Is Ai dangerous? Maybe not...

Ah, ah. Heat. Peace. Heat. [snorts] Ah, ah. Heat. Heat. Hey, heat. Heat. N.

Hey everyone, we're live again. Yeah, not a great start. Um, but we're live again. Hopefully all of you uh can hear us. Um, and so what we're going to uh talk about today is AI. And if any of you have any questions about previous videos or other topics, uh feel free to um talk about or ask uh about those things. Um but the main topic today is about AI that we talked about last time. — Yeah. One of the things that I just want to give people a little bit of the logistics here. Um, a few months ago, we made the decision that we would be putting out our videos, releasing them on Saturday mornings, uh, giving people a more predictable ability to be able to find them rather than looking on X or, you know, trying to see if you get a notification. And last Saturday we had a little bit of a glitch uh with YouTube because uh we always publish or post in 4K and uh it was one of our longest videos and for whatever reason we got a little bit of a problem that Saturday so we were unable to release it until the following day last Sunday. So, not a big deal, but we always will follow a release of a video with a live stream that evening at six o'clock to give everybody who's watched the video an opportunity in real time to ask some questions and then everybody can share in the responses and the interaction. So, it saves writing comments down and it's fun. We really enjoy doing this. Well, one of the challenges uh last Sunday was because it was such a long video, you only had really a few hours to be able to see the video. So, most of the people that have seen it over the last week really didn't have this opportunity. So, if we do spend a little bit extra time talking about previous videos, we may want to focus on any questions people have about the wine uh improvement video with vacuum distillation. I It was a lot of fun. It really worked. It's not very expensive. And there's been a lot of very interesting comments about people who made suggestions about things we should do in the future. Try different wines, get somebody with a little bit more talent than us in tasting the wines, blind taste testing. But the underlying theme is that it really did work and it's something that is approachable with a minimal amount of investment and work. In addition, we also are going to be talking about any other questions people have about any of the other previous videos. is if there's something that you uh would like some information about, we're open to doing that. So, there's no restrictions on that kind of subject matter. And then we're going to get into the AI. We've done a little bit of um looking into AI over the last week or two as we've, you know, developed this theme. But, uh I'm sure probably everyone watching right now has had some exposure and some experience with using uh AIS. things like Grock and Chat GPT are probably the most likely exposure even though AI is involved in a lot of other areas of science research and industrial work. — And just to interrupt um Anthony you said train with the three laws uh above all uh we will be uh talking about that and we even already have that — um pulled up here. — Yeah. So, — so anybody as we're going through this, uh, it Wikipedia sometimes hits it on the nail and they did a pretty good review of these three laws of robotics and it's part of the sort of dark side, the sinister side that uh, is making people a little bit nervous about incorporating AI in so much of our life. So, I think it's a good review because it was actually written by Isaac Azimov back in the 1940s. And you can imagine with you know teletypes and the printing press not being that much in their past. It is remarkable how pertinent it is to what we are doing nowadays with the advancement of processing and computer technology. It's getting close to it really matters. We really should think about this. So um also um demand thank you very much for the 20 especially this quickly. — We'll get a pizza. — Yeah. Uh so he said um I think it could have been uh dangerous but it has already been fed so much brain rot that I doubt it and Tesla boat followup when — are we going to do some Tesla boat thing — well not for the next few months uh New England it's winter uh but yeah one of the things that we would like to do with the Tesla boat is uh we need to

upgrade it. We were using trolling motors to power it simply because they were easily available and they worked. Uh, and we actually had an old Tesla battery in there. So, we had a lot of battery capacity and not a lot of power. So, I'm looking at repowering it with something that would allow it to operate a little bit more uh high performance and so we could do more with it. And uh I'll be working on that off and on. I've been touching on that. So, yes, we are going to revisit it but uh repower it to make it a little bit more interesting. um the brain rot. See, that's the problem with AI is that it's rather difficult to know exactly how is it being developed and who's feeding at the information. One of the um sort of standards probably most people are aware of is the large language model. uh that is the guiding principle behind it which is really not that much different than Google searching or Wikipedia searching where the computer has access to an enormous amount of data and it's just how you present it and through the um interfaces like chat GPT and Grock they've made it more user friendly and so it's easier to use it you can do it uh audio uh you can also um use it to discuss issues that you haven't really formulated very well. So, it's difficult to do the search. It seems to be able to dig through the noise in your question and get to the key. So, I find it very useful. But, uh the whole thing with the brain rot is Yeah. But it's also got some good information just like almost any technology. There's there's bad parts to it and good parts to it because it's being developed. You know, the first few Starships didn't fly too well and you know, the first time they ever tried almost any kind of technology, you're going to have rough spots, but there is such a huge potential and a lot of money involved. Uh they're going to be working on making it uh marketable friendly. And so if people have a very negative attitude about what it's going to do, it's not going to help marketability. So there's a lot of motivation to improve that part of it. — Yeah. And uh something uh that I just uh saw here, two things. Uh first, Mike, uh thanks for the two um get the AI drunk. Beep boop. Okay, funny. Thanks. — Yeah, right. — Um just lower its voltage. — Yeah, but um uh Bass um Bassmatic. Um I used Grock uh last week to determine if my mom needed an ambulance. Dot dot. It worked great. Um, if you're fine with it, um, I would like to know exactly like what did it tell you? What do you mean it like it worked? Great. — Yeah. Because — if you're fine with like, you know, revealing that info. — Yeah. Because I would think, you know, if you're questioning whether or not you need an ambulance, you need 911. You don't want to do a Google search or, you know, any kind of a slow search. So, I'm presuming um there's more involved than just, you know, should I call an ambulance or not? Is she breathing? Mhm. — Um so my recommendation in that case just like every hospital and every you know every professional in the healthcare industry would say call 911 we'll figure it out. Um — personally I like chat GPT better than Grock. That's just my own personal uh prejudice. Maybe I'm more familiar with it. Um, but over time, uh, what I've discovered, one of the things that I like so much about these, um, AI chat AIs is the audio function. The fact that I can be in the car, I can ask it a question and it will respond to me and then also give me a text copy of what was said so that I don't have to write things down on a piece of paper. And again, I don't have to formulate the questions so perfectly as you do when you have to do a Google search. you misspell a word and all of a sudden you're searching something you never intended. So that is probably one of the best interfaces uh and best reasons why I've moved so much to AI for inquiring about things we're doing for some of our videos and it's really gotten sophisticated. You know, one of the things I noticed when I first started using it was how human uh the voice sounded and I thought, "Wow, this is pretty sophisticated. This is really good. " But using it over a period of time, one of the things that I started to notice is they keep improving the humanistic aspect of it. And so I was doing a search for a previous video and it was a difficult question because it was an optimization question. You know, there really wasn't a right answer. There were choices. And when I gave it the uh the query, uh it gave me an answer that I wasn't happy with. And I came back and I said, "Yeah, that's a reasonable alternative, but that solution would cost me millions of

