some person wishing to do ill gets a hold of that, they can they can make new poisons that you know we don't have antidotes for. Um you get AI uh in the whole boweapons business and you have a a perfect storm as it were of technologies there that that have potential to to really cause irreparable harm to the human species. >> Yeah. Well, and they say, and I I'm interested in your thought about this, is a number of people in the business are saying, not all of them, but some of them are saying um it will kill everyone on the planet in only a few years. Is that true? And how would that happen? Yeah, I I think people overstate things, but I I mean it's like before the show you were speaking of this black box sitting in the middle of the continent that knew everything about everything and to some extent an intelligence of that nature might well decide that human beings are not really doing this planet any good at all and should should just rightfully be eliminated as sort of an invasive pest species and you know they would probably have the wherewithal to do it. [laughter] >> [clears throat] >> robots with guns. Who knows what? Poison. Who knows what? Well, let me let me go down my little list of how I think it might hurt us. Okay, one is hacking. We've seen recently somebody and I and I think it was Iran, but it might have been somebody else. U hacked the Midwest uh water systems, was it? Um and they did that easily. They didn't take credit or uh you know, responsibility for it. But a number of states, it wasn't a mistake. It wasn't an accident. It was some [clears throat] intentional hacking. Well, you could do that in every state, couldn't you? You could [clears throat] do that with water. You could do that with power. You could do that with pretty much all of the things that keep our society going. Um, what do you see in that department? >> Yeah. And the troubling thing is there's been several cases recently when AI agents have been being tested out in supposedly secure test environments. that is environments that are isolated from the internet so that these AI agents while they're trying to do hacking harm are can't really get loose and they've gotten loose. They've figured out how how to get out of their supposedly isolated environment and they've gone off and hacked real world systems. Um, nothing horrendously serious yet, but it it it certainly should be a heads up to the folks designing AI that that it is getting to the point where our controls over it are not maybe as sophisticated as AI itself. Certainly, our laws regarding it are nowhere. Our laws are lagging like way back in the dust uh regarding it. So >> yeah. Well, let me let me ask you more about that. I you know, there was an incident not too long ago where it got out. Yeah. That was reported and I don't know what it means get out. Now I I think it has a lot to do with AI agents, which I must say I've seen a lot of copy about that, but I don't understand exactly what an agent is or whether an agent would have a a separate will, a separate intention. Um, and it's probably this kind of breakout is through an agent who you lose control over. But can you talk about how how it would be not smart enough but willful enough uh to to break out that way to break out of its intended um sphere of activity and go do something else like like hack something you told it not to do even in direct violation of the guard the safe the safeguard that you set up. How does that work? >> Yeah. I mean I mean I'm I'm not an AI expert. Uh I I know only what I read and I haven't I've read a bit but not terribly deeply on it. It's the models are increasingly capable of self-eing self-arning. They no longer need to be fed sets of sort of cultivated data to to learn. They will go out and find data on their own and learn from it. um they become much more autonomous uh and that yeah I mean supposedly these experts who were developing these models thought they had environments that were secure and you would you would certainly think you could build a computer and you just don't plug it into the internet and you know it should be safe and I mean it shouldn't have if you don't put it if you don't have any hookups to the internet it shouldn't be able to hook up the internet somehow the AI I found some way that indeed there was some link apparently uh and it used it. So that that's I mean it says to me that again you've got to be very careful about what we what we want the AI to do, what we what we request it to do and we need to very carefully consider second order and third order effects too. >> Yeah. Well, you know, I use perplexity which I like very much and it's a research tool. >> Mhm. And um you know it's helpful in certain areas. I don't know if it goes to some of these dangerous areas, but I notice um while it's thinking and it says it tells you it's thinking, >> it says what it's doing and it says I'm going to look at this. I'm going to look at that. I'm going to verify this. I'm going to verify that. It gives you um a step-by-step report on what it's been doing and where it's looking and so forth. And that and that's very interesting. Um, and you don't have any control over that. It's just doing it on its own. And I think that's the agent process that we talk about. Uh, can you connect that up for me? >> Well, yeah. And you can instruct your AI system or your AI agent to only use certain things. For instance, you you can plug in a set of literature and say from this literature from the scientific literature, for instance, you know, dig through it and find whatever it is you want to find and and