What Is Going On With AI? - Interview w/ Robert Scoble

I have to ask you first. This is a legal thing. Are you AI?

I am a biological neural network.

Good answer, good answer.

I've been joking with people. I'm a biological neural network trained on expensive data by Microsoft Corporation. And ChatGPT or GPT4 will actually parse that sentence and explain what I mean by it. It's like, what? Wow.

It's like our own little attorney now. I love it, man. So, all right, I got a question for you. When's the singularity event? When is it?

Where is it?

How do I get tickets?

Oh, man. Do you have Parkinson's? You might not get it earlier. Oh, no, I think it won't happen all on one day, right? It'll, it'll arrive for Neural Link, right. When you get jacked in and get wires on your brain, that's, that's pretty much getting there. Yeah. Right. I still think that's decade, decades, two, one, two decades away. Unless you have Parkinson's there. You already get a wire on your brain if you have Parkinson's. Right? And I met the surgeons who are doing that and they said, well, now we have a conversation with people and we're, we're like, well, you need one to fix your Parkinson. Do you want more? Do you want more wires on your brain? And it's like, yeah, jack me up. Put, put 8 in there or put 10 in there. Neural Link is doing something like 32,000 on, on the pig's brains, right? Little electrodes. They've been doing all sorts of animal research to figure out how the brains work, how to hook up to them, how to talk to them, how to write software that talks to brains, to biological brains, stuff like that. Let's say it's a decade, right? 10 years. Yeah. We have to go through glasses first and we're not going to get glasses for three, four years. Normal everyday people. Five, six, seven years, somewhere around there, once we get everybody on the glasses. So that's a decade. Then we can start thinking about neuralink because we'll have the data for, on everybody. Because these glasses are going to be watching your eyes, going to be looking in your ears, right? Going to be watching your hands. So all of a sudden it has been watching everything you do. So it'll have the data to really integrate into your life when you get the wires on the brain. Interesting. Maybe it's 15, 20 years. Yeah.

For more or next year, I mean, right. Things are speeding up.

So the problem. No, it's not gonna be that fast. Because cutting open your brain and putting Wires on your brain first of all requires FDA approval and they turned it down so far. But. But it also is $150,000 surgery and it has a lot of side effects. Right. If you have Parkinson's, you'll sign up for that. Yeah. Right. But it, if you're a normal neurotypical person, you probably won't sign up for that anytime soon because unless you're a billionaire. And even then, if you're a billionaire, why would you sign up for it first? You know, it's like, let some other people test this thing. Yeah, maybe Elon would do it, but I doubt even Elon would do it for first. Right. If you get a pair of augmented reality glasses on your face, you're probably going to think it's the singularity, I bet. Particularly if it's, I mean, you're seeing how good chat GPT is at having a conversation with you. Right. So if, if there's an AI and it's just always listening and having conversation with you and putting information on your screen or hey, you know, I don't have a good example. If I was holding a Coke can, hey, how much is 20 of these on Amazon? And I told you, right?

Yeah.

You would be like, whoa, this is a new, a new world we're in. Right. Yeah. Apple's going to come with a bigger product first this year, then a more consumer version of that product next Christmas, then the glasses. Okay, Right. So three years from now, me and you probably get a pair of glasses.

Now when I think of glasses, I think of the Google glasses that were trying to come out. I don't even know how long ago these are.

Yeah, Google Glass was an early version, but it was only a little tiny screen. It didn't cover your visual field. Okay. These are actually going to have a visual field that covers your visual field. So it can put spongebob on the table. Right. Or it can extend my monitor and have two more virtualized monitors off to the side. Right. Stuff like that. And once that arrives, then you can bring virtual beings. So having a Spongebob that's talking to you. Right. Because Spongebob is being driven by chat, by OpenAI. Right.

So you have been following just about everybody in the space. When did this start in your mind, this new.