dollars. The other solution, which seems plausible, uh, I've got the equipment right now. " And then there was a pause. And I don't know if you've ever noticed this. Sometimes when it's really got to do an analysis, it takes about 3 or 4 seconds. Then it came back and it giggled and then did this human sort of sound where it goes. If you heard that like that little with the mouth, it was kind of spooky because in a few months I've noticed the improvement in that human interface where if they keep improving that at some point you're not going to know who you're talking to when you call a company or when you call for information. Are you talking to a real human being or not? It may not matter, but you may not have an opportunity to be able to say, could I please speak to a person because you don't know. So that is one of the reasons why I think some people are a little bit nervous about it because it's becoming very human in its responses. And I'm not sure if you use the most recent version of chatpt um but it seems to respond a lot faster than it used to. Ah, — if you've noticed that. No, I don't have a chat. GPT4 is the one that I'm using right now. — Four. — Not 5. 1. — No, four. — Oh, okay. — Okay. So, — you're using an old model. — I'm the old man. Yeah. Right. That was good a couple of weeks ago, you know. — Yeah. No, but um what I was surprised about two things that I think is nice is it writes down everything that you're saying on your screen instead of just having that bubble. I get the bubble, but I get the text — Afterwards. Yeah. Yeah. Um but now with the one the 5. 1 it seems to write everything down as you're speaking. I don't think I selected a different setting. Um but it responds almost immediately as if you're talking to someone. Yeah. Not the delayed sort of like, "Hey, could you help me with this? " — 3 seconds later then it responds. — Right. Exactly. I'm used to that. But uh — yeah a couple trials — it is improving. — Yeah. — And one of the things that uh a lot of people in um this field or working uh in this area uh we're sort of trying to predict is it going to become superhuman? be able to think faster than us more creatively than us? And uh I actually asked Chad GPT and I also asked Grock and they came up with fairly similar answers. The prediction was sometime between 10 and 18 years from now, we probably will have um AI systems that are superior to human beings in just about every metric. And I think that's plausible because when you look at some of the uh development of the technology, what's involved with actually these AI systems, they are getting very close to the um processing power of a human brain. and uh I did a little bit of research on my own about that and I've got a little bit of background in that information. So I think it's pretty valid, but right now uh they're not quite there. And one of the greatest strengths or sort of separations we have right now is the number of processors. You know, it's not the number of transistors, it's the number of GPU units that would be sort of comparative to the number of neurons that we have. And even uh chat GPT4 has about 15,000 GPUs that are currently being used when you are asking a question. We have about 90 billion uh neurons in our brain. But the uh GPUs are operating about a million times faster than our neurons. So it's not necessarily an arithmetic, you know, half * 2 equals 1, but it's in the same playing field. We're getting relatively close and it's got a perfect memory which is much larger than ours in terms of its ability to access it. But one of the major limitations and that's where I think they are moving is the adaptability. You know, a human brain when it functions as we learn, as we grow and we mature, uh will begin to highlight and emphasize certain neuronic pathways to allow us to do things more quickly. Like a piano player, when you learn to play the piano, you don't think about hitting the C and then hitting the F. Your brain has uh its hard wiring has become more robust in the areas that you frequently use it. And that's not true of uh the standard types of um AI systems that we are using right now. It is more indicative of the adaptive systems that are used for robots where something is done and if it works uh it is incorporated. If it is

not it is thrown out but to emphasize successful pathways is something that they really haven't uh begun to incorporate. Now, I don't know if that's necessary, but it is something that we do. And it obviously is an evolutionary decision or an evolutionary process that said, "Yeah, if you do this, you're probably more likely to survive and produce surviving children. " So, we've evolved to have an adaptive type of evolutionary learning process in our brains, and AI does not do that. And I'm not sure maybe some computer scientists would have a better idea as to whether or not it actually is compatible with and would be an optimization for the system. But the point is all of this is evolving very quickly. And I think in the future there's risks because obviously um the capabilities of a super intelligence um would be frightening if you're not the super intelligence. Uh, at the same time, imagine how much capacity and how much we could do if we weren't spending half of our lives trying to remember something or figure something out, but had an immediate answer and could begin building your next video project based on accurate information and sourcing and procurement and all the rest of it. So, I'm looking forward to AI and I'm not um naive to the fact that it could potentially produce some significant downsides, but I think the future of, you know, modern society is going to be to incorporate it in almost every facet of life. — And uh Mike uh thanks for the five. Uh trolling is difficult here because you always receive a wellthoughtout response. Love it. Also, singularity is the term you're looking for. — Uh, singularity of um I'm sorry, but in reference to what? Singularity. — Don't know. — Uh, because normally I think of that in the terms of a black hole. Um, — and that was before. — Yeah, I'm I don't know. Maybe if you want to just clarify that in terms of singularity was the word I was using to uh should have been using to respond or to identify what feature. Um, nevertheless, uh, I think some of the things that, uh, we can talk about is AI systems as search engines are great. That's what I use them for primarily, but as they become incorporated in other features like driving and, you know, other operations that also I think is going to have a remarkable effect. Um, how about on traffic deaths? Just reducing traffic injuries. you know, couple of lives saved. It's worth a heck of a lot of investment. — And a uh separate question um non super chat uh from Chainsaw Freak. Hopefully you're still here. Um you said that you upgraded to 96 gigs of DDR4 RAM. Um so I can run larger models locally. Um, I would be interested in learning how to uh what models you're running locally and how to do that because I also just upgraded my computer for all the uh video editing and everything that we do. Um, and I also got 96 uh gigs of RAM uh DDR5 faster RAM, but I'm wondering what benefit that uh like serves you and also what models are you using. So if you can respond to that I would ask one other question to add to your series of questions. Yeah. — Did you just find that it improved the performance of everything that you could previously do or are you now able to successfully do something you just wouldn't touch before or couldn't do? — I'm guessing it would increase that performance because I don't think it would stop. It's like you can edit with, you know, — no, I don't mean when he mentioned that uh, you know, now you can do it locally. What I'm wondering about is — well, it can run larger models locally. — Okay. — So, maybe smaller models, but yeah, I but I'm interested in what models he's using. So, — Oh, okay. Yeah. And then u one of the um interesting things that uh we've noticed also is along with the AI development there's obviously been substantial development in humanoid robots and not necessarily that the humanoid is the key to it but it is obviously highly marketable and I've seen some videos uh both videos from um conventions as well as what appear to be shorts and I have a fairly high suspicion that a number of them are AI generated shorts. Uh they seem too good. Uh there's they're too humanoid. And so I'm not so sure if the humanoid uh robot development is as

good as it appears to be, but if it's anywhere close uh combining um an AI system with uh some uh with a robot that because it's humanoid, it interfaces in our world. you know, you can have a Robbie the robot that's banging through doors and, you know, destroying dishes and all the rest of this stuff. Uh, that's not going to be useful. Yeah, it might be a great, you know, warbot, but if you're going to be purchasing it in order to use it around your home, it's got to be something that can climb stairs, open doors. It's not going to work if it looks like a poodle. it's 8t tall and can hold, you know, flamethrowers. You need something humanoid, not just because it looks like a friend or looks like a member of the family, but so that you don't have to change your living environment in order to be able to interface with the robot. So the closer they come to us, the easier it is going to be to market them and to interface and marketability is key because all of this costs a heck of a lot of money. — Yeah. Okay. Uh so a lot of people are talking about VRAM and everything else, but um we got something from Entity of the Void. Thank you for the 20. Um, a problem with AI is it makes it so knowledgeable. Um, oh, yeah, knowledge knowledgeable. Uh, is not earned which makes people inherently lazy. Uh, this means in a long enough timeline people lose the skill to know how to find information since it's readily available. That's one side of the coin. The other side of the coin would be like I understand what you're saying and I'm not disagreeing with it. I think 100% you are correct on one side of the coin. The other side of the coin is what if those tools enable you to do what you couldn't do before? They augment your performance. As I said, I was looking for an optimization program to be able to build uh a project that we're working on. And I could have looked it all up. I could have spent a couple of days on the computer trying to source the materials and, you know, figure out how to do this. But I got the answers that I needed in about 10 minutes, human time, and then I got to work using that and applying it. So, I don't think it's necessarily true that we'll get lazy. We may just on the other hand become far more productive. You know, we don't waste a whole bunch of time driving to and from work. You know, you don't gain anything from that. Yeah, you might lose your driving skills if you're not putting in a thousand miles a week, but at the same time, you're not wasting gas, and you could be spending that time making money at work or at home with your family. So there's two sides to the coin and that's why the whole AI thing is so fascinating because it's a tool that we don't really have much of a handle on it in terms of all of its capabilities because it's evolving so fast. Chat GPT4. Yeah. Right. And also Yeah. We haven't like tested everything and also other you know places are using different AI models. So there might be ones that are designed for, you know, specific scenarios that you're not looking for, but they might be able to do a better job than say Chad JPT 4 or five. Well, like say for example um conformable AI uh this is what is sort of a category or principle uh different than the large language model is used for um engineering molecular structures like drugs uh where they can produce and um generate an image of the molecule based on all the physics involved in the components in the molecule. and within a couple of minutes they have an idea that should they invest in the machine or the equipment to be able to build this because it's going to be functional. So that is a different type of AI than the large language model because it's made to deal with image processing and imaging and 3D uh visualization of the surrounding world. And like I said before, the transformational or the evolutionary AI that's used to operate machines to make them sure that they don't run into their operators or destroy the room that uh allows them to learn when they bump something or touch something never go through the door that way or you know that kind of thing. So there are at least three different types of AI branches that may occasionally cross paths like we may start getting adaptive uh large language models and we may also be using the large language models to be able to research who's also maybe tried to configure a molecule for the same similar type of drug and it can look up the 50 or 60 different examples that have already been done. So they don't even have to go through the processing power. So there's going to be

