AI & Robots, Robert Scoble, Wrote two books about Spatial Computing. Media pioneer. Tech blogger

Please welcome Robert Scoble. Yay.

Hi, how are you doing? Let me get my microphone over here.

Robert Scoble, of course, tech master and AI master. You're the master. So the reason I, I know of you when I first knew heard of you is because you worked with my husband at a company in San Francisco called Rack Space. And even back then you were reporting on all kinds of new technology and your reporting has always been different. It's not just this company's coming out with this and this has happened. It's always been really a deep look into the what, what are you doing, how are you doing it and why are you doing it? And I've always been fascinated by the stories you come up with. And I'm really excited to announce that you have a new show coming out too. Is it unaligned I see on your shirt? It's right there. Tell me about it.

It's an AI term. If AI goes unaligned, it's gone anti human. It started, you know, killing people or doing something naughty. That's the fear long in the future that it might do that for humans. It means that you're not attached to any organization. So it's a little fun name. But I'm interviewing a bunch of AI entrepreneurs, people who are using AI to run their businesses and build, build things.

So, okay, cool. How do people find it? Is it a YouTube show or is it a podcast or what is it.

Beyond X.com and YouTube and Facebook? It'll be everywhere, but it'll be. YouTube is where you put video. Right.

I was reading a story just this morning that on Wednesday there's going to be a very secret closed door meeting and all the big wigs are going to be there. Musk is going to be there and Zuckerberg's going to be there and the Google folks are going to be there. It's a, yeah, it's a meeting in Washington D.C. and it all has to do with AI. Yeah, yeah.

Figure out how to regulate this and how to keep some of the nasty stuff from happening. Because there is a lot of dystopian nasty stuff that could come with AI, joblessness, wealth inequality, and then this existential threat, you know, down the line where it could, it could decide to get rid of us.

I was reading a quote from you and I don't have the quote in front of me, I'm sorry, so I'm going to mess it up. But it was along the lines of, listen it, we're going to feel pain before we, you know, get to the good stuff about AI, that there's always going to be lost jobs. There's going to be the bad parts of it too. Before we get to like the, you know, the reason why we switched, the reason why technology moves us along.

The productivity gains that are coming are pretty interesting. I mean, yes, as people who drive taxis and trucks over the next 20 years, their job is in real threat. But we're going to have less tragedies on the road and a much more productive economy and be able to use cars and trucks for things that we don't really use them for right now.

What do you mean about productivity down the road? What else is coming? What. What does America look like in. In 10 years and 20 years?

There's people who are building factory systems that use AI. I mean, let's just get a little bit more real. Surgeons are using AI in the HoloLens to help them do their surgery. That makes them more productive for. Right. And there's a lot of AI that's coming after a whole raft of jobs. I have a list on x.com of 2200 AI companies and they're going after everything from psychiatry to driving, you know, and yeah, this, this is something you should pay attention to. It's.

I mean, are we going to be able to think for ourselves anymore or is all the thinking going to be done for us? Well, for a while.

I mean, the large language models that are underneath some of the systems, like ChatGPT or Google's Bard, they do hallucinate or technically confabulate errors. They make things up. And so as a human being, if you're going to use AI to write a college essay or something like that, you got to check all the facts.

Yeah.

So you still got to engage your brain. Maybe 10 years from now you won't have to engage your brain.

And that'll be now we have to. Yeah. So I recently saw this story that you, I think maybe it was a month or so ago, maybe not that long, that you had posted about going to. I don't know if it was a convention or a company and looking at robots that were kind of humanoid robots and your fascination with them are these robots that are human like and using AI and are like, does every house gonna have one? What's going on with this?

Eventually, yes. The one to watch is actually the optimist from Tesla because he has the ability to get it into your house because he has the autonomous car. And so the autonomous car. I mean, I have a Tesla in my garage. It just Drove me all around Silicon Valley. Right? That can bring a robot to your home and, and the technology and the data set that.