Siri? When Siri was announced, it was, they told me it was one of the first AI apps to go. It was certainly the first AI app on the iPhone. Some people say the Roku or the, some of the robots that did vacuuming had AI and stuff like That a little bit before, but right around that era was the beginning. And then I started tracking it and it started showing up in my oven. My June oven has an Nvidia card with a computer vision in it, and it showed up in my car. My Tesla has, you know, computer vision in it. And that's when I started following it last year. It started getting serious. Right. Big things started shifting. And then that's when I realized, okay, I need to pay a lot more attention to this community. I. I care about new things. Right. All the way back to childhood. I like new things. I like talking about new things. I like playing. I mean, if you told me here's a new restaurant, I'm in. Right, Right. I like new things. And this community is kicking out more new things than any any community before. And it's growing at a pretty good rate. I mean, I found 40 something thousand people in this community to follow, so. Right. And these are all people. These are a lot of people with PhDs and. Yeah. Going to Stanford or Carnegie Mellon or some school like that. Right. And coming out with some pretty serious skills and then, you know, working on a team. I mean, a lot of these kids are writing papers around about neural radiance fields now at Stanford and other places. And so they break that with three or four other people and they leave and then they start a company. Right? Yeah.

Let's say a friend comes to your door today and says, robert, I'm in, man, I'm in a. Where do I start? What do I do? What would you say to your friend?

Go to open AI and start figuring out how to get GPT4 running. Right. I mean, that's the first one. Then figure out, you know, once you start getting used to talking to that, there's a lot of other tools that do very specific things like edit a video or listen to a video conference like super normal listens to a video conference like this and writes a transcript. Right. As we're talking. And it does pretty good at that. Then at the end of the call, it looks through the transcript for patterns, like did we talk about tasks or themes? Right. And it makes some notes about the, about the transcript and tells you the highlights of the transcript. And it does that in 300 milliseconds. And it does that on a Macintosh. Right. On the neural Network, on a M1 or M2 processor.

Mindset. Let's talk mindset on this. Because as you want to embrace these tools, some people are like anti AI. They're not even looking into it. They're scared of it from sci fi movies in the past. What is the mindset shift that needs to happen for people to start embracing AI?

Get over your fear, because it's coming anyways. It's gonna take your. If you're scared you're losing your job, you're really gonna lose your job if you don't use it. Right? Because the first people who are into AI and are really using it, they're becoming more, Much more productive. So you're going to be the last one to get fired if you're really productive. Right? Your mom's going to go, man, that guy is kicking out some work or that girl. Right? You know, and. And I sat down with an optometrist on Friday and we just did some simple searches like, can you write me a blog post about Invisalign braces? And it did, man. And it's like, whoa. All right, so what does this all mean? Well, does that mean that her web team gets fired? Right? Maybe. But maybe she'll go in and say, I don't. I still don't want to deal with this. That's why I hired you all. So you got to get your productivity way the hell up. Right? You got to use the tool. And I just, you know, I can show you you can create website insights with us. You better start talking on this every day. Or. Or I am gonna have to lay you off because you're not staying up to date. Right? Sort of like somebody who says, I'm not going to use a cell phone. You don't use a cell phone world today. You're gonna get booted out of the corporation. Right?

Yeah, Right.

It'll be weird, right?

Especially if you're in sales and you don't want to make any sales calls. It's like, I don't use the phone.

Yeah, yeah. You know, I can never get a sales guy. Let's. Let's talk about this.

Yeah, yeah, of course there's a wave coming.

Right?

And, and my question to you as you've been diving deep into this, is this wave going to be a wave of new AI native apps, or is it going to be existing apps integrating AI? Which wave is coming? Or is it a combination? Both.

Okay. I mean, last week, Microsoft and Google announced that they're integrating these large language models like GPT into their services that, you know, Microsoft Office. So soon you're just going to start talking to your computer and go, hey, computer, I need a slide deck made this morning. My boss just assigned me a presentation on Wednesday and I need to get right. And that's sort of how you should be talking to it too, because it uses some of that data to figure out like, oh, okay, you have a presentation on Wednesday. What's the topic? You know, it starts having conversations with you and then it goes and builds a slide deck. And then you could start talking to it and saying, hey, I don't know, I don't like the background. Can you find me a, can you, can you go over to stable diffusion and generate me some AI art to put on the background? Maybe like a pink forest or something like that. Right. And it will, right? Boom. Yeah. All of a sudden you have a pink forest on the background of your slide deck. Right. And somebody who works like this is going to be much more productive, get a lot more done every day, and therefore they're going to be harder to fire than somebody who doesn't even know that this. Right.

Yeah.