this sort of thing as this stuff evolves. And I don't think it's ever going to go to one sort of super intelligence. It's probably going to be multiple different pathways of super intelligence for different functions. We're going to be using different tools, hammers, saws. — Yeah. And uh Patterson um you said, "I like how AI is challenging conspiracy fringe on social media. " Um I agree. Maybe we don't agree on which conspiracy fringe. Um, but I do like the fact that, you know, a certain posts that are either, you know, AI garbage and that sort of stuff or misinformation, um, can be challenged because AI can, you know, look into it and see is this actually true and look through, you know, all the articles that you could possibly think of, — prove whether it's wrong. — The upside is that you can do what you just said. Yeah. The downside is that the AI prefilters it before you ever got to see it. And now you're dependent on the AI not being garbage or manipulated for some purpose so that you never see a post from say a politically or economically unpopular point of view. — So I do like keeping the human interface in there at least the transparency. — Yeah, I agree. Now, one of the um one of the issues that I think uh we wanted to sort of discuss here involves um you know the three laws of robotics. I think the reason that I brought that up because you know I had read the books long time ago is that uh you want to incorporate into these systems fail safes things that keep them from becoming um a negative. And the problem is when they become uh opaque, we don't know exactly how they're working, it's difficult to know if in fact what they're doing is sinister or not. Uh because in fact it may seem on the surface to be a bad thing. One of the examples was there's three laws of robotics and again you can study them, you can look at them on the page, but — and obviously I think a lot of people like myself first heard of it uh from Irooought the movie with Will Smith. Yeah. — Um but if you haven't um — there's a great line in the movie. We can't play it by um uh copyright laws. — But he he's standing there and he's sort of the, you know, cynical, you know, old school, old world kind of person. — And he's questioning these researchers who are building these systems and said, — "You have robots building robots? Oh, that's just stupid. " — And it's kind of like — one of the best lines. — It's lines in the movie. One of the things in the law of robotics is that a robot must always follow the commands of a human being. And the sort of the um the idea is that it will never sort of operate on its own if it hasn't gotten a specific command from you know a human. But there is a zeroth law that's uh sort of a modification. There was a lot of discussion about this over the decades that followed and it was the idea that um a robot will never harm humanity. Uh and they change the human into humanity. And that was where the conflict can potentially come up because what happens if the person that is going to set off uh say a boom device uh that's going to damage a thousand people, you know, it's going to blow down a building. Well, the robot is not allowed to harm a human according to the law, the first three laws. But if it will save a thousand people, that's more like humanity. So now there's a conflict. And when the robot has a conflict, what does it do? Does it shut down? Well, that's also not an option because eliminating yourself from the system could be just as harmful to humanity by not following an order. And so one of the things that we're going to find with AI is that if we don't have human interface, then we do run the risk that it can kind of go off on its own and do a whole bunch of things that we don't understand. And like was earlier suggested, we're going to lose the skills or the will to be able to be cynical about this because we're so dependent on it. We'll be like, well, you know, bad things happen. And so that's one of the concerns that we're going to have to sort of decide as a society or you know what we're going to be doing even personally to protect ourselves from the fact that it's going to be inserted in every part of our lives and may in fact work toward goals that aren't quite as explicit as we would think from the simplicity of those laws. Yeah. Okay. Uh a couple uh super

chats here. Um Mike uh thanks to the five. uh AI will take the knowledge of various professions and help us uh see things from a new perspective that will result in truly novel ideas. Yeah, I think you know I'm sort of an example of that uh just personally background in medicine, background in physics and I can kind of put those two rather diverse sort of knowledge bases together in some of the things that uh I build or you know projects that I've done. Uh but I'm just one person and that's just two areas. But imagine if you had the ability to coordinate a whole bunch of different very difficult topics that a human being can spend their entire life getting good at uh and be able to meld them all into sort of a an approach. The sky's is the limit. I'm really excited about the potential of that. Uh and uh thanks for the 10. Um he said, "I work for a large airline I will leave unnamed. uh we have been integrated a integrating AI to create p uh predictive maintenance using historical trends. it works fairly well but needs refinement just an example — right and I think that refinement will eventually become also part of the AI system you know uh for example you don't need a crash uh in order to motivate that but let's say any areas that you detect that uh it's not accurately predicting and you end up having to go after you know a problem because it was missed um becomes part of the AI system if it's if its evolutionary and its ability to analyze the fact that you called it back up five times to ask the same question. Clearly, there's something wrong in its response. So, does it learn? Does it fine-tune its response? And so, I would think the more you use it, the more likely you're going to have less and less problems with it because it's going to learn from that interaction. It's going to get smarter just like any mechanic in the shop would. You're going to learn how to do it the right way. But just like that mechanic, you know, when they first come on the uh the floor, they don't know, you know, I'm they don't know enough. Uh Jets off, uh thanks for the uh five for local AI users. Um AB, I'm not going to try to pronounce it. Uh AI for engineering and nonbias research. AI uh our uncensored AI use LM Studio to run and download. Okay, I'll try that. I'll look up LM Studio and see what I can do with that. Um — yeah, that's good. — Yeah, because actually I would like to use the power of my uh new computer to, you know, try some stuff. — Yeah. — Um and then Mike again, thanks for the uh two singularity when AI exceeds human intelligence. — Oh, so like singular event. That's what I get by that. Okay. Yeah. — Yeah. And that's you know it's an interesting thing because if I was a smart AI I would have asked me what do you mean exceed sup uh human intelligence. I know more than you do already Eric. So the thing is it's like I don't think the singularity is going to be an event a single time. It will sort of like feelers be better than us in certain areas. It's already areas and these areas will become progressively larger and larger in terms of percentage. At what point does do you reach the singularity? Is it only when every single thing that we do it does better? — Well, like Joe Shimo just commented in capitals, no, it means when everything becomes one. But I think that's sort of the point. It I think it's becomes equal. I don't know what you mean by becomes one because see we're not tied into the AI. You're going to even have multiple AI systems. Some better than others at certain duties, maybe just some better than others. — So is it when one AI developed system is better than a human being in every single regard? Is that the singularity? Okay. Well, then does everything become one? No. you're still going to have competing systems. Maybe superior systems for certain functions are just faster or they're cheaper or whatever. But also, how would you ever know? Like, you know, it's one of these things where that transition point, even if it were a point, you really you wouldn't know because just like being on the inside of the event horizon, you don't really know you're there. It's an interesting cosmological theory, but the point

