Is Mind-Reading AI Coming Soon? My First Real AI Nervous Moment

0:00Today on the James Altucher Show. Data is oil. Data is the gold of this AI revolution. Imagine you have an AI that has all of everybody's thoughts also. So it's not just learning on tweets and texts. It's learning on the 60 ,000 or so thoughts times 8 billion people that are thought each day around the world. This sounds like amazing science fiction and magic and everything that one could ever have dreamed of. Or it could be the end of the world. This isn't your average business podcast, and he's not your average host. This is The James Altucher Show.

0:57So something just happened, which for the first time really makes me question whether AI is going to be a utopia or a dystopia or even what's called a newtopia. So a utopia is a perfect society. A dystopia is where technology is used by government or corporations to control everyone, sort of like 1984 or the Hunger Games. And a newtopia is somewhere in the middle where a government tries to use technology to make a perfect society, but really they're controlling things and are corrupt behind the scenes. And look, I'm an optimist. I tend to think that every advance in technology is worthwhile to society.

1:44There's going to be bad consequences, but the good consequences will usually outweigh the bad. So nuclear energy is a great example where, of course, the consequences are devastating on nuclear energy. We've seen this with the two droppings of nuclear bombs in Japan in World War II, but also nuclear energy can potentially make cheap energy for the entire world. and we've seen this with the internet where you could look at lots of bad consequences of the internet, but overall the internet has improved the quality of life of everyone. And look, right now we're getting information about the revolution that is happening in Iran simply because of internet connectivity and things like that are invaluable for a free and democratic world.

2:25But AI, there's been a lot of nervousness. Is it gonna change society for the better, make new cancer drugs, give us all the information we want at our fingertips? Or is there opportunities for AI to turn into kind of like this dystopian Terminator, Blade Runner kind of society? So this is the event that just occurred this morning that makes me nervous. And I'm gonna explain every aspect of it and why we really need to think about this. And again, I am not a pessimist. I have never once looked at AI in a pessimistic way, but I just wanna raise the question. So OpenAI, which is the giant AI company that powers ChatGPT, for instance, they just put$250 million, a quarter of a billion dollars into a new company called Merge Labs.

3:16And so before I describe all the details of this, there's a couple of things that already are weird. First off, OpenAI itself just raised$20 billion. So why is it interesting that OpenAI raised$20 billion and they just put$250 million into Merge Labs? Well, usually when a company raises money, it's because they don't have enough money. They need money to build technology or to market their technology or whatever. It's not usually the case that a company raises money and then immediately puts some of that money into another company. Now, Merge Labs and OpenAI have one thing in common. Sam Altman is a founder or co-founder of both companies.

3:55So when I looked at news articles on this, most of the news articles did not mention that. I am not really against a conflict of interest, particularly these are two private companies. They could do whatever they want and if their shareholders are okay with it, but it's just a little bit of a weird thing. And now the question is, OpenAI, we know what they do. They make ChatGPT, which a billion people use. Merge Labs, what do they do? They do something very interesting and it's similar in some respects to a company called Neuralink, which was started by Elon Musk. So Neuralink and Merge Labs, what they do is what's called brain-computer interfaces.

4:33So I'll describe what Neuralink does. They put a chip in your brain that uses AI to understand what, in some very limited cases, what you might be thinking. So an example is, if I'm thinking I want to move my right leg, then some neurons will signal and the neural link chip will recognize that signal. They've used AI to learn what those specific neurons signaling mean. And they'll say, oh, the chip in the brain will say, oh, James wants to move his right leg. So if James, God forbid, is paralyzed, then perhaps the neural link chip could signal to an Iron Man-like exoskeleton that is wrapped around James's leg and body, hey, move the right leg.