If your boss comes in and starts talking about AI, you don't know about it. Well, you're, you just mark yourself as, as somebody who's not curious, not up to date, not using the latest tools, not pushing yourself. Right. Not, not trying to figure out what, what is coming. Right. If you do are that, that person, you're probably not going to get fired that day. You might get fired tomorrow. I mean, seriously, programmers are telling me over and over and over, it used to take me a month to write code, this kind of code. Now it writes it in less than an hour with me. Right. That's a huge, huge productivity shift if you're doing anything with text, even answering customers. Right. Hey, can you write me an email to an angry customer that, you know, and just tell it some details about yourself, you know, oh, I run a glass shop and I, you know, we had a problem with the customer. It'll write you a damn email. Email and do a pretty confident job. Maybe you have to edit a sentence or two in a, you know, an email that's, you know, 20 sentences, but you're still 19 sentences ahead right now.

What's gonna happen to school right now? Like what it needs to happen for the education system to embrace this new era.

Complete overhaul there. I mean, some teachers are, are taking to it and, you know, working with their kids with it or their students with it, if they're older. At Berkeley University, I judged a whole bunch of student projects in design class and many of them were using, you know, stable diffusion to make their art. And after talking to them, oh yeah, we used GPT to write the text and, you know, help us Out. Those are, those are the smart ones, right? And that's. The smart approach is to pour on the gas. Use this tool, use it to learn. But you still got to learn to write an essay, because that's about making your brain think in a certain way, right? Or if you're a computer scientist, sorry, you still got to know how to do algebra or calculus, you know? Yeah. You can ask a computer to show you how to do it and help you and explain it. It's really crazy, right? It'll write code for you, and then you can ask it to explain the code line by line. And it does. It's like this line of code goes to this API and hooks this up and puts a variable into this memory location. It's like, what, Right? So it helps you really, really, really learn fast, right? And so it should be a tool that you use there. But it does have bias. It does have, it does generate mistakes once in a while. Here's one. If you ask it, what's the history of Silicon Valley? How did Silicon Valley start? Questions like that, it gives you the standard answer, and the standard answer is wrong. Even, even if you go to Hewlett Packard's garage, it says it's the birthplace of Silicon Valley. There's a plaque out there, a historical landmark. That's wrong. It started in the transcontinental railroad in the Sierras, which gave Leland Stanford the wealth to buy Stanford University, which started Silicon Valley. The problem is Stanford doesn't like to tell that story because they. 1300 Chinese or more, probably three times that amount, died in the tunnels on the, in the Sierras building the railroad because they were the first ones to use nitroglycerin, which is a highly unstable. Just explodes. Right. By the way, Chat GPT will tell you all about nitroglycerin, right? What's really interesting is it knows the Chinese's role in the building, in the transcontinental railroad. And if you ask enough questions, it'll hook it all up and say, oh, yeah, this is the true history of Silicon Valley. It, in its ingestion of all of human writing and history, right? It, it figured out the truth, but it doesn't present that to you at first. So you have to know, you have to know the history to be able to pull it out. Right?

How does that work? Is that political? Like, why wouldn't it just tell you straight up?

Because the rich people wrote the history. Rich white people wrote the history, Right? They don't like to tell the history that we stepped on a few People to get here. Yeah, right. And, and it's a dirty history. And Stanford University VR team doesn't like to talk about that history. It's not, it's, it's not how you start the parent tour at Stanford University of explaining. Oh, you know, this university was started by killing a few, few thousand Chinese building a railroad. No, that's not how they started. Oh, this, this university started. He Packard and a couple guys got rich because they came here. Yeah, right. That's how they started. So, so it picks up those, those signals from people, right? It picks up those biases. Yeah, right. And this is the problem. If you, if you don't know that you, you know, it's hard to pull it, pull out the true complete story. Then in other places I actually bullshit. So it just makes, you know, it makes up a fact, right? Because it's, it doesn't know anything. It just knows what's the next likely word, right? It's just. And it's pretty good at that. But once in a while it goes on a bad path and sort of wrecks the, wrecks the thing, right? And gives you a bad answer. And it does it very confidently because it's like, ah, this is the next word, right? You know, I, I'm just picking the next word. I don't really know anything. Right? And once in a while it gives you some. It's really fun if you, if you know there's. You can ask it to remove the.

Okay, right.