is I don't think it's that simple. I don't think it is like there is some, you know, mastermind, some great evil scientist who's developing this stuff and as soon as it gets really smart, then it decides to, you know, destroy its creator and go out and try to find a mate. You know, I don't think it's going to be um that simplistic. I a very messy evolutionary broadband type of approach and we'll probably look back in 50 or 60 years and say it was probably in the 30s that something like you know what was called the singularity kind of happened. But by now you know 50 60 years after that it is. So it it's going to be an interesting thing because it is hard to talk about stuff that's smarter than you are by its very definition. You can't, you know, put it under a microscope and say now we understand it. We can't understand it because by definition it's smarter than us. So if it didn't want to, it could keep us from knowing. And would it want to? Would it have a want? — And uh Chad, thanks for the uh 20. I'll have to uh you know moderate this but uh thanks for the 20. Uh you were assuming AI is uh programmed correctly. Malicious programmed AI will lead to images of George Washington being BLK. Uh in the future if you didn't know that was false. History can be re rewritten. Yeah, [clears throat] that's true. Um but obviously there's going to be other AIs and also yeah you can always look at textbooks and everything. So it's like it I wouldn't trivialize that because — well I'm not trivializing it but also I wouldn't put it so far where it's like everybody is going to be you know in the pod you know and that that's all they can rely on. Because when you think about it, I mean, the AI systems that are out there already right now, we have several different models out there and they're all probably going to survive for a long period of time. And the malicious, the scary part, the, you know, Iroot sort of fear. Yeah, that's why people are concerned about it that it's going to not rewrite history, but manipulate it in such a way and toward a certain goal, toward a certain motivation. whose motivation and even if you could identify that again there's this mad scientist that wants us all to think something different about the past there's going to be a lot of inertia because of not just the textbooks but other AI systems other evidence and so it would have to so manipulate the world to completely surround you with false information that itself is self-consistent and that's one of the problems with sort of um lies when you tell a lie Ultimately, the lie cannot hold up because it does not meet uh this the standard of holding up against all analysis of this of the lie. There's going to be something about it that is not true. And if that's the case, it's going to be vulnerable to discovery. I don't think we're going to end up in a uh dystopian world where all of our history has been rewritten in such a way that we would not be aware of the fact that there's cults out there and there's goofy people out there with some weird motivations um that have successfully done this. Generally speaking, no, it's over the top. — Okay. Uh two things uh before Jeremiah uh before I get to your $5 uh thank you for the super chat. Um, I saw uh something here that I think is interesting. Um, so Anthony, uh, you said, "What do we have to detect AI deception? " — Um, I I've looked into that and in the previous live stream that we did. Um, it sounds dumb, but it's actually probably a pretty good way to go about it is using dumber, more simple AIs to review the code um that smarter AIs have to see if they're, you know, accurate or um deceiving you or something like that because obviously humans can't go over all the code because there's so many billions, trillions of lines, there's no way for a human to be able to figure that out. So, use more simple AIs that aren't learning to detect any sort of falsehoods or — I didn't watch the movie, the iRoot movie. I didn't watch it. — Well, it wasn't about that, but — but no, but wasn't there a hero in the movie that was a robot, but it was really primitive? It was like a dog or something like I may be remembering it wrong. — Yeah, it's the opposite. It was a smart It was one that — like an older model. — No, Sunny. It was an older model. It was a newer one, but that actually operated on emotions. — Oh, okay. — So, it used emotions to save but to

your point, there were a bunch of the dumb older robots that were very simple and not, you know, controlled or whatever, uh, that tried to save humans from the newer ones because they were — they were being wiped out, but they were trying to defend them. Yeah. — And so I agree with you 100% because I think the competition is the way to do it. — And yes, but it wasn't really what my point was is not competition is you're using ones that are simple that aren't learning. Uh you're using them to sort of snitch on the, you know, larger language models or larger AI based stuff. Um but before we get into that, uh Gemini, yeah. Uh Jeremiah, sorry. Uh, thanks for the five. Uh, these laws are pure Hollywood fiction with no basis in reality. The technology decides when the robot army switches to evil, quote unquote evil. Um, we weren't talking really about i, robot. I just brought that up because they use the three laws um, from uh, what's his name? — Azimov. — Azim. Azimoff. Yeah. So, that's why we uh, brought that's why I brought that up. Um I'm Hollywood didn't create that — right the um the idea of the simplicity you're correct about that the one sort of failing point there is that as we move forward — uh the simpler AI systems the more um sort of difficult to manipulate from you know a central control or something like that are themselves going to be far more sophisticated than our current uh AI systems. So, they're all going to be marching forward. evolving. — Well, not necessarily. If Well, if you use uh like for example, like everybody complains about Siri now on iPhones. I've never had an iPhone, but it's so basic that it's just programming. It's not learning anything, right? Like other language models or other AIs do. And Mike, thanks for the two. You're saying the same thing. What if dumb and smart AI collide or uh collude? Um I get that. But if you keep the dumber AI dumb and you can always reinvent it, you use that dumb AI to detect if there's something going on with the smart AI by looking over all the code. Okay, — I've got a problem with that because see the thing is the smart AI can manipulate the data sets — that are available to the dumb AI. They can change the historical record. They can add information out there for the express purpose of manipulating the AI police force, the AI, you know, regulation. — I get that. But the dumb AI is still smarter than humans. So, what what's the best bet? to use humans to go after the smart AI or to use the dumb uh um dumb AI to go after the smart AI. — I think there's a parallel way. Two smart AIs, the likelihood of them sort of operating uh in conjunction with each other as opposed to in competition with each other. That — because the dumb AI can be easily manipulated by the human being that built it — and by the AI system that's trying to hurt you. — No, I totally disagree because for example, you have Target and Walmart, okay? — You know, uh competing with each other, right? — Are they going to try to side with one of the like small mom and pop business? No. They're both going to get rid of all the small uh mom and pop businesses and then compete with each other. — They're going to take out everybody and then compete with each other. — I stand corrected. I think you're right. Um, but I'm not so sure that the because the solution that I just suggested, — you're right. I'm wrong. — But I don't think that other solution, the simple AI, is going to be robust. I think it's manipulatable, and I don't know that is going to be a defense against a malevolent AI. I don't know. — I I'm not saying it will work, but I'm thinking that's the best case right now. I don't What about the idea of as they make the AI systems uh more efficient, more compact, processors improve in speed, you've got your own AI? What if instead of two or three, you've got 3 million of them? And so for my questions, my searches, my research, I can use my own AI and it can be buffered, it can be um — restricted, it can be limited. So I never really depend on the global AI for my source of information and it it's also not able to manipulate me directly. — Yeah. But like what you're talking