5:24The Neuralink chip is marketed as a way of helping people who are paralyzed, helping people who are blind. I've seen examples, there's examples online of helping people who are paralyzed use computers. So if they think of a spot where they want the cursor to be, the cursor will go there. Again, what the Neuralink chip does is it uses AI to recognize patterns. So it says, oh, when James is thinking of moving his right leg, these neurons signal. When James is thinking of moving his left leg, these neurons signal. So it uses AI to learn these patterns so that later on, after the learning system is built, it's able to help James move.

6:01So in those cases, the Neuralink chip or the brain-computer interface is reading my neurons. But sometimes the Neuralink chip or a brain-computer interface could hypothetically write to the neurons. So what would be a useful instance of this? Well, let's say I'm blind and the Neuralink chip sees that I walk into a room with a bunch of flowers and it knows which neurons will make me think of an image of flowers, it'll write that image into my brain. So now I see flowers. So potentially the Neuralink chip or a brain computer interface could help blind people see again. So these are the most talked about use cases.

6:42They're usually medical use cases to help the paralyzed, to help the blind, to help the deaf, to make people who previously have lost some abilities, some physical abilities, get those abilities back. So now, how does the brain-computer interface work? Well, I described to you how it uses AI. It studies millions of patterns of neurons in your brain and what thoughts might be attached to those patterns of neurons, and it builds a learning model similar to how ChatGPT is a learning model built on all the data in the world, like Reddit posts, books, tweets, other data. That's all the data that's fed into the large language model that builds ChatGPT.

7:23But a brain-computer interface has a different form of data. It has the neurons that are associated with different thoughts, and it learns which neurons or which patterns of neurons are linked to which thoughts. And so that's a brain-computer interface. And then once it's able to translate your thoughts into a digital form, it's able to then do things like get a computer to do something or get an exoskeleton around your body to do something. So Neuralink does this by putting a chip in your brain. The chip is able to look at the electrical signals of the neurons and then its AI gets to work. What I'm about to describe is a little bit scary, but it might not be.

8:01So what Merge Labs does is they take cells and neurons from your brain and they reprogram them. So genetic engineering is a field that's been around for 20 years. You could look up technologies like CRISPR on how to edit genes. They change the genes in your neuron to be more sensitive to ultrasound. Then how do they get these neurons back in your brain? Well, they put the new neurons into a virus. The virus puts these neurons back into your brain. And I might be getting this a little bit wrong, but at a 10 ,000 foot level, this is what's happening. A virus is injected into you that puts these genetic engineered neurons back into your brain.

8:47That aspect is correct. And then what it does is it puts a device that's very sensitive to ultrasounds on the base of your skull. What they claim is that Neuralink is invasive. They're putting a chip into your brain, whereas Merge Labs is non-invasive. It's all biological. The neurons, yes, they are genetically engineered. Yes, they are put back into your brain via a virus, but it's still normal biology. There isn't a foreign invader like a chip that's sitting in your brain. Instead, they have a small ultrasound reading device kind of attached to the base of your skull, so it's outside of your body, it's outside of your brain.

9:28And when it wants to know what you are thinking, it basically sends ultrasound signals to these genetically engineered neurons that respond accordingly. Almost like the genetically engineered neurons are like sleeper cells inside your brains that wake it up when they get the right signals from the ultrasound device that is attached to your skull. If the device is looking for, oh, what is James thinking? Well, now it is able to listen to these new neurons in my brain that are very sensitive to the ultrasound signals that the device on my skull is sending to the neurons. So this all sounds like science fiction, but it's very real.

10:09This is a real company started by Sam Altman, who also runs OpenAI. They just got$250 million from OpenAI and a few others. I don't know whether it's a good thing or a bad thing that they are genetically engineering neurons for their ulterior motives and then injecting them back into my brain. Oh, even the word ulterior is a little judgmental. Now, let's just think of some basic use cases. So obviously, for these brain-computer interfaces, what it allows people to do is think thoughts that could then interact with AI models. So we described some of the medical uses. I could think, I want to move my right leg.