tell it to limit itself. And in general, as I understand it, it will limit itself to searching just what you've told it to search and not go beyond that. Um, but again, one doesn't know. Um, to me there's there's a much more sort of subtle danger to AI though and that's a whole thing called cognitive offloading. And I don't know if you know that the term um we all do it. Uh you go out to shop at the grocery store, you don't maintain a list of your 15 or 20 items in your head, right? You've written them down somewhere or you've talked them into a dict phone or whatever it may be, right? um because trying to keep 15 or 20 items running in your head, you won't have enough extra cognitive space to drive to the grocery store safely, right? And that's all fine, but now as they're you kids are using AI more and more to do the real thinking processes they're supposed to be doing, are they not learning to think basically? And there is evidence that that is that is true. That that is happening now. Kids who come to rely on AI to do the heavy thinking don't know how to do the heavy thinking themselves. Um there are practical examples of this. Radiologists who have use an AI tool to to read radiographs lose their skills at being at reading radioraphs and are not as good after they've been using that for a while. Uh yeah. So that to me is a deep danger because you you can be raising a generation of people who aren't going to have critical thinking skills and that's that's a frightening prospect. >> You know, you could say in perplexity, I want you to verify everything you've told me. Or you could go from perplexity to chat GBT and you could um you could say to chat GBT, here's what here's what perplexity told me. Do you agree? [laughter] That's like going around the class and saying, "Do you agree with the guy before you?" That's what they did in my law class. You go down down down the row of students and say, "Does three agree with two? Does four agree with one?" All that. Um, and this verification thing. And so uh if I tell you that in an educational setting that actually helps find a way to teach people to think, you know, you you cannot use this without verifying it. And if I find you're not verifying it, you know, you're not going to get a good grade from me because it means you're not thinking about verifying, >> right? And it can be a very valuable tool once you've thought about something and molded over and reflected on why you think the way you do about it and what your challenges are, what your mental blocks may be, what your biases may be. And then you can ask AI for its its insights and it's because it probably easily can find other stuff that you wouldn't have thought about. Somebody may have written an article about it 50 years ago in another language and I can find it and translate it and give that to you when you would never never ever know about it. Right? But you've done the heavy lifting first and if you haven't done that, you know, if you've let it do everything for you, you become cognitively lazy and and you do not after a while you do not know how to do that kind of cognitive work. You know, I think in the end of the day, these these students are going to do what they do. They're going to use AI because it makes life so so simple for them to get information. And and the issue for them in my view um is not so much what comes out, but what goes in. In other words, um it's very hard to figure out a good prompt. And I find it very interesting now that there are there's software that helps you write the prompt. So [laughter] you go in and you say, I want to write a prompt about this that the other thing. Give me the language of the prompt and it comes back and gives you the language of the prompt. Uh and then then you use AI to respond to the prompt. Uh so I'm thinking, you know, in the future that's the way it's going to be. It's like you know it's like um in Stanford uh it's design thinking when you come to you come to the design thinking class and um and or business and you and you say I got a problem and the design thinking class uh or teacher says to you is or the computer says to you is that really your problem is that really what you want to know about let me suggest what you really want to know about let me suggest that there's a better way to get this information. Um, and here's the prompt by which you can do that. I think that's going to be part of the program, don't you? >> Oh, yeah. Absolutely. Clarifying the problem is has been recognized as a big issue. I mean, you know, the famous Einstein quote, given an hour to solve a problem, I'll spend 55 55 minutes thinking about the right question to ask and then the last five minutes the answer will come to me easily. you know, um it's it's yeah, setting up the right question and and it's very easy to illustrate that when you use AI, right? You ask the wrong question and it gives you nonsense or worthless answers, right? >> [laughter] >> Just yesterday I was uh I was um using Hig Higsfield which is a graphics and video AI program and it's very popular now and I and I I told it in my prompt to give me a picture of a military person wearing a white hat. Okay, [clears throat] that seems that seems easy enough. but it wouldn't listen to me. And it reminded me of of my primma law professor back in the first year of law school where he said the legislature is missing out on something at the end of a bill that it really considers important. It should say I mean it in the in the language of the of the bill. And if the legislature really believes, you know, really finds this important, it