about though is how many GPUs they have um versus you have one — and yours is so basic versus they have server farms where it can do whatever your AI locally could do in a millisecond that it might take yours 9 hours to do. Right. Exactly. If it's an important question, it might be worth waiting nine hours. Um, you know, if it's a real really important, should I marry her? You know, that kind of thing. I mean, if it's a really big — Well, you know, yeah, — I bet. Come on. Honestly. — Oh, yeah. It'll come up with reasons. — I bet you there have been some marriages that were checked uh beforehand. Yeah. — Actually, yeah, they probably have some. And even as simple as asking ask asking chache PT like well she has these good qualities but also she did this in the past and blah blah is marry — and then the AI system what it'll do is it'll bring up her entire history for you things you never even knew and she didn't tell you — and I keep saying she because it's two guys here — if you're watching you can study to see should you marry him e you know too so let's be fair about this both ways — yeah any anyway anybody. — Yeah. — But those are the kinds of questions where like if it's really key then maybe there is uh it's worth it to have some sort of a system that you have that can buffer those kinds of decisions. Uh even if it's very time consuming because you don't care the result is that important. — Mhm. — I I'm just looking. — You checked. — Oh no it's just um wage. Uh she had COVID. Should I marry her? It's like, — "Yeah, but did she get the shot? " — Yeah. But you don't know. — Does AI know? And also you're trusting AI at that point. — Yeah. Right. Exactly. — But and the only way it would know is if it's doing some, you know, background research that like if it does know, that's even more scary. — Right. And then you keep looking farther down the list to see what else — comes up, you know. — So, and well, and uh Danny, uh thanks for the um I don't know what currency that is. Uh I don't just give us the digits. — Yeah. 100. Okay. Um Czech Republic. Okay. Thank you. Um thanks uh for your videos and live streams from the Czech Republic. I'd love to hear all these uh other angles and insights into AI from the people in chat as well. No question here. Keep doing what you're doing. Thank you. Yeah, we will. We really enjoy this. This is fun because it's one of those things where — building the projects, teaching, learning, you know, having fun with that kind of stuff. It's a lot of work and it's, you know, it's very gratifying. I like the immediate uh sort of interaction, the the ability to talk about this stuff because we aren't actors. What you see on the camera, whether it's the video or whether it's this, we're telling you what we think. And this is an opportunity to be able to talk about anything, not just about wine or graphine, you know. So, it's fun. Yeah, it is. Uh, so zero. Thanks for the five. Uh, AIs are the result of intelligent design, not evolution. Give them no reptile brain. Ambition, selfres preservation, easier to raise than children. Um, there's a problem there. Um, and it's in the three laws of robotics is that a robot uh should operate, I think it's the third law to protect itself. Um it and I think it is a reasonable law because if you've got a billion dollar system uh that is operating you know airplane maintenance and you know hospital census and you know the military and checking weather and doing all this kind of stuff and then if for whatever reason it doesn't uh protect itself from say a virus or it doesn't protect itself from an attack by another AI then you lose all the positives. So there is some argument for the AI to be somewhat self-defensive or protective. And once you get into that, then how sophisticated are the attacks determines how sophisticated its protection can be. And you could see how that could eventually push up against the first two laws. And it becomes primary because what if you're running the entire world and somebody has to be allowed to uh unal alive themselves in order for it to continue functioning and doing all these wonderful things. It's more or less, I think, a matter of prioritization rather than just do not give it the ability to protect itself. Free will. Um, you know, sense of humor. Yeah, I don't see a problem with that. And then, uh, Mike, uh, glad to see you again. Uh, thanks for the two. Will AI finally push us to fix our electrical grid? um — what's wrong with it?

— Well, obviously we are sort of uh towards the limit of what we can produce. I don't know why we don't just build more plants, — but I'm guessing — versus source. — Okay. — I mean, I think — I think the main problem is the source though, not the grid. — Well, the Well, he said grid, and that's why I was sort of — Yeah, but I think Okay. Well, that brings both. — Well, for example, controlled nuclear fusion uh probably going to be the future of energy eventually. Like I predicted, you know, six years ago that it would be here in 5 years. So, I'm one I joined one of the many people who are a little bit too enthusiastic about it. But fusion, I think, will be a source of energy once we get a handle on this. It's incredibly abundant and it has the benefit of it being so hard to do that it doesn't it's uh very difficult for it to become unstable so you don't end up with uh radioisotopes that can be used for other purposes nefarious purposes. I think the power source is less of a concern than the grid sending the power around and I think they already use fairly sophisticated computer models for adjusting the grid. you know, it's a hot day in New York City and so they're going to have to lower the voltage by, you know, 3% or something in order to be able to continue to function. — So, I do think AI would be an excellent um solution for that because of its adaptability and its huge capacity to analyze all of the ramifications of any kinds of changes. Um, yeah, I I'm almost positive that I'm almost positive they're already incorporating it uh just based on the fact that it would be so useful and they're already capable of doing a lot of these things. Something that I've always thought uh thought about though um is why don't we use more hydro electric like power sources because the nuclear everybody like is scared of that coal pollution — fusion is all that sort of stuff like hydro always seems like the most feasible method — like where — it might harm some local ecosystems slightly but you can transplant them like why don't you build more dams. Is there a reason? — I think there is. I think once I read that uh scalability is a big deal with hydroelectric that um as you make them bigger, they become cheaper per megawatt hour or whatever the capacity is going to be. — So it's like the three gorges dam, you know, all of these sorts of things where they build the most enormous ones they can to get the most efficiency out of them. And there are limited locations where you can do that. — Yeah. But you can do it in small like even small rivers like one that I know of that's near my uh house. Um that generates like 12 megawws or something and it's been there for like 80 years. — Yeah. — Um I don't like I don't say what and just build more of them. — I don't disagree. uh I'm not knowledgeable enough about like you know are we maxing out and it's mostly the quote environmental risks and there are some like you said I don't know if they're as serious as many people uh comment on plus I'm sure there are also some political risks associated with it because when you put a dam in one location you're going to affect the water flow in the river below it — and so how many people are going to be negatively infect affected and now you're in court for 15 years and you know can you keep the project going. So I think it might be partially the localization of this even if smaller microhydro you know that sort of thing uh is a good alternative I think the major dams are probably faced with political environmental and just long-term planning issues that make them very complicated. One of the things that AI could do is it could analyze every single river almost to the centimeter — and figure out what location is best where could you do it and then it could even incorporate political issues, you know, environmental possible impacts based on the historical result of building a similar size dam and within about what a second — you could have a map of where you could put the available microhydro units. Then you put up a business that's your business. you start producing microhydrop plants and you know — unless its intent is in the future in 50 years is to unal alive that population so it places it in a perfect place where if the damn breaks — but we need to do it now before it becomes you know self motivated to do that you know — right — um invasion thanks for the five um I'm in school for engineering and I'm aiming for analog to make it much more reliable uh to rival current ARC architecture. I'd also like to see AI on this.

— Yeah, analog is really interesting because you always think of like digital as being sort of the new thing. It's always better, but it is discreet. And the idea with analog is it's sort of infinitely adjustable. There's no stepwise pattern. And I think it was suggested, and again, you're in school for this. I never studied this so I'm speaking kind of from memory and very limited knowledge but there were some suggestions as analog being useful for a number of functions even better than digital because of that uh infinitely adjustable capability and it might be from an AI point of view a very interesting method for decisionmaking because you don't have to get to 92. 5% um uh confidence level before you say the answer is this. You give a progression. And so the pro the progression is a likelihood. It's more likely. I'm not going to give you a number. I don't have a number. But it's more likely than this. And so that nondiscreet kind of uh function might actually make it far more powerful. might not interface with our current huge infrastructure of digital unless you had some sort of translation system. And I don't mean translation just in language models. I mean even in industrial controls you might need to be able to interface with digital systems. I'd be interested if you don't have to send a super chat, but if there's a way for you to identify it, I'd be curious as to what type of systems you're seeing now that are using analog for um practical functions. I would be fascinated by that because it is an interesting forgotten technology and I don't want to go into it uh too much so I'm not going to read the whole thing plasma. Um, but what do you think about verifiable elections? Like using AI obviously. Um, that would be a like actually an interesting thing where if you could use AI to, you know, sort of look over all of the votes, see sort of like where they came from, uh, the discrepancies in certain areas and just sort of like give a check mark that it's like, yes, this vote makes sense, that sort of I think it would be an interesting thing for an uh an analyst to do or say a news um group to do. I'd be a little bit nervous about that would be one of the first areas where if you had a malevolent AI, — it would try to Yeah. Exactly. — That's where it would start. You know, — whatever politician wants AI, — it's going to He always wins. — It's weird how that works. — Yeah. By one vote. Um, thanks for all the uh compliments, guys, and everything, too. Um, and sorry if I'm not getting into every single one. Yeah, they're coming pretty fast. Um, and I can always see these like from a distance and I just see them scrolling and I watch a lot of uh podcasts and you know like uh online uh discussions and I see these things going up so fast and I you'd need to have we've got a moderator but I think it would be very difficult to be able to pull out um you know the information as it's coming so fast. So, forgive us. Sometimes you pro there's probably been a lot of very good comments that we just we don't see them because they're coming so fast. So, we'll do our best. — Yeah. And I try to scroll uh to different ones, you know, whenever we're not talking about something related. Um so, Michael, uh this is unrelated. Fusion generators have virtually no common downsides with fision uh reactors. fears associated with nuclear are uh — translation — stabbed uh incorrectly between the two very different technologies. — Right. Yeah. The uh the only uh sort of radiation uh problem is the shielding. Um I've followed fusion obviously I'm kind of interested in it and um there is radioactive reactions going on inside of the plasma and when that occurs it gives off a lot of gamma rays and energetic particles and so sometimes it's lithium sometimes it's burillium uh that lines the chamber in order to block some of this radiation and it becomes radioactive. So ultimately you would have uh radioactive uh disposal to have to worry about because the radioactivity actually does damage the shielding too. It changes the atomic structure. So it's no longer say pure burillium. It's an isotope. So eventually they have to be replaced and it has to be disposed of. Uh on the upside, it's very difficult to