What do you, what do you mean your, what do you mean your Tesla can be, bring a robot to your home, like it just knocks on your door one day, what happens?

Yeah, yeah, my Tesla might go, might open the garage door, go pick one up someday and then come back and it'll deliver it to my house and then come in and go, hey, it's nice to see you. What? You know, do you want me to fold your laundry or do your dishes or, or clean your windows or clean.

We laugh, but you're not, you don't. You're serious.

No, I'm very serious. Right now it's hard to get people to see that world because these kinds of robots are, are not done yet, not ready for that kind of work. But they're getting closer and closer. Stanford has a robot that does 1600 tasks, right? So you just say a few more years. I mean these, these, this stuff is learning pretty quick. So right now a robot like that probably is going to cost about a hundred thousand dollars. Oh, but it's 26 little electric motors. How much do those really cost? Not much. I'm expecting if when they make them at scale they're going to be able to reduce the price to less than $5,000. When it does that, that's when you're going to have one in your house or a lot of people will.

What's the time span that you think that will happen, Irene?

I think that's 20, 26.

What?

Yeah, it's going to happen fairly quick because the learning that is happening in the AI world is moving at a fright frightening pace. Right. Elon just showed a system two weeks ago that learns by watching videos, millions of them, and learn to drive a car. A car, a self driving car is a robot, Right? It's just not one with fingers.

Are we, am I wrong to be scared of it? Because I look at the army of rope of Tesla robots that we just saw on the screen and my first thought is a Will Smith movie, right? The iRobot, they're all coming after me. I mean, am I wrong to look at that and think, believe me, I'm.

Scared a lot of times about AI and robots. So no, you're not wrong to be scared of this technology. It does bring some, some new problems that we have to figure out. But this is really going to assist people as well and help, help them live their lives.

Oh wow. Okay. So within what the Next two years we're going to see accessible robots that can come into our homes.

And two years, probably in, in a factory situation or in a higher net worth situation. Right. Somebody can afford a fifty thousand dollar robot. By the end of the decade they're coming down in price and there's going to be a lot of them, not just Tesla, there's other companies working on them as well.

We saw them doing tasks in the video. Do they talk to you?

Yeah, that's a large language model. You can already get ChatGPT on your iPhone and talk to an AI and it talks back and it's pretty damn good.

And so the robots will do this too?

Oh yeah.

Okay.

One thing I love doing is pulling out Chat GPT and just having it record, you know, while, while we're talking. So there's a little microphone down here. Now it's recording and if as it listens to you, it will start talking back to you. As soon as we send the recording to, to it, it's really pretty amazing at what it's doing. Hey Chad, GPT. Can you say hi and, and tell us a little bit about what you can do for human beings? Right, so you send that up and then you can watch it right, you know, right back.

Oh wow, look how fast it goes.

I'm chatgpt. I can assist with a wide range of tasks, answering questions, helping with research, tutoring on various subjects, providing writing assistance, brainstorming ideas, and much more. My primary goal is to provide information and facilitate productive conversations. How can I assist you today? Right.

Oh man.

So then you're gonna have a virtual being on your phone that you're talking to like this and it's going to talk back to you and it can talk back to you in any language. It's really crazy. And you can talk to it in any language.

So we know the robots are coming. That sounds ominous. We know we have the Chat GPT. What else is on the line coming our way that maybe we don't haven't really, you know, been delving into?

There's 2200 companies I've found There's actually more, a lot more than that. That's just what I found on x.com and these are companies that are doing, helping you with writing, doing really weird, weird video editing by videos by the way. When I import them into a program called Descript, it writes it, the AI writes a transcript and then you edit the video by copying and pasting text from the transcript. And then it also, if you have a fan noise in the in the room or something like that. It gets rid of that. The AI gets rid of that. It's so there's a lot of things that AI can do to help people do all sorts of tasks. And there's thousands of companies already that are doing all sorts of things from helping people run factory floors to run retail stores to you had to talk to IT and get some answers or some college essays written.