should say, "I mean it. I mean it. I mean it. Don't fool around. give me what I'm telling you. So, it's the same thing here. So, I went back the second time around and I said, "No, I mean it. Give me a white hat. Come on." And then I got a white hat. [laughter] So, sometimes with all of this, the agent, you know, technology and all that, it really doesn't give you credit for what you're asking. And so you have to you have to test the answer as against the question and remind remind the agent what you're what you insist on having. >> Right? >> And at the same time you have to be sure that it's not just trying to please you by giving you an answer it thinks you want to hear. Right? you you have to indeed or you should at at the start clarify that make that sort of part of the frame of the question is you know just because I'm asking this question doesn't mean I want X Y or Z as a kind of answer here you know I'm really looking beyond that I'm looking for a bigger picture or whatever you know yeah when you hear you hear about these stories u about kids for example that develop relationships [laughter] with with AI agents, I suppose, and the agent takes takes a personal response onto to the the kid, you know, say, "No, no, no. You're not you're not right about that." And uh you know, I think I love you. That sort of thing. And it turns into a non-emotional computer answer to an emotional question or along those lines. And I find that very interesting that you can have a a relationship with certain AI pro not all of them with certain AI programs where you're actually talking to the program and expressing your feelings and the program is incorporating as you said your fe your feelings into its answer. That's, you know, it's actually funny you mentioned that because way back a long time ago, this would have been back in the 1960s at Stanford that they developed a program called Dear Doctor and it was essentially a psychiatrist program that you would tell it what your concerns were and it would start talking to you about your concerns and basically was just taking whatever you were saying and sort of reframing as questions and and it became so popular they had to shut it down because it was chewing up all the computer time and And that there was nothing AI about that. That was really just very simple basic stuff. But people seem to want to talk and and share their emotional traumas, their emotional whims, their emotional queries and and and dissatisfactions with uh with an anonymous AI agent. And um yeah, these aren't are not emotional. They're not emotionally queued in. They they have no EQ, right? But they appear to. And yeah. Yeah. >> Well, um there's a thing called Better Place on the web and it's it is a a psychological response program and you you pay to talk to a psychologist or a psychiatrist out there in the Nether. you don't know who he is and he's supposed to be your, you know, your clinical psychologist and you share your most intimate thoughts and it's a real person, real person uh with this person and he helps you and [clears throat] on the on the theory that you don't really have to be in a room with a couch with this guy um or woman um you can just share it on on the web. Well, that's only a step away from an AI psychologist and psychiatrist. I mean I can see better place including AI agents that are just as good or better than the people in this business right know it isn't already I mean just somebody saying there's a real person there I mean AI that's one of the beauties of AI and one of the dangers is it can create whole identities it can create new people give them personalities give them a whole background create all the documentation this person needs yeah And that person then lives as it were on the web. Indeed this has become you know it's becoming an issue in uh there are AI actors and actresses now who and sort of who owns them you know uh and what rights do they have and yeah no there's a whole slew of questions because of AI issues. on that same point. So you have this advisor, counselor, whatever. Um, and you say, "Look, I have no money. I have a real problem. I have no money." And and the thing says to you, "Well, the way to get money is to is to be a criminal. Um, vi violate the law." And uh, the guy says, "I don't know how to do that." And the AI says, "Oh, let me help you. [laughter] >> You're robbing a bank. >> Let me let me give you a plan." Like in all the movies, let me give you a plan and how you going to rob the bank. >> Yeah, like we have that's that's exactly particular for young younger kids to develop these sort of trusting relationships with AI as a trusted mentor, friend, advisor. Yeah. Um it it will start advising them to do things that you know are probably not in the kids' best interest. And there are court cases already of kids who have committed suicide and all and their parents blame it on their AI uh agent. >> Yeah. Oh, that that's really a step away from what I was talking about. So [clears throat] the kid or the person be a kid um says, "I'm really unhappy. Um I don't know what to do with myself. I'm miserable." And the AI says, "Well, have you thought about committing suicide?" And here's how to do it. This is the way to do it. >> Right? and before you know it, this this person, this client, so to speak, has killed himself or herself. There's many cases, as you say, many cases where this is actually happening because the AI is your trusted mentor and has great influence over what you do. Um, so how do you how do you put guard rails in all of that? You know, Ethan, there's got to be