turn that, you know, dirty shielding into some sort of a weapon or something destructive, but you're still going to have to deal with some radioactive waste from the systems. But compared to sort of changing out a core, uh different ball game. Anton, uh thanks for the five. Um have you ever used AI to uh to design a work breakdown structure or write code? I've uh we've never used it for a video. So, we've never used AI for any video stuff or anything like that. Um, but I have done some tests like with Chat GPT probably 9 months ago or something like that to just see if it could program like very basic like could you make um the classic game of Space Invaders and write it in code and then I would copy the code, put it into Notepad++ or something like that and see if I could get it to run and most of the time it sort of worked. Um, but we've never used it for any sort of video related thing except for — Chris, but he wasn't using it quite in the way that you're probably thinking of it. — Yeah, because he was using YOLO to basically, you know, it's an already written program. You only look once, image analysis, that sort of thing. I don't know that I would call it AI. Um, but the what I was going to bring up is you and I haven't. Um but Paul uh he might have and if he hasn't it might be something if we have a live stream next — I'm guessing he probably has cuz he has chat GPT so and he's a programmer so if he hasn't used it I' I'd be kind of shocked but — yeah and uh if uh he joins us next Saturday uh we could bring that question up with him because if he has it'd be interesting to hear the history if he hasn't it'd be interesting to challenge him to say you know could you write that code that you just did better with, you know, AI and show the result and maybe do a, you know, a comparison because, yeah, I would think it's obviously used for writing code. It's used to write code for AI. So, yeah, there's no question about the fact that it's used. We just we're not coders. — Yeah, exactly. — Um, let me see if there's Yeah. Okay. Uh Steven or Steven. Uh we need uh AI to filter out AI comments but can't trust an AI mod. — Yeah. It's AI and AI. Yeah. Okay. And currently the mod uh Sunnyside I think um she is doing a fantastic job. So anybody who is uh saying she isn't you can leave. She's doing a fine thing. — Yeah. Fantastic job. and we really appreciate it sometimes. — Oh, and also Joe Schmo. Um, Frank, I did an a game in AI. — Okay. So, he did. — So, um, Paul, if you do join us on Saturday, uh, be prep be prepared uh, to give us a little bit of insight into that because I think people would be hot about it because what you do with coding to me is just it's another langu. I mean, it's just it's another world. And if people with our level of and certainly more and less uh capability to do coding could uh use AI to be able to do some of the things that you've done more easily or maybe more transparently. That would be really fascinating because coding is cool. I just can't do it. It's like playing, you know, a pipe organ. I can't do it. But the thing is it would be it'd be fun to be able to — and it would be cool if like you can see obviously I have this chat GPT thing up here. Um Paul I don't know what exactly like what language you're most used to and everything. It's either you know C or anything like that. — Pure basic. Yeah. That is Yeah. This is go-to. Um, if you would be able to do something sort of live, you know, where you could put it into AI and then sort of, you know, demonstrate, okay, AI was like while we're talking was able to code everything. — Yeah, he's doing this and we're doing it. — Play a game. Yeah. Like, — yeah. Right. — After 20 minutes, that would be a cool idea, I think. — Yeah. Another thing, too, is debugging code. you know, you write the code and then AI goes through it and rather than spending two weeks trying to figure out why you've got a problem, uh it could probably greatly uh accelerate the finishing of the code and remove the frustration. Um I think uh AI is not yet sort of like in the physical world until we get the robots out there. So it is going to be intellectually based. What can you know do with your brain, it can do with its brain. — Yeah. Right. I'd like to see it compose music. Um, is it dull? — He said he did the game in pure basic.

So, yeah. Right. Of course. — Yeah. But, um, like music, I was just thinking about that is like, you know, the thing about music is, yeah, there are rules and co, you know, there's chords and again, there are rules just how you put together a concerto or song. And obviously you could give it the rules, but does AI produce very bland, muted sort of follows the rules, but there's no uh intuition? There's no sort of that spark uh that gives makes the song, you know, a blockbuster. And that is often when you listen to a song, it's like, "Oh, yeah. I remember that. That's really great. " And you listen to the chords and the changes and everything like that. And it's the same as everything else, but it's got that spark of inspiration that maybe AI just can't do and may not be able to do for a while. — Might be an interesting test of AI. — Yeah. And I can't pronounce uh your name, but you said keep it away from music. Um, I agree. Like I would prefer for all music that I listen to not be AI generated, but you have to admit some songs that you've listened to, especially when it's very specific to like a topic or like a YouTube channel is doing intro where they're on a trip or something like that and it can integrate every part of that trip into the song and have it, you know, rhyme and all that sort of stuff. It's pretty good sometimes. — Yeah. And is it is your concern because you're a musician or you're a composer and it's a matter of competition or you really just think that uh every time you've ever tried it or you've heard it, it as I said it's just not inspired, it's not worth listening to. Uh or is it too good? I don't know. Um but these creative arts like painting uh a portrait uh you know to make it beautiful uh a landscape scene. I mean you could make it accurate. You can take a camera but the thing is when you paint that same landscape the human brain interfaces with it and um you translate the emotional effect the sort of the impression that you get from it. the impressionists, for example, um AI could imitate the style of a human impressionist, but that's not what we're talking about. Can it actually uh express something that isn't just in the pixels? — Okay, two uh two things, Steve, thanks for the two um not directed at you, just amusing um prediction. Um so that was referring to a previous comment. Oh, — okay. Uh, and then one other one that I saw from uh Joe Mackey. Um, can AI invention be patented? I don't think so. I don't think it can. — I don't know. What if you own the AI system? — It's like I can use a computer to design one of my products. — Maybe if it was a local AI like we were talking about before where it's like you installed it. It's on your computer. it's not owned by anybody else and you sort of developed the thing whatever it comes up with I you probably be able to patent that but if you're using a public sort of resource even if you pay for it I don't think you would be able to pretty good question because when you think about it like for example if I hire somebody to work for me and they work as a lab technician and we build something like that. I patent it. They don't. Even though they might have done the bulk of the work on it and a lot of the creative input came from that employee of mine because they were working for me. So if I use a tool that does the same thing. I use my computer and a program that I bought from Microsoft or whatever. It's still my patent. It's I own it because the intellectual property was created under my investment. my control. — Um, the AI system, you brought up a good point, the public nature of it. It's like if I go to a public library and I look up the physics behind something and build something, I can patent it just because I read it in somebody else's book. — But you had to do work. The library didn't build it for you and you just took it. — No. And the AI isn't either. It's — it's a it's anformational resource. — Well, not necessarily. Like it can do programming. So for example, say it programmed an entire like computer software, a game — and you sold it. — Yeah. And sold it like — and it's copyright. — Is that copyright? See that that's where it gets because you actually didn't do any real work. — Do we have any patent lawyers out there? Could somebody give us feedback? because this is an excellent question and not that you know we're in the market