Here's one thing I did. I give it a thousand words that I wrote, right? And then I told it to edit it and it changed the meaning of one of the paragraphs, said Gavin A. And I said, can you please remove the. You just put in. And it started changing the correct, the correct paragraph. So it knew that something was different about that paragraph and the way it picked the tokens than the other paragraphs. So it can start editing itself, you know, but it didn't do a perfect job. So you, this is where the human still adds something to the equation, right? We, you guys still understand that you might know something that it doesn't and be able to fix it or improve it. The guy ran IBM Watson, which is IBM's AI early day stuff. He told me what, what we showed you on TV is IBM Watson beats the human being at the game of Jeopardy. Right? Yep. That's, that's what, yeah, right. It was one of the earlier AIs that started beating humans at things. Okay, you said what we didn't show you on TV is IBM Watson. And a human being beats IBM Watson alone because we bring some weird shit to the game. Right. We're biologic. That's why I call myself a biological neural network. Because we're different than the computer. The computer is really light switches if we. The bottom of it. We're very different than that architecturally. So we bring some special sauce to the game. And I don't know where all that special sauce is, but once in a while it shows up human fixes a mistake or whatever. By the way, that makes the AI better over time. So 10 years from now, I don't expect it to make any mistakes at all. Yeah, right. And in fact, it is getting better. GPT4 is better than 35. Right. And gives you less.

So you just see you working away like Judge B.D.

Cut the.

Sorry.

Yeah. You know, by the way, it's really good at generating. So if you want some. To tell it to create some. And it will create me in a nighttime story for my kids and make it a complete. And you know, it. It does.

Just like humans. We too. I mean, it's trying to be like us.

Absolutely. We're good at it.

Exactly.

Particularly if you ask somebody something they don't really know very well. They. They don't like to admit they don't know very well. Yeah. Start. That's sort of how AI bullshits too. If you. If you get to the edges of what it. Of the knowledge that it, you know, ingested from the web and from books and stuff like that, you know, it'll start hallucinating here and there. Right. But it does it so confidently in between facts. Right. So if you don't know your topic, it looks right, right, right, right. And then there's some. And then. Right, right, right. And, oh, you know, if you don't know your stuff, you're going to turn in an essay with some teacher's gonna go, oh, you didn't find the. You get a B. You know, are you gonna see?

Yeah, yeah.

You're not good at fighting. That's first rule. If you're going to use the AI, you got to find the bullshit.

That's the first rule. I like it. That's good stuff, man. So let me ask you about language because I heard that Dall E2 is trained with English captions. So a lot of countries like India are saying, learn English right now. Because a lot of these programs were trained in English. Is that going to matter much longer?

No.

That's what I thought. That's easy questions.

It's pretty good Actually, if you give it an essay and say, convert this to Japanese, it does. And it does a pretty good job of it. Maybe not completely perfect, but it, it's pretty good. Right. So I've been around the world. I've been to, I don't know, 60, 70 countries, right. And I've never needed to learn very much language because I've always been hanging out with the nerds, and the nerds always speak English because all computer manuals and all, you know, computer programming code. Right. Usually you're writing in English because that's how the system was set up. Because, you know, it was started here in Silicon Valley. Right. So, you know, we all, we all wrote English and then as more and more people came on the world, they had to hook into our systems. And so same thing with chemistry. Chemistry was largely started in Germany, Right. So a lot of the old chemistry texts were written in German. So if you're a chemist, you had to learn some German to go through the old books and learn your. Learn what's going on. Now, that's not so true anymore. But because the translations are pretty good, I mean, I went to the Shanghai Disneyland in China, and I was talking to people with Google Translate. Right. You know, and maybe it was simple stuff like where's the bathroom? Or, you know, where's this ride at? Or. Right. Or how do I find a taxi? That kind of thing. Yeah. But it did perfectly at that kind of thing. And it does more than that. Right. And certainly now it is really amazing. I mean, language.

Would you consider this the greatest opportunity in our lifetimes to evolve faster than we've ever evolved or use a skill set and scale it up faster? This is.

Yes. Yeah. This is an accelerant for human beings that's going to lead us into a new. I, I call it a new renaissance. Somebody else actually called it that for me and said, we're in a new renaissance because we're getting all this new art and new photography and new writing and new and new knowledge. Right. And we have a thing to talk to about anything, and it gives pretty deep answers most of the time. So it's, it's an accelerant for human beings for sure. I mean, if we didn't have to worry about feeding our families and earning a living, we wouldn't, you know, we would find these tools very, very fascinating. Right? Yeah. Because they help you build things, they help you do things, they help you think about things, they help you write books, they help you, you know, do what you want. Right.