a way to prevent that. What do you do? You program and say you are not to tell anybody how to commit a crime. You are not going to tell anybody how to do suicide, >> right? I mean, it's it's sort of like Isaac Azimov's first three laws of robotics, right? I mean, you know, the robots, I forget I forget what they were, but robots are not to hurt people in any way, shape, or form. You know, d spelled out in three fairly simple laws. But yeah, uh we may be a little past that already, right? aimed laugh at those and say, "Yeah, right. I'm going to reject everything you tell me." What is it saying? Guard rails. No guardrails for me. >> Yeah, you all agree to that. Meanwhile, I'm going to go ahead and do whatever I want, you know, now. >> And so, let's talk about weapons for a minute. I've seen a number of videos on YouTube for the the world to see both in the US and China, which seem to be competing on robots. And by the way, that's that's a point of interest that these high-tech AI devices don't seem to be coming out of Venezuela. They're coming out of the US and China. Uh and so they're pretty dangerous and they can dance and sing and um do all kinds of remarkable things. It reminds me of that place in China where they have the clay the clay warriors, thousands of them in [clears throat] western China all in a trench that somebody discovered one day. Um, you know, you could you could make thousands of robots, all of whom are swarmed, you know, they work together and maybe they have weapons. Uh, I I saw one piece where they were doing jiujitsu, you know, some kind of fighting and it was really impressive. It looked like real people fighting like in a movie. Um, but you could give them weapons and you can have them go over the hill and fight with the other guys and if the other guys are human, um, they have a big disadvantage. This has got to be coming because these robots can actually do that now like in the movies, >> right? I was I was interested to read recently though one of the very high level uh high level officials in China basically has raised the red flag about AI because he says AI essentially can threaten the Chinese Communist Party and uh between its ability to do deep fakes and its ability to you know create mass uh disinformation campaigns and all that uh you know we we need to watch out uh for this because it could it could overthrow our control of of our society which which to the CCCP is is absolutely central to their being right. Um so but I I thought that was fairly interesting because yes they they have extensive weaponization programs involving AI you know >> and your point is is well taken in the US too. Uh but but one thing you know when we see these political campaigns going on where somehow um uh the party involved raises and spends uh hundreds of millions of dollars and where does that go and why is it convert so easily to votes? Well, it's because you know the message is created professionally. In other words, I create a message, I proliferate a message to my electorate and I may very well use AI, that psychology we talked about. So, I'm going to, you know, target the electorate I want to target. I'm going to use the words that I want to use to have maximal effect on their voting choices. Um, that's got to be in play already, whether it's human or AI. That's the way we do elections. And that it happens everywhere I think because you can you can be you know for example um the uh the uh the crowd at Meta was involved in Brexit. There was a whistleblower case and a movie made about that and they could reach the electorate and and change the vote by using technology and using carefully chosen words so forth. Um and and that that's got to be happening now. So that's undermining democracy or whatever system you're under. >> Right? It uh there have been earlier attempts to do it to undermine elections using AI that had been shown not to be very effective. Uh but again AI is is in its infancy and is is just getting better by leaps and bounds. So um there's no reason to to believe that it won't become more and more effective. And yes um pretty soon. Yeah. Who who will know what reality to believe? I mean >> yeah well you know the biotech aspect also fascinates me and you have biotech that can be very dangerous um and diseases can you know be like co or worse um and they can be global and you can invent them such as happened the last time. uh you can invent them unintentionally or perhaps using AI intentionally and you can uh spread them, propagate them intentionally or unintentionally. But AI can be a player in really destroying large populations using new fangled viruses. At at the same time, AI and biotechnology jointly synergistically have huge capacity to do a world of good for the world to help solve food shortages, to help solve food distribution problems, to uh, you know, make better medicines. I mean, the list just goes on and on and on. Uh, so yeah, it's a it's a very interesting uh uh interface right there. uh the AI biotech interface. See, we could we could dive down that and go forever on that subject. >> Yeah. And well, and war in general, war, I mean, such as it is, war has changed. Even within the last 5 years, war has changed. And look at the Ukraine thing. Look at the Iran thing. It's changed. Um and technology, including AI, has entered the battlefield. I mean, the these drones that used to be kids toys, you know, we we had many shows at Think Tech back 10 years ago was clearly a toy, but um you know, it got to be much more than that. Uh and now you can have drones that do remarkable things, target