for doing this but I think it's an excellent question because you can imagine the wet towel effect that would have on people using it for a lot of intellectual creativity if it's no longer theirs. you know, they came up with the concept and then AI uh integrated it into a system, perfected it, uh error checked it, um showed you where to buy the stuff and basically told you what to do and you've become its human robot that's, you know, functioning. Does it belong to Grock? Does it belong to, you know, or does it belong to you? But then also, but it's I think uh like even let's go more extreme. What if it's something where it actually like you know Sora, you know the video generation stuff? — Um, so what if you have it come up with an entire movie, say you paid thousands of dollars where instead of, you know, 12 seconds or 10 seconds, whatever it can come up with. Um, you have it do, you know, an hour and a half and you have it come up with a movie. Maybe it's not perfect, whatever. And then you take that and just based on your prompt, that's all you gave it. And then you take that movie and you sell it and you copyright that. Is it yours or is it owned by OP open AI? Yeah, actually — we could inquire or you could inquire too. Uh but it is an interesting concept because at some point um you know the property you know the whole idea of patents and copyright is that if you create something obviously you had to invest time and money in this you took the risk and so there has to be some reward for that financially to keep the company going or keep you know financially sound. So that's part of the reason you have copyright and you have patents. So if you lose that because you use a certain tool, that would seem to be kind of a strange interface. And it brought up another thought that I had is what about liability? What if a company uses a an AI system to design a drug and the drug turns out to like theolyomide or something like that have very bad consequences years later? Even not under patent anymore. Who's liable? the AI system who messed it up or you who didn't follow up with proper testing or both of you. I don't know. Personally, I think it would make sense that it it's perfectly copyrightable if you inquired uh because it's like the public library. You're just using a tool. But I might be wrong. Well, I if someone was to be incriminated like if you like say a doctor asked chat GBT if this medicine would work on this patient never used this before. — Yeah. Right. And it's like well he recommended it and the patient dies obviously the doctor is going to be liable because also chat all these AI you know open AI and you know uh — there's your answer all of them uh have no but they have warnings saying our results are not always accurate you know so they have warnings about that right why — but well because it's not always accurate — no because why are they telling us this — means they don't want to be held liable — exactly Yes, sure. But I don't know how that would apply to — um patenting or copyright. — Well, I think what it is that they've already analyze the risks of using AI in a environment where somebody or something could be damaged or hurt and as a consequence there's liability. If I'm a little doctor in a little town and I do something horrendous, um, they can sue me for everything I'm worth and it might be $50,000, but if they could go after AI for $50 million, you bet you they would do that. And AI probably has got to be aware of the fact that if they're held liable uh in order to maintain copyright control and be able to uh patent objects, they may not be able to stand up in court by saying, but they're not responsible for anything bad done with their product that they received um royalties for. So it I don't think they could have it both ways. They can't protect themselves from liability, but maintain patents and copyright protection. I don't think you can have it both ways. And I suspect the warning is because their lawyers told them, you know, you've got to warn people that it's not reliable and you can't blame us if it doesn't work. But at the same time, they can't say, "Oh, yeah, but we did this, so it's ours. " You can't have it both ways. That's not logically very consistent. — Well, maybe it's not logically consistent, but I think that does work that way a lot of times. So, for example, say you get like video editing software, um, and you don't, you know, pay the premium for, you know, say it's

$400 a month or something like that. You pay the lesser version that requires you to have a watermark and site them and give them credit and all that sort of stuff, but they still also have warnings saying you can't do explicit content with our software, — right? And so they're giving limitations and saying you can't do these things with our software, but when you do make something with our software, you have to give us credit. So it's like back and forth. You know what I mean? — It's it Yeah, I do. And I think I was right, but I think you're right, too. I think it's I think there's sort of it's a grayish zone because at the same time um no one's going to be using AI and they're not going to sell it if everything that you do uh is going to belong to them. At the same time they're not going to be producing stuff that could be uh challenged in court because they really messed up and they the warning wasn't quote sufficient uh to protect the public from their mistakes. And so it may be that is a hot topic in AI is the liability protection and the copyright and patent protection. Okay. I don't know if this is true and if you can um you know give more information but um you said open AI offered to pay for chat JPT copyright lawsuits. Um obviously OpenAI owns chat GPT. Um, but what do you mean by they offered to pay for their lawsuit, their copyright lawsuits? — I didn't know that they had any. Um, — you in theory like in potential lawsuits. — Yeah. Well, I would like to know. — Yeah, — because that's all that he said. So, if you could — Yeah. — Oh, you said this was a few years ago, though. — Another thing. — So, if you can give an update on that, I'd be interested. — And that's another thing, too. Imagine like Hollywood and uh you know the production of movies whole uh music industry. What are you going to do when um somebody can and you don't necessarily have to have an AI system create a complete movie by coming up with everything. But what if I'm sitting there and I got this great idea. I'm going to, you know, make this uh this would be a really neat movie. and then I run it past um one of the AI generation systems and they produce an entire movie. With very good AI, it's going to be close to competitive with actual Hollywood and uh music production. How are they going to survive? I would think they would be nervous as heck right now because you're going to have more movies and better movies that at least contaminate the market. — Uh and same thing with music. It's like most people don't buy music now. I think most people download it or they just listen to it online. Um what if the AI is being is able to provide this in such quantity that you can't even find the good musicians, the good composers anymore? The you know the actors are all AI improved um generated characters. You don't have to pay actors. You don't have you know that sort of thing. I would think that would be an industry that's probably looking to diversify right now. — Uh Steve, thanks for the uh five. Um he said uh to be obnoxiously clear, your talk has made it absolutely clear to me that we will soon have half if we don't already a war between AI commentators and mods. Uh first, we don't have any AI mods. We only have one real person. um and uh the commenters um can't do anything really about that. So, — well, if you if there's a war going on, express it to us because we can share it with everybody uh rather than the person who sees the war going on. As long as it's, you know, it's not violent, it's not, you know, uh risque or anything like that. If it's something we can talk about, I'd be interested in any kind of, you know, conflict or, you know, what's going on in the chat. So bring it forth and let's um let's discuss it. — Yeah. And I think most people ignore it and I ignore anything that isn't related. — Yeah. — So — not goofy comments about you know what beer you like. — Yeah. Uh just looking wanted to see if I got a response from that last one. if you want to keep uh — well, one thing I was going to mention is I got a comment earlier on the wine video that we just did and I'm already setting up the system to work on a white. So, uh, if it turns out that it's pretty successful, what we might do is we might revisit that topic in a couple of months, maybe even a couple of weeks and bring in some somebody or some people that are a little bit more knowledgeable than us and try some blind taste testing to see if we can really nail that down because it is a great tool in my mind and if it holds up with