You had mentioned earlier surgeons doing things. And I was reading a story this morning about Meta's virtual reality technology actually being used to train surgeons. The story today saying doctors are practicing at home While wearing the Q2VR goggles that simulate medical procedures. And medical schools are also now adding VR to the curriculum. So this is technology coming into play.

And it's been here for a while. It's just, it's hard to get people to buy headsets for VR. At Stanford University they have a VR group and they took 360 cameras out on the football field and captured the football practice and then they put the quarterback into a VR headset so that he could watch his practice over and over again. And he started improving his stats because he could see things that were happening on the field in VR. He could stop or slow down the game so that he could see what was going on better than he could in real life and learn the path patterns.

Wow. Yeah. Today I was also reading an article that Tim Burton has come out and he's saying, listen, even if the AI copying my art, the results is pretty good. It's missing humanity and soul. So for all of these things, for all of these help with writing that we're going to get for all of the, you know, assistance and you know, trying to do things. Are we missing a key piece of humanity when we introduce AI?

It's funny, the founder of Amy Aimi, he's a generative AI music company. He said, you know, AI can make music now and it's okay, but it doesn't test that well with human beings interest. If they add just a little bit of human performance into that AI, it tests like a real song, people like that. So that's the trick. I, I say you got to sprinkle some human sauce onto it, right? Same, same thing. If you're going to turn in a college essay from Chat GPT, if you just turn it in the way it kicks it out, well, it probably has an error to there that was hallucinated by it. So you got to fix the error. But it also has this kind of cold pattern to it lack of soul might be a way to put it. You got to add some human sauce to it. You got to make it a little bit more interesting, a little bit weirder, a little bit, you know, a little bit more human. And you gotta add a little bit onto on top of what the AI does, and then you'll get A's on your essay.

I'm wondering what the, what the human jobs will look like in the future.

I don't know, you know, I mean, I have a list of 1500 AI artists that I'm watching, and this is, this is a crowd of people that's kicking out the most incredible art. Every few seconds, a new piece shows up. On my tweet deck. I have a screen where I can watch x.com raining down like the Matrix. And this whole crowd didn't exist six months ago. Right.

So you think people will find a way to use AI to make money.

For themselves and have fun and entertain people and educate people and educate themselves. Right. My favorite prompt of Chat GPT is please show me all points of view on X on a topic. You know, and it does. It's really good.

When you look at all these companies that you mention, what do you think best practices should be? What are some of the rules that maybe haven't been in place until now that ought to be.

You have to understand this new technology. If you're a business, you have to understand what the risks are of intellectual property being sucked into open AI. Right. If you copy a spreadsheet into chat GPT, all of a sudden OpenAI has your spreadsheet has everything in it and has the technology to understand everything in that spreadsheet so it can see patterns in your business that maybe you didn't even see? Well, is that data going into a training set for the future of OpenAI's training? Yes. If you are going in the front door, if you're coming in through the API door, if you're a professional, you can use the API there. The rules are different. They don't ingest your private data. But most of the businesses are building systems where these large language models are behind their own firewall so that they can protect their intellectual property and keep it from going into some of these AI systems because they're getting very smart.

I've read that there's a kind of an issue with AI where the two words predictable and unpredictable are used. So creative, they say, can be wonderful, but also unpredictable. When you look at AI, is it a predictable science or can you not predict what this humanoid robot in your house, folding your laundry is going to do next.

Oh, you're going to be able to predict gross things like, oh, you know, please go and do my laundry now. Right. And it will go and do your laundry now. Does it do it exactly the same way? Does it stand in exactly the same place? No, this AI, in fact, if you talk to ChatGPT, if you ask it the same question, you'll get a slightly different answer both times. It's a little unpredictable, but it is answering the question for you, Right? It's like a human being. If you ask me the same question twice, I'm going to give you a slightly different answer both times.