10:51and even if I don't have a right leg, if I have an electronic or a computer right leg like Iron Man does or the Bionic Man does, then I can think I wanna move my right leg and then it will move the right leg. The other thing it can do is very simple thing. I can control digital devices with my brain. So for instance, if my brain computer interface is on the internet and my thermometer is on the internet, then in the middle of the night, when I wake up, If I'm too hot, instead of getting up to turn down the thermometer, I could just think I want the thermometer to go down and it will go down. I'm giving a simplistic example, but I can control any device.

11:33I could turn on the oven, turn on my car just by thinking, and I could turn on the TV. I could switch channels. Let's get a little bit more sophisticated. I could start playing games just with my thoughts. I saw a video of someone actually playing online chess just by thinking, not actually moving the cursor. Or I could look things up on Google. I could look up information. Or thinking about the AI aspect, what do you usually ask ChatGPT during the day? So let's say somebody is talking to me about, I don't know, Abraham Lincoln's, the Gettysburg Address. I could have OpenAI in my brain. I could have ChatGPT in my brain, summarize the Gettysburg Address.

12:15and I could think to myself, give me a smart response that criticizes what this guy is saying about the Kettysburg Address, and it could come up with that thought for me. Or, you know, Chachapiti knows every kind of psychology and persuasion technique. So I could say, given that my boss is saying A, B, C, and I want to raise, what should I say back? And Chachapiti could put the thoughts in my brain of what to say back, word for word. So this is an example, again, of both reading my thoughts and also writing to my thoughts because sometimes I don't wanna just control devices, I wanna also get information, and so that's the writing.

12:54So another thing is communication. So instead of me going to a computer and typing an email, I could think, oh, I want to talk to Bill. I wanna send this message to Bill. Here's the message. And the message goes through the brain-computer interface, goes on the internet, goes to Bill's brain-computer interface, and then goes to his brain. And now Bill has got a message from me and he could send a message back. Guess what? That's normally called telepathy. But as Arthur C. Clarke, the author of 2001 A Space Artist, is famous for saying, any sufficiently advanced technology is indistinguishable from magic.

13:31So what we now call telepathy, for all we know in just a couple of years, is gonna be real science and real technology that people are doing every day on a global basis. It's instant global communication from brain to brain. So another thing potentially is immersive gaming. So if it could write thoughts to the brain, now let's say I wanna be, I don't wanna be in my home, I wanna be climbing Mount Everest. Well, it could write thoughts, I could be in a game which writes thoughts to my brain computer interface, writes thoughts to my brain, and now I'm climbing Mount Everest. So that's a little more sophisticated than anybody is capable of now, but the future eventually arrives.

14:17So that's what's going to happen. And again, in terms of human-to-human communication, maybe it goes beyond thoughts. I can share my emotional state with somebody or I can share an enormous amount of information with somebody like, oh, give my therapist my entire childhood in one brain dump. So on and on. But what could be a little nerve wracking is there might be subtle ways to influence or persuade someone of things. If you could truly write thoughts and if there are neurons which are related to your political preferences, which there probably are, to be honest, we don't really know. I mean, we're going to learn a lot more about neurons as these technologies develop and which neurons are associated with which kinds of thoughts.

15:00But it's not unreasonable to think that some of your opinions are wrapped up in some of the neurons you have and that the AI will eventually be able to distinguish which thoughts are related to which opinions. And again, we can both read and write with this technology. But what does OpenAI hope to get out of putting$250 million into this? Well, this ultimately is the real future of AI, which is not just me sitting at my computer and tapping to chat GPT and getting an answer. That's old school. Eventually, people are going to laugh. They used to actually like type into computers to chat GPT. Eventually, all 8 billion people on the planet, in order to keep up with everyone else, you're going to have to have a brain computer interface.