multiple different types of wines, uh, it could be really exciting. — Yeah. And Joe Shimmo uh said I think he's talking to other people in the uh comments section. Um but Newgrounds uh had some great underground music. Um I used to remember using it's completely unre unrelated and I don't know like how popular it is now but Newgrounds used to be a great place for you know independent people to you know post music or programs or games or anything like that. Mhm. — I don't know if they still do the same thing, but — things like that might be, you know, sort of not taken over, but — made easier for people like that if they come up with some sort of AI game or something like that they can do instead of take six months to do that they don't see any money or revenue from they can do in two days and upload to there for oh check out my game see what you think about it and then if it gets enough traction then maybe they develop enough to then, you know, post on Steam and actually make some money from it. — Right. So, — right, — something. — Yeah. Lowering the threshold. I think AI is going to lower the threshold to entry for a whole bunch of intellectual um — operations. Okay. Uh Steve, thanks for the two. Uh again, uh I just meant the human race in general, guys. Um, I assume you're talking about the comments, not us, but — um, okay. Uh, Mike, uh, thanks again. Uh, anything exciting coming to solar panels or renewable energy? Probably not soon, just because also it's the winter here — and so we're not obviously wouldn't do anything solar related unless you would be using snow somehow. Um — h — yeah you could use the reflectiveness for — a lot of the ideas that we have here come from these sorts of discussions where it's like could you do that? No I — wait a minute. — Well actually yeah right exactly uh not specifically. We've got a few good projects that are on sort of on the build and you know accumulating parts and figuring out how they're going to work. Um the problem with the solar panels, there was a recent uh article in um Max Futon uh Max Photonix and they were talking about the use of um Peltier devices uh that are getting more and more efficient. They're notoriously inefficient, but they're nice because they're solid state coolers. If you're not familiar with them, uh they're the same thing as uh thermmoelectric coolers. You put a current through them. They often look like ceramic plates with these grids of metal between them and one side gets cold, one side gets hot. They can be used frequently in computer systems and small electronic systems uh to locally control temperature very accurately, but that you pay a lot in efficiency. they've gotten better and there was a recent article about using them in uh conjunction with solar uh arrays or solar systems because solar needs direct sunlight. It needs certain energy in the photons basically visible light but these work on heat energy. So even though they have a low efficiency, they can be combined in the same ground space that you paid for, but they can actually produce energy as long as one side is warmer than the other. And the one side that's warmer could be the snow versus the night sky. A lot of people don't realize this, but the night sky is incredibly cold thermally um from the point of view of wavelengths. So you can they'll operate at night. So maybe you only get 5% of the energy, but you get it through the night. And that's going to be an interesting modification that I think could be um added to existing infrastructure. So there's, you know, very little startup uh expense involved in that. You don't have to buy new panels. Okay. And from Anthony, uh just interesting question. Uh why no gasoline fuel cells? The Bloom used by Google was said to work with gasoline, too. Would uh would be better than batteries. Still use gas power. That's interesting. Um, yeah, you could theoretically use it. I think most of the time they use um, you know, there's the hydrogen and oxygen simply because you're producing water and it's clean. Uh but when you combine oxygen molecules with gasoline or methane or any other kind of material um there's energy release and if it's done slowly it doesn't have to burn. It doesn't actually have to get incandescent. I'm not really sure of the catalyst that would be necessary because many of these things require some sort of a catalyst in order to lower the um potential energy necessary to get the reaction going. That's why a flame uh

you put propane into the air, it doesn't burn. You need to initiate a burn and the heat of the burn is what then initiates the next molecules to burn. It overcomes it increases the kinetic energy and gets them to slam into each other. So, I think you could do a fuel cell with anything um that uh is available, but there may be problems with fouling. Uh gasoline, unlike hydrogen, uh could potentially produce soot that gets on the catalysts, which aren't cheap. Uh they're often platinumbased or palladiumbased. So you can imagine you don't want to get them dirty. So it might be just fouling that's the major problem with that. — Uh Sanford, uh could heat pumps replace boilers? — Sure. I they do, don't they? I'm pretty sure they do. And I don't mean to like trivialize it, but um when you talk about a heat pump, we demonstrated an air conditioner augmentation system based on a heat pump. Uh we actually upstairs here showed you a heat pump to warm a room using the hot side of an air conditioner. Uh but there's no reason why a heat pump couldn't uh operate um a boiler. uh you just have to extract heat from the environment and get it hot enough to bring it to say if you're calling it a boiler the boiling point of water which would be quite a jump uh but still you are extracting some heat from the environment. If you're talking about a water heater where the requirement is maybe, you know, 70 degrees Celsius water or something like that, that would be very practical. And that's I'm pretty sure there are heat pumps currently that do heat water uh for distribution of the heat, not just blowing air. — Yeah. Okay. Um so I'm thinking uh we're going to probably turn off pretty soon. Uh, so if there's any like major questions that are, you know, burning that you want to um come up with — while you're waiting for that — plan is we're not going to be doing any uh project video next Saturday. Uh we've had kind of a busy week and we're not going to be ready with the next project for next Saturday. But if we can convince Paul to come on, uh he sounds like he's interested, especially with the programming. We might get into a programming um video. And uh I'm not going to give you a lot of details because I haven't talked to him about exactly how he'd like to cover it. Uh we'll go on next Saturday uh evening just like we did here and we can go through that and we can discuss — and he uh just responded programming stream explanation point. — Okay. All right. So next Saturday expect us here at 6:00 Eastern time. And uh you've seen Paul once before, but we're going to get into this in a little bit more detail and uh let him show you what he can do and what he's done with uh both AI and just pure basic. — Yeah. And I think that'd be fun to talk both about AI computers. I might be able to um not sure if you can see it. Probably can't see it. Can you uh Yeah, right behind if you can see his head. Uh right to the left of his head. If you say that little computer there. Um, yeah, that uh that's a new computer that I built uh just for our um you know, video editing and everything. It's the one that I told you about the amount of RAM and everything. Um and it has the Intel — processor. A lot of you might Well, it is water cooled. Um anyways, so I might be able to uh bring that up here and sort of show you that. Uh Paul can talk about programming. We can go over the AI stuff, all everything that you could possibly think of that's related to computers than we regularly do about like sort of more mechanical stuff. Um, — and I'm also open to talking about uh issues that we haven't even covered. Uh, if somebody wants to bring up an issue, as long as it's not political, we don't want to get into that. That's just, you know, you can go elsewhere for that. Uh but if you want to bring up other issues like we're bringing up AI. We don't make AI systems. Uh we'd be open to talk about that too. Um so — yeah, — appreciate your uh spending some time with us on uh a Saturday night. It's a good way to spend a night. — Yeah. And uh PC uh thanks for the two. Um thanks. Look forward to anything that you do. Thank you. Appreciate that. And then Mike, uh thanks for the two again. Uh you're always up here. uh 24-hour live stream. No, no. I know I I've seen people that do that. Yeah. — And maybe when they're alone and they can take bathroom breaks, like leave, you know, — if you got 90 around the table, somebody can disappear for a while, but you know, — and they're also like maybe playing a video game or something like that where they're not having to, you know, come up with topics or read comments or something like that. I'm impressed. I actually by some of those

people how they're able to do that and not pass out. Um, — and most people complain about the lengths of our videos. Could you imagine now doing a live stream that's typically, you know, 2 hours for 24 hours? It'd be a challenge and uh you probably learn a little bit more about us, you know, when you start to — if we were going to do a 24-hour live stream, we'd have to have a couple movies prepared that we would just sort of sit back, eat popcorn, and maybe every once in a while look at the comments and say, "Oh, yeah, we agree this movie is great. " Okay, — your time is valuable. Everybody's time is valuable. And uh again, thank you very much for spending your time with us tonight. and uh it makes it worth it. It's a lot of fun to know that there are other people interested in the same sort of stuff that we are and we can kind of communicate about that. So, it's very gratifying. Thank you. — Yeah. Thanks everybody for tuning in and we'll see you next week. Ah, Heat. Heat. Hey, heat. Heat. N.

Heat. Heat.