That's true. I mean, I guess in that way it's rather human. Like I'm looking at though, that the robots we looked at and all the advancements to kind of help make our lives easier. And even in the Bay Area, there are companies that are now doing the. The Jetsons flights. Like, you can come have a vertical liftoff at your house and they'll take you to sfo. I just. It feels like finally the future has come. All these little things that we saw in the movies of the past are now almost here. And I don't know if people are going to cope.

Well, yeah, we'll deal, we'll deal. I mean, when I was a kid and I first got my first computer, you know, an Apple II computer, back in 1977, if you told me that there would be one of those underneath every light in my house, I would be like, you're insane. But my little lights on the wall here have a computer under them that's more powerful than that. Apple too, and connected to the Internet and has microphones. I mean, these little lights that cost a couple hundred bucks, right? So my entire world is science fiction now. Everything is changing. And, you know, as a human being, you gotta start catching up because this is gonna catch a lot of people unaware. Maybe not today, but in a couple years when you get a pair of glasses out on at an Apple store or something like that, you're going to be talking to the AI and it's going to be talking to you and it's going to be doing a lot of things.

There's a lot of money in AI. We've seen some companies, you know, just, yeah, take the money hand over fist. Is the new hub of AI San Francisco? Or as I read, is Seattle an AI hub? Where is the future of AI going to be?

Definitely San Francisco is where, where the Nexus is. And the activity there are Hubs around the world, Tel Aviv, Beijing, Seattle. You know, there's a lot of AI people all over the place. But the big, big dollar work is being done here in Silicon Valley. And you see that, right? With the autonomous cars. You go to San Francisco, you see autonomous cars everywhere. Now you don't see that in the rest of America.

Yeah.

Or the rest of the world.

The story I'm reading says AI has Forbes, AI50 has Seattle startups. There are a bunch of startups in Seattle and then there's also, as you mentioned, a lot in San Francisco and Bloomberg has come out with a list of, of the 10 AI companies to watch right now. And I think OpenAI is on the list. What looks like, who looks like they're emerging to be the leaders in this space feel to you.

Open AI is certainly one that I watch very closely. I mean there's a whole lot of, a whole number of competitors. Anthropic is one that has an AI that works similarly to Chat GPT but can take much longer documents. So you can feed an entire book to it. It can read the book and then it can spit a book out back out to you. It's crazy what's going on. Right. So there's a whole variety of companies that are in that top tier that are trying to compete for the mass consumer over the next few years.

Are other company countries beating us when it comes to AI? Are things that are happening in China or other countries around the world? Are they competing with us or are we outpacing them when it comes to this new technology?

The Chinese ll, The technology underneath ChatGPT is called a large language model. Right. It's a machine with a trillion knobs, basically. It's very complex new machine. And in China they are definitely competing with us, whether they're ahead of us or behind us. You know, most people say they're a little bit behind, but they're starting to kick out large language models that are pretty interesting and they have a lot of manufacturing data that they can feed to their model and train their model on. So I have a feeling if I want to learn how to make an electric, a high speed electric train, I'm going to have to download the Chinese LLM, right, and use that to get the engineering data because they've built a thousand electronic trains there, we haven't built one. So they're going to feed that data into their large language model and train it to be smarter than ours is about that topic. Now would I use the Chinese one to learn about Tiananmen Square. No, that ain't going to be in the Chinese one, at least not for a while. So you know, there you have to download or use something from Chat GPT or, or download your own meta. There's many that are, are coming. In fact, I just saw a whole system yesterday that runs on your phone without being connected. So that's what's mind blowing about this for me. This large language model compressed all of human knowledge. In fact it read 20 years of my blogs ingested that, ran it through a bank of Nvidia cards to make this model, which is less than a terabyte in size, that can run on my phone. And now I can talk to my phone without being connected to the Internet and get answers to literally any question.

That's fascinating.

Yeah.

When you go and talk to people at these companies because I know you go out a lot.

In other words, someday a new series coming that you're actually gonna, it's actually gonna understand you right now.

