path: root/transcripts/humanityplus-at-caltech/alex-peake-intelligence-amplification-ai

What will happen when we make machines that make us to make them and they make us into them? There are three different Moore's laws. There's 3 uncanny valleys, 1 coming of age story for humanity and technology, and 2 species, and 3 high stake questions.

First of all, an auto-catalytic reaction, the best example is life. An autocatalytic reaction is a reaction where the products of the reaction are the catalysts, so as the reaction occurs, the reaction begins to go faster. We can be right in the middle of an autocatytic reaction. There are two specific As and Bs that I think are autocatalytic. There's strong AI- which we don't have yet- and there are people who are aiming at it. Well, we're going to have machines that will learn things. We're going to have machines that will learn more than just a few things. There's one way o predicting this if this might be possible. One of the things that we don't know with this graph is when the software will exist to exploit this capability. A lot of people are projecting that the challenge is not at the hardware, but the software challenges. The reason that the Humanity+ and Singularity project is the difficulty solving this-  though there's tons of money being thrown at certain forms of AI, like military AI, and other narrow AIs, they are not the AIs that we would like or that would be like us.

Everyone has this slide in every AI talk- the terminator possibility or unfriendly AI where it is a military AI and we don't look foreword to it. There are two specifes that don't get along. Either way, we kind of see AI as the next logical step, and it's the showdown between us and our creation. Merging or getting out of the way is ienvitable. There's a friendly ... shit.

Strong intelligence amplification. We are all intelligence amplification users- the internet, information tech, everything we use to learn and grow is intelligence amplification. There's a difference between Google Answers, and a complete redefinition of the way brains are shaped and grow. The most profound vision and most momentum on this idea is the Diamond Age by Neal Stephenson. He composits that the makerbot- the rise of molecular manufacturing- earl gray hot, which is right hand in our lobby in a limited form, we're heading towards that goal. Nanotech and making things for free, might lead us to Earl Gray Hot star trek utopia. They don't have to compete for stuff. How much does the ship cost? Well, we no longer use money, instead we work to better ourselves and the rest of humanity. Before the girl can ask about that, the Borg attack. We now know that this might not be a Star Trek thing. We can have makerbots and not have utopia. The Diamond Age shows why. People might not make Earl Gray, they might.. into.. complete spoiled brat mode. The idea in the Diamond Age is that people split: the feets, the rowers and your labors; they are considered 3/4ths of a person in Greece. They are the logical conclusion of relativism and total freedom, they lead towards liestyles where we destroy ourselves. The ultimate tribe in Diamond Age is neovictorians, and the thees call them biggies. The whole idea of tribes in a post-scarcity economic world, it's not the economics, it's the intelligence, who you know. The essence of triablism is building a culture a shared striving for excellence, and an infrastructure for education, that other tribes seek out to join your tribe. What happens is that their schools become their competitive advantage. How do you teach 8th graders nanotech? You have to radically aggressively teach the tech, the cohesion, the values that would make a society strong.

The problem is that this is.. when you go for a perfect system of education, you might get it. You shouldn't reinvent a wheel. You have to make an education path.. nobody cane scape from.. might be conformity that couldn't invent the stuff that makes nanotech age possible. It might get stuck there. One of his characters sees his grand-daughter, so he makes something that makes humam inds and the .. make physical matter.. he makes something call the Young Lady's illustrated primer. The illustrated primer exists to solve the problem. How do you on a mass scale shape each individual person not to be the same, but to be free and easy? Make the person independent and free is a bigger challenge. The approach was a fairy tale. They open the book, it unfolds for them, an opportunity to decidew ho they are going to be, it's personalized. The primer leads through the bigger question, once you have the mind wide open, what should you make in the mind? What should be the content? The over-arching value that Stephenson emphasises is the .. conformity.. thiefs non-conformity.. You have to teach them something else. Subtlety. That's elusive to teach, it's the biggest challenge in humanity. When we have really dangerous choices to make. If we're going to do all of these problems, and all of this tech, at 100k degrees, and 10 kmph. We can't just be bold and go boldly into strong AI. We have to go subtle. There's a tech story that shows the coming of age, and it's 3dFX.

I was personally involved in 3dFX, and it tells how it effects our children as well. This was me at 13. My dad taught me about inventing stuff, and my mom taught me about art. They knew that they could only teach me half of what I needed. She introduced me to the amc. They had a floating point unit, that would be good for science. I started sim city playing, I started playing sim life, and how genes and alleles could be shaped, and how you can .. I wanted to make things, like art, they were static images, I wanted to make games, I was into fantasy games, and that made me wish the world was magic, HOgwarts and magic spells. You can look them up in old books, and they didn't work. In my cubbard in the stairs, I was trying to discover science and programming. There were real magic, and there were magic words, you had to open your mind to the computer and change you in its image, then once you had it in your mind, you could change the computer, and then you can change it in your way. We had this tech, this intelligence amplification, and it's done great. But it's still not what.. so we learn this stuff, and pretty soon you can make it do what you want. It's addictive, so you want to keep learning.