It's Interesting. You talk about making a living too. Do you feel like the basic income needs to be developed some kind of safety net here? If this happens faster than we think and there's a huge unemployment, I mean.

We should definitely be thinking about that because there will be people who fall. I mean, there already are people. There's tens of thousands of homeless in San Francisco alone. So there, there's people falling off the, off the table all the time. And we have to have a better answer to that. The other way to look at it is this is a very, very powerful education technology. It's a very powerful interaction technology. So the system could figure out like a mental illness just by talking to you pretty quick. Right. If it kept track of you, you know, like. Right. If you have dementia and your dementia is getting worse and worse, it can track.

All right.

How many times are you having trouble remembering your kids names? Right.

Yeah.

And it can keep track of your dementia and help you skip. Frizzo down at USC is doing research with dementia sufferers with virtual reality. And it really does help them because when you have dementia, you get frustrated, you can't remember your own. You know, you're talking to your own kid and you can't remember them. Right. That frustrates the hell out of you, makes you depressed, makes you irritable. Right? Now you're miserable human being that you can't remember names. Right? Right. And pissing everybody off around you. Yeah. Well, you take them into VR and remind them of who you are and remind them of your child. Right? And all of a sudden your brain gets refreshed and some neurons are still firing. And that helps you be a happier person and help you remember things. Right. If you have an assistant standing next to you telling you, hey, you know, you're talking to your boy, this is your kid, his name is Patrick. Right. And you know, and here's what he was talking to you about for the last two hours, right? Reminds you who you are, what you're doing, what you're you're supposed to be doing that's coming. I mean, augmented reality glasses are going to help people who are blind to see. It's going to show them everything in their homes. Hey, hey, Siri, where's the pepper? In my house. Oh, it's over here. We'll take you there. Right? Because it knows soon with computer vision and the camera on your head, it's going to know every object in your home. Wow. Right?

Yeah.

So now it can tell you you'll need to unlock your iPhone first. See, you'll need to unlock your iPhone first. It'll also help deaf to hear because if you can't hear and somebody's talking to you, it can, it can do a transcript right in front of you and tell you, oh, that person, here's what that person is telling you right back to the blind person. If somebody hands a blind person a two dollar bill instead of a twenty, it can, the computer vision can say, that person just handed you a two dollar bill and you can go, hey, you trying to rip me off? Right, right. And, and, and so it can help you see and it can help you hear. And these are going to be very, very powerful technologies for those. And then it can help people like with dementia and, and a variety of other things. Pfizer's head of R D me one time that they're doing all sorts of research on a bunch of different problems, human brain problems, you know, dementia and depression and stuff like that. Ptsd, stuff like that. With, with virtual or augmented reality, you can really, really, maybe not fix those things, but certainly improve the human brain condition for those things. Right? I mean, my washing machine stopped working one day and was flashing a code at me. And what do you do? You go to YouTube and start searching on this brand with that code and all of a sudden a video pops up showing you, oh, you probably have something stuck in the pump is what that means, right? Here's how to take apart your washing machine and fix this. This, right? And I did it. It was all right, but it was. I had to watch a 2D video and do it on the 3D machine. Right. With augmented reality, it can convert the video to take this screw out, take this screw out, take this screw out, take the back off. It should look like this, right? And it's on top of the machine. So now I can teach anybody how to take apart a machine step by step, right? With augmented reality glasses, that improves a lot of people's lives. My friend owns a scrap yard, you know, hey, here's how to take away that, apart that washing machine and save the engine out of it, you know, and use it for something else, you know.

Yes, yes. Oh, man. Robert School, I just want to thank you, man. I. This is my first interview with you and I would love to check out, check up with you from time to time as I do videos of different stories and get, get a little piece of what you think. And I hope I have better questions as I evolve in the space too. I was trying to bring some good stuff. Hopefully it didn't seem like I'm too much of a new.

What kind of questions you ask. You know, you ask an AI influence. Yesterday, I had it right. Pretend you're an AI influencer. What would you do? And it's pretty. It's pretty good, you know? So you start playing with it. You start getting some ideas. Maybe it's not every question, but it gets you started, right. And gets you over your writer's block or your fears or.

Yeah.

Whatnot. Yeah.

Such great advice. Such great use cases, too. I had two grandparents that suffered with Alzheimer's, so they both forgot who I was before they passed. So that hit deep for me. Robert, I want to thank you so much for chatting with me.