It does.

The old technology doesn't understand you a lot of times, but the new one does.

No, but when you, when, when you go out to these companies and you talk to people about what they're doing there and how they're creating things, are they secretive about it or are they so excited they can't wait to tell you?

The small companies are excited because they don't, they need customers, right? And so they're, they're much more open to talking. Everybody's trying to keep secrets, right. If they, if some engineer figures out they, a way to make one of these models better than, you know, OpenAI can make it, they don't tell you about that until it's out. And then they brag about it, right. You know, say, hey, download our new model. It does something that OpenAI doesn't do.

You know, Wes from Florida writes, the problem is only the big companies with lots of server power can host your large model AI systems and therefore they will control how the AI behaves.

That, that's sort of true. But there's new AIs coming that use one tenth the number of Nvidia cards. So let's say OpenAI needed 10,000 Nvidia cards to make its model, which is billion dollars worth of models. Tesla has a couple just spent a couple billion dollars in their data center to run all their AI computers. A couple 20,000 AI computers in data centers, a huge amount of numbers. The newer AI learns faster and learns with less data. So one of the trends right now is, yeah, you're probably right about the big ass large language model that's running at Google or running at OpenAI. But there's a lot of work to do with smaller models and there's open source models like from Meta. Meta is kicking out large language models that are open source and so you can run those without being connected to Meta. Right. Which is pretty interesting.

A lot of people in the chat worried about jobs, about what people are going to do, that there are good things beneficial to humanity, rights flow. But underneath it all can't stop thinking about what happened to the Terminator movies and other sci fi films. Other people are worried about unemployment numbers and income inequality. And what happens in is, you know, does this create a bigger divide in our society?

It's, that's one of the real problems we're going to have and we're going to have to have a conversation about policies in America. My dad escaped poverty by getting, going to the army for a few years and getting a GI bill which paid for his education. He got a PhD at Rutgers because of this. We need, we need policies like that. I, I don't like guaranteed minimum income, but that should be on the table to discuss as well. But we do need, need to worry about that long term because some people are going to find it very hard to change from let's say a truck driver to some other kind of job in society. There will be plenty of jobs, there's plenty of jobs. It's just they're very highly skilled jobs which you need an education for. If you want to work in AI or building robots or stuff like that, you got to, you gotta be a much more technical person than a truck driver. And that's a real problem.

I'm running out of my Robert Scoble time, which is upsetting to me. Before I let you go, I want to ask you, and I'm seeing, it's so interesting. I'm seeing AI stories from all sectors. AI now being used in real estate and renovating houses, in trade data, all of these things. And people are wondering now that it's seeping into everything we're doing, is the potential for AI to cause evil, big or no.

Last September, I sat next to one of the top AI safety researchers for 10 hours coming home from the UK and he spent most of that 10 hours freaking me out. So it is possible. I was like, I was like, how, how would it go unaligned, right? How would it do that? Become anti human? He's like, well, it's. AI is already better than a surgeon that's seeing tumors and scans. And a surgeon is the most. Amongst the most highly trained humans we got. Right. So if the AI is better than them at seeing at their job, at this specific job, maybe it gets better at doing things than all humans. That's a scary world.

Where does that leave us?

Well, that's an interesting question, but where does that leave the AI? Does the AI decide it doesn't need us, you know, 20, 30, 40 years from now? Or five, you know, when this AGI, this artifact, this true generalized intelligence shows up and it. It's looking like it's going to show up in the next decade?

Freaks me out. Is there anything that I.

You can talk to it about your problems, though, you know, like a therapist laid off. What should I do? It'll. It'll lay out a whole plan for you to get another job.

Oh, that's so funny. Is there anything I didn't ask you that you've been looking at or that you want to mention, because I'm sure that you're aware of technology that I just don't even. I can't even wrap my head around?