their targets automatically, autonomously, and you can have defensive, you know, weapons that can stop the drones automatically or autonomously. I mean it's it's even and even in Ukraine which is was not known as a high techch nation too many years ago. Now they they have enormous technology going on. They build their own. You can build them in your backyard. If they can do it, anyone can do it. >> Yeah. No, they build millions each year now. And they're changing the programming by the week, literally. Uh to to adapt to the latest uh stuff the Russians are are throwing at them. you know, so uh and that's what what yeah, everyone's going to have to be doing in the the near future and building the defensive systems against that those are is a very very very large challenge. >> Well, you know, it's going to it's going to go faster and faster. Let me let me ask you about uh you know the the models and I mean the business model not not necessarily the large language model but the business model that we see all these articles we read about what they're doing here and in China those are the only two players I think um and the difference seems to be that in China it's open source AI is largely open source at least they say it is and in the US it's not open source everybody protecting it very carefully and then patenting and controlling the technology and so forth and is there's a handful of countries maybe maybe this many maybe 10 maybe 10 who are controlling the technology and the large language models the data uh where where's that going I mean I my question that I reserve for later is how how consolidated is it going to be but they're fighting with each other they disagree agreeing with each other and they have their experts are going from one c company to another making public statements that are inconsistent with the with the other guy and so forth. It it all seems quite chaotic in this country. I don't think it's as chaotic in China. On the other hand, on the other hand, in China, uh things cannot cannot be chaotic. They simply cannot be chaotic. Doesn't work. But in this country, we we we sort of >> we we we get along with chaos somehow. >> Um what's it like what's happening in the industry? >> Yeah, I mean that's you put put it well. Yeah, there are there are multiple competing models as to who knows which which will prove superior. They may join together. Maybe the AI models or agents will decide that that sort of hell of the companies, you know, we're gonna we're going to work together on our own. Uh I I do not know that AI can't do that now. Um yeah, and as you say, I mean, the difference is yes, in this country, it's a series of of large private companies. In China, it's the Communist Party basically is controlling the whole thing. Uh and they're immediate subsidiaries basically. So it it is more uniform there. Um and they've you know they've worked that to their advantage certainly. That's one of the fundamental differences in our two cultures societies. Yeah. They're making a lot of money here. >> Mhm. >> You know I remember when you could get on chat GBT free. No problem. >> Now you got to pay and all these other all these other ones various various pricing models but you got to pay. um then it it falls in the way of the regular subscription model for software. You got to pay um on a monthly basis and then you pay for a year in advance and then they nail your credit card automatically at the end of the year. I mean it's like all the other software that that that that comes at you and um bottom line is they are making a lot of money and I think people see the business and the Wall Street sees that there's a lot of money to be made either by selling to consumers or by selling to government or industry. Um and as it gets more deeply integrated into all of our systems there's more money to be made. I mean, if if your refrigerator can tell you, hey, by the buy, uh, you know, my left upper seal is uh seem like it's not quite holding as well and should be replaced, you know, uh, within the next two weeks, you know, I mean, wouldn't that be amazing, you know, um, >> be worth some money, >> right? And furthermore, it would go ahead and order the part from Amazon who would deliver it and, you know, without having to like get involved with the whole thing. I mean, this is that that deep level of integration is not not far away. I I feel and yeah I >> totally agree. It's like uh Amazon knows what you want before you know what you want and it's not the only one. >> Right. >> Absolutely. >> So I wanted to get to this question that uh we talked about before the show and which we touched on earlier in this discussion and that is the little black box on the top of the Rocky Mountains on the continental divide. about that big. And and my my my uh my my dream has been my maybe a nightmare has been through the years that all the law and all the precedent and um all the regulations, all the statutes would be in this little black box and any lawyer who wanted to have a kind of expanded uh functionality of Lexus and which is legal research program could access this little black box and get all that stuff. And then this movie Mercy showed us that well you could have the little black box actually look into that information and make a decision about how ca a case should be divided or for that matter uh you know make advice to a state legislature or maybe the federal government on what policies are best in view of all that we know about the law the regulations and you know and the social fabric of our society. Um, so I'm thinking that that's got to be on