15:48And people say, well, I won't use that. To be honest, I probably would use it because you're simply going to be a billion times smarter if you can control all the devices in your home and all the devices around you with your brain. And B, if you have access to ChatGPT and Google and other technologies right in your brain, you're going to be infinitely smart about all subjects. And you could argue, well, it doesn't really make you smarter, but it kind of does. Like when I have access to ChatGPT, even during the day, I'm able to learn things much faster. I'm able to process information much faster.

16:26I'm able to respond to information much faster because I'll still form my own opinions, but I'll be able to get a lot more information in order to form an opinion much more quickly. And I'll learn from the opinions of others because chat GPT has all this stored up. But again, so on the one hand, you can see why open AI slash chat GPT would want to invest in this because what will be the AI model that the merge labs brain computer interface talks to? Well, it'll be chat GPT and chat ChatGPT slash Merge Labs could potentially make trillions, not just billions, but trillions of dollars from this.

17:03So from that point of view, you can understand from a profit-taking standpoint, this is why OpenAI would do it. But there's a more insidious reason. Right now, ChatGPT learns by reading all the text in the world and forming a giant large language model or AI from it. So it reads every book, it reads every news story, it reads every Reddit post, every tweet, every LinkedIn article, every Facebook post. It reads every piece of data it can get. It watches all the YouTube videos, and then it builds gigantic AI models. Data is oil. Data is the gold of this AI revolution. Nothing really separates ChatGPT from Twitter's Grok, from Google's Gemini.

17:49They're all basically the same software. What really separates a large AI model is the data it has. They all have the Reddit posts. They all have all the books. They all have the tweets. But imagine you have an AI that has all of everybody's thoughts also. So it's not just learning on tweets and texts. It's learning on the 60 ,000 or so thoughts times 8 billion people that are thought each day around the world. So it's estimated that people think about 60 ,000 thoughts a day, 8 billion people, multiply that together, and that's how many thoughts there are per day, every day. And imagine an AI that is learning on everybody's thoughts every single day.

18:36Look, it's using that to get better. So it could better service people who, for instance, have medical needs or better service people who have real positive use cases for an extremely smart AI model. But again, there's the obvious questions about privacy. I won't get into that. And there's also the questions of, do you really want an AI to learn on everything you do? You might not, and maybe that will never be fully possible, and maybe there will be settings so you can turn that part off, but it's going to have to learn on some of your thoughts in order to get good enough to read your thoughts to be useful.

19:11So where that border ends and where the border begins where, hey, this is a little inappropriate, remains to be seen. But I think as much as I am optimistic, and by the way, if this brain-computer interface was available tomorrow, I probably wouldn't get it tomorrow, but maybe the day after tomorrow when I see the first results, because I want to be smarter. I want to control my thermometer from my bed and play games and have telepathy and have access to all the world's encyclopedias at a moment's notice. So I would want this, But I do question there's a small number of people and companies that are going to control all the data in the world.

19:53And once we hand over that data, we can't ever get it back. It's all going to be in these AI models that are going to know everything about human existence and how brains work and how to potentially write new thoughts and opinions to the brain and know everything everybody's thinking. So if somebody doesn't agree with their government, it's almost like the movie Minority Report. Maybe they're arrested before they even do anything just for thinking things. And look, there's been books, 1984, there is all this discussion of thought crimes. And so this has been predicted by science fiction writers for the past century.

20:28So that's it. That's the news. And what's scary is that news like this is happening more and more frequently. To conclude, I'm an optimist. I'm hopeful. I'm excited. this sounds like amazing science fiction and magic and everything that one could ever have dreamed of, or it could be the end of the world.

21:03Pluto TV has thousands of free movies and TV shows. This is the mindset. Free. This is the mantra. Free. This is the... With movies like Joe Dirt, Pixels, and 50 First Dates. This is awesome! And TV shows like Survivor, SpongeBob SquarePants, The Fairly Odd Parents, and Ghosts. Pluto TV is always free. Huzzah! Pluto TV. Stream now. Pay never. You're welcome.