The thing that's about to come in November is multimodal AI, and that means that it can look at your camera in real time and know everything that it's seeing. Berkeley's AI group showed me that an iPhone can already see hundreds of thousands of things. So if you aim it at a Coke can, it goes, oh, that's a Coke can. If you aim it at a Toyota RAV4. Oh, that's a RAV4. Right. If you aim it at a thing of milk. Oh, that's a carton of milk. Right. And that will bring us a whole raft of new assistance in. In society and a bunch of new things, new fun things to do with our iPhones.

Do you have to enable that technology or does it. Will the camera just see everything and go for it?

You have to. It's not going to happen automatically. You got to download.

Yeah, like, I don't want that. You know, sometimes I don't want people to see everything going on. Not that I'm doing anything wrong in here, but 10 years from now, is.

Your phone just listening to you? It's possible. It's really possible. These lights have microphones on them. Could it be listening? That's possible.

I was reading a story that you.

Have to agree to let it do that.

Right. I was reading a story that Google was saying that they held back on technology. They had this technology that could identify the face of everyone.

Yeah.

And they pulled back and they didn't release it and they didn't move forward with it because I don't know if it was, you know, they didn't know how to control it or people were concerned about it or what. Did you. Have you seen that? Have you?

Yeah, I interviewed the founders of the company that Facebook or Mark Zuckerberg bought to do his facial recognition. And I was like, this was 10 years ago. I was like, you know, where is this leading? It's like, you shouldn't be scared that we have the algorithm. The algorithm is not that big a deal. You should be scared that we have 2 billion people's photos. In other words, we can trade it to see everybody's face with this data set. Right. And you're seeing that. Right. I. I was filming one day and somebody walked in my video who was a friend, and it said, oh, you know, Jeff is walking through your video.

Crazy.

Yeah, it can do a lot of things. They're being very muted because they don't want to freak people out. If they freak too many people out, we'll call our congressmen or congresswomen and get some regulation. Right. And that's also happening. One thing that your, your audience might find fascinating, the. This technology is an extraordinary listening tool. As I demonstrated it, there's psychiatrists who are using chat GPT to listen to their therapy sessions with their patients approval. And I did this with one of the psychiatrists and I talked to her for 30 minutes, had a conversation sort of like this, where we were just talking about, you know, things in my life and how things are going. And it wrote seven highly detailed scientific notes on the conversation and figured out all my mental illnesses out of 30 minute conversation.

Wow.

Right. So, yeah, it's scary, but it also helps the psychiatrist provide a much better quality of care because she has the ability to track you over meetings. And the AI is helping her.

Wow. She.

She told me on Friday she thinks it's going to be malpractice to not use AI soon because it's so good.

Interesting. But I wonder if it will tell if it will kind of develop mental illnesses for people that don't even necessarily have them. Like, what kind of mental illnesses does say you have? And are you sure? And what if you weren't even going to a psychologist and all of a sudden it, it hears you and it says, you know, I think you have xyz. These are things you didn't even.

Or dementia or autism or. Right.

So interesting. Oh, man. All right, Robert.

The new world. We're going into. I'm trying to track it as fast as possible because we're moving fast. There's of lot a. A lot coming.

So tell us one more time, how can we find your new show, Unaligned?

I. I would follow me on x.com because that's where I do all my work. I have lists of all the AI companies, all the AI artists, all the investors, a bunch of other things there. And I. I'm watching 70,000 people who do this kind of AI and I'm constantly re resharing their posts.

So thanks, Unaligned. There you are. And what's your Twitter handle? Is it at Scobalizer?

Yeah, it is. Last name and then I Z E R at Scobalizer.

The new show is called Unaligned all things AI. It's something we all have to keep up with and I can't think of anyone better to help us do that than Robert Scoble. Third time's the charm, my friend. Thank you for being here. I'm going to talk to you again next year.

I love you. Thanks.

I love you too. Thank you for coming on.

Hi, it's Mark. And I thought that was great. Hit the notification bell. You'll know whenever there's a new video being dropped. And please subscribe to our channel to help us save the universe.