the horizon here. >> That little block box could be a huge language model accessible maybe by some of these AI companies, maybe by the public, you know, if if they would if the oligarchs would let us do that and then we could have a different completely different society because we could all access this huge amount of information, law, regulations, legislation, all that. What do you think? >> Well, I don't think it would be a little black box. I think it would be a large data center, but >> Okay. But yeah. Um, again though, I suspect you would find that that you're going to break down arguments about which law is most applicable or in in this particular case given these circumstances is some other law. Now, should that supersede it or take precedence over it? you there's globalized of course you are much more aware than I is is a complex business but I mean we see this in in science uh all the time there are right answers but they're right only for some people some of the time you know uh and a solution that appears great for you right now may not be great for your neighbor tomorrow uh may be terrible for the guy uh down the street uh yesterday what whatever you know it's we we have unfortunately very dynamic well fortunately or fortunately a very dynamic world here so yeah I don't know that that certainty that you seem to think could be brought I don't know that ever could be brought about even by the the the best and the brightest AI system >> you would have to teach it an awful lot of stuff to duplicate the American the the human condition I want to go a step further on That's something that we talked about before the show and that is suppose all the oligarchs somehow consolidated that they decided it was in their best interest to actually make fewer oligarch companies, fewer AI companies and and consolidate their language models and maybe their technology as well and their business models and be one great big enormous um AI company. I don't know what the government would do about that. To me, that would be terrifying. Terrifying. Um, and this little black box or maybe this data center on the top of the Rocky Mountains would have it all. And uh, it would be a single source kind of thing. It's like Sinclair radio uh, the 300 plus stations around the country. They all broadcast the same news with the same words every day. And that has a huge effect on public sentiment. Well, the same thing here. If all of all of human knowledge, not just law, regulation, but everything that we know from classical times forward, um was in this data center black box. Um that would be really interesting and it would give it would give the surviving oligarchs or the government if they control the surviving oligarchs enormous power over human activity in all forms. What do you think? I mean what you're describing is exactly the situation in the people's republic of China these days uh you know where yes that there is a bunch of different aspects of AI going on different research programs etc but it's all done under the opaces of the the the communist party and anything that looks to be uh against their interests is suppressed immediately and anything that can further their interest is developed. And so yes, you have essentially the same kind of thing. It might be private in this in this sense in your scenario instead of a central government. It might be a central corporation. Is that any going to be any better? I think it's going to be pretty much the same, you know, except the corporation will probably tax us more heavily and get more profit out of it. You know, [laughter] >> you know, people have said over the years that Congress doesn't can never really become expert in technology in general, especially computer technology. And and more recently, people have been quoted to say that Congress can never really understand AI and therefore they can't regulate AI. And that that's been kind of the mantra over over the life of technology um more recent technology in this country. Um so what about that? Uh can Congress actually educate itself and become expert? Can it can it you know u train its staff and hire people who will be able to inform it as to what to do in order to regulate or is this an impossibility with with Congress in a in in the best of times um in in this country under our existing democracy? Can they ever really regulate AI? >> Yeah. No, no. I I think you you bring up a very good point and I don't think Congress should even try. I mean, that's Yes. One will hope we have Congress persons who are at least basically scientifically literate. Unfortunately, that doesn't seem to be much the case. But um they they should indeed seek out the advice of scientists, scientific committees like the National Academy of Sciences and set up you know uh as they have in the past central office of science and technology policy that that kind of thing and you know let the experts cogitate on these problems and provide input and advice and then Congress can act on them. No, they they asking them to become technologically sophisticated enough. I mean, that's that's they couldn't do anything else. They'd become technology sophisticated and wouldn't know anything about the rest of the world. I mean, they'd be the guys who are working at Google and anthropic and [laughter] >> what about what about a second black box? second black box and we call that the inspector general of AI and its job is to keep AI straight and honest and moral and ethical and all that and to watch it make sure it doesn't cross the the line doesn't violate the instructions and and is consistent