I got a powerbook and took it to school, and they didn't know what to do with me. We had teachers scrambling to do stuff with me. So they sent me to meet Dr. Dude, Professor of Computer Graphics. Dr. Dude, that was his hair at the time. The longest when.. Dr. Dude was known as David Frackyou, Dr. Dude was his real name. What he taught me, he looked at my mac, but if you're in university science, you have to know unix. Hi Alex Peake, my name is .. you can have any name, as long as it is 8 characters. alxpk. You have a machine name when you enter this world. There was this movie or it just came out, Jurrasic Park, and it blew away with the graphics. It's a unix system! I know this. It's powers the whole park. It tells you the whole thing. I have to find the file. This was the same computer in the graphics lab. I was using unix in the university. In the movie, you can actually spot the logo. Silicon Graphics was the dinosaur t-rex of computers. Something was happening just underneath its feet. If you have a computer bigger than a person. They had workstations. PowerBook. They were about to suffer the same fate as the dinosaurs. They had really big weird shadows. They bragged that they had a z-buffer in some of the machines.

Thsi was not a platform that did real-time graphics. They didn't care about this. They wanted to make movies, that's where the money was. What happened was that, they got this movie, movies are just going to get better and better. After a while, we reached a wall. The Uncanny Valley, the characters started to look creepy, and we started to want the old way. Then we reached the low. Indiana Jones and the Crystal Ball. the jumping of the shark in movie graphics- we didn't want the graphics- we wanted the story. Graphics needed to go into this rodent sized computer that was nothing compared to the SGI. It has this little killer app called Doom. Doom had Doomed the previous era of big tech graphics. They laughed at Doom, that's not real graphics, that's 2.5D, and there's stripes. It was a lot cooler than anything on the SGI. It was real-time and fun. Quake. Quake almost captures the feeling.. itw as more like Asteroid. Quake was big enough that it delivered a market big enough, and when the hardware, it turned this into this. And when you get 3D accelerated graphics, there were moer than doubling the power.

OpenGL was made by SGI. It enabled not just prettier way to kill people, but also beautiful characters, and brought girls to the art, who were excited to identify with characters (Tomb Raider). The idea of characters and games, grew graphics. Not just .. 256 MB of memory, then 256 cores/processors. What happened is that Moore's law became obsolete. Nvidia started to call it Moore's law cubed. Moore's law was bottoming out because of processors. GPUs were using parallels. Pentium 2 wasn't really just two pentium.. they pretended to give you two, they gave you addresses. 3dfx went to the voodoo2 with 3 processors on each card, and they could double that to 6. The graphics were photo-realistic. Graphics (Crysis) were basically perfect. The graphics cards are growing. Parallelism. So right now, there's characters like, total photo-realistic. We're at the uncanney valley that the movies were at. some people think that this has been crossed. The games look physically like real characters. These are like models with no brains. There's a guy with .. Unity Developer, it's going to be a huge game engine. Jessie Shell. He points out a Moore's law for Ai. Step one is the exploitable property- Ai makes intelligence amplification games possible, it's the biggest opportunity, if you get a huge number of developers sharing, you get a lot of- single to parallel, completely laughing the CPU. This is useful for memtic mentor ai.. insatiable demand. That demand is intelligence amplification. The market size of eduxation is 10x, and will be .. in 5 years. The prophecy is that someone has to go make the hit. Quake to graphics. I can demonstrate it, I am going to demonstrate it at the end, and go to my point. My point is that our game is going to use games as a cschool and explore this. Andreas Schmeil. Having a boring life, to a challenge that makes life ..

Sexual reproduction paradigm. Cloning doesn't work because you need the red queen hypothesis- bacteria constantly changing to survive. Red queen hypothesis shows that you need new DNA to.. we have a lot of people trying to control your mind. Memes, sexual education, and symbiotic reproduction is about to happen with ai, in the same way that sex happened with two people. We're going to have a say about kids.. and they are going to make better ai because of these games.. they can destroy standardized education, and that's in order, we're going to be scared because we won't know if you would trust your kids about ai. Why should we trust you? No child abuse will happen on an ai's watch. How much will we let them protect your kids? Your ai knows better than you about the child. The ai might have to protect your kid. TV. In some ways destroyed a lot of kid's lives, and we do not want to end up being zombies. The challenge is.. how do we get games to take us out of our .. connecting people to things beyond games. Will they love computers or us more? Imagine the kids grow up and it's their mentors.