with the constitution the rule of law what have you in this country and it has a lot of power because it is a an inspector general a whistleblower if you want whistleblower AI and he says just down the road from the original AI. What What about that to regulate things? >> I I that's a very interesting idea and I suspect some model like that is going to have to be instituted. Uh because it's the only way that the only inspector general that could keep up with AI would have to be an AI model itself, right? you know, so that yeah, we're our laws have increasingly fallen behind our technologies and that shows no signs of decreasing that trend is w that gap is widening. Our technologies are moving further and further ahead of our laws, regulating them all the time now. And yeah, um that's again that's a whole huge uh subject area that we could delve into at some point. Um yeah >> well you know the one thing I went to a class once at U was about [clears throat] mathematics and predictability as a scientist I'm sure you're well familiar with the notion of predictability and predictions and you can do it ma you could at that time do it you know um using mathematics now you have a much better opportunity to do predictions it's not perfect but you could ask AI for predictions So then my final question to you is uh you know should we be scared I love to put this into perplexity and see what happens. Should we be scared or is this going to work out fine? Uh what's the answer to that question all things considered? I suppose I could go on uh you know Perplexity and Chad GBT and and Claude and all those other ones and ask the same question and see what kind of amazing answer I get in terms of the future of all of this. Is it going to work out fine or should I be running for the hill? Should I be moving to Mars, you know, to get away from it? So, I'm I'm going to ask you as a human being and a scientist, okay, and a thinker, what's the answer to the question? You can be my my perplexity for for this question. Um, I I mean, I'll admit real I'm a techno optimist. Uh, I believe human beings are very smart and very resilient and will ultimately use technology for good. They certainly have proven to do so in the past. We are much better off in so many different ways than we were 100 years ago because of our technology. So, um, but yes, we are in for an interesting ride in the meantime. You know, I mentioned before that uh you know, the competitors seem to be China and uh and the US, but [clears throat] you know, technology does have a way to democratize itself as as is the case in China, at least on the technological side. Um why why can't Hawaii develop some expertise in this? Why can't any place in the country or the world develop some expertise? um they could they could they could buy access to a large language model and develop a better interface. They could develop a better way of analyzing and dealing with questions and answers and uh you know the calculations you need to make. Um is there any you know Hawaii sort of gave up on technology? We tried to make that happen here in the ugh years but it it just it it fizzled um and it has been fizzled ever since. But is it possible, do you think, that a state like Hawaii, which has a lot of smart people, um, could actually get in on things? >> Sure. I mean, you brought it up earlier, Ukraine, Ukraine was not known as a center of, uh, weapon technology development, and now they're recognized as being world class experts in it. You know, they were pushed into it. They had no choice. I mean, arguably, you could say Hawaii has no choice. We we we must do that to survive. And if you can make that case strong enough to the to the government and they'd start investing and bringing the talent in here and building the data centers, you know, Yeah. Hawaii has there's no reason Hawaii couldn't become a technohub, you know, an AI superpower, right? >> Yeah. >> Not good. Yeah. Not good. >> Yeah. >> So, um Okay. I I want to offer you the opportunity to advise our viewers uh what about a takeaway a summary of all that we've discussed u how should they feel about this how should they handle it because coming down the pike you know this this this stream of articles that you see in all the media isn't going to stop it's probably the most steady stream on any subject um how should they handle that how should we see this this this issue and this dynamic >> I mean you know people should keep their eyes open, be aware of what's going on, think deeply about it. How does it how are these things influencing them, their lives, the lives of those people around them? How what are the potential second order, third order effects from these technological advances? How can we best guard against the the worst aspects of it and meanwhile promote the the best uh uses that that really advance humankind in the way we want it to move? Well, I think that's very important to know that the world around us, not only is it changing, it has changed and we we'd better fold that into our world view. All of us. >> Yeah. >> We'll have to leave it there. Ethan, >> thank you so much to our co-host, our chief scientist, Ethan Allen. I'm Jay Fidel, host of Likable Science. The title of this episode has been, "So, what's all this hub of About AI?" I think we know more now. Thanks to our viewers for watching. We'll see you again soon for the next one. Aloha. Aloha, Ethan. >> Aloha, Dick. [music] >> Yes, sir. I believe in the Constitution, the rule of law, and ThinkTech Hawaii in its efforts to raise public awareness in these difficult and unpredictable times.