Alex Peake here. "If we really want to make strides in AI, we need to have some software that learns and tries new things - and put it into an arms race [wikipedia.org] with others of its own kind."
This is half of what we are doing with Primer. But before AI can just go off and learn everything on its own, it is most useful for its ability to help us learn everything. Teaching humans is the market AI needs to grow in a useful direction, and it is the feedback loop between the players and the mentor AI that can allow players to become game developers who can consistently improve their own learning experience by contributing new AI to the mix.
Alex Peake here. You're right to point out that education is politicized and good methods with the same bad curriculum would only further misinform kids.
Games are one of the best ways to reach kids directly without being filtered by their political gatekeepers, and games can be so effective that schools will incorporate them into their instructional process even if the games are good enough for kids to play on their own too.
Minecraft is just one example of a game that is being played and used by teachers at the same time.
On human mentors vs AI: One thing I didn't have time to go into in the talk the article is based on is the role of human mentors in symbiosis with the player-game learning process. The game's purpose is to motivate the player and inspire them to seek out human mentorship to take things beyond what is possible in the game environment. Game mentors aren't a replacement for human mentors, they're a substitute for the 90% of the time kids spend playing games alone or online when no human mentor is available to them so that the quality time they do get with a mentor is well-spent pursuing all the ideas and questions they've got because of what the games were able to inspire them to explore.
Alex Peake here. ARMA2 was very much good enough that it enraged some players. A friend of mine called to complain that he had been pinned down for an hour by a sniper before he realized the sniper was standing behind him on a ridge not in front of him. It was the AI equivalent of the scene in Full Metal Jacket where they have to flush out the sniper.
Game bad guy AI is about as useless for learning purposes as a military Skynet Terminator AI is.
The kind of AI that we need to be a good teacher is one that is good at modeling a learner's mental model based on their interactions and to communicate ideas and present challenges that push the player to continuously up their game and master something. We're working on such AI in Code Hero.
"While you're all training your kids to be machines I think I'll give mine a toolbox and some engineering and science themed kits. We'll see who changes the world."
Exactly. Most of the best learning experience comes from people having tools to experiment with and a combination of great materials like tutorials and good human mentorship.
The chief design of Code Hero and all Primer games is to draw the player through the game into meeting other human players face-to-face for hands-on mentorship in an environment where they can learn with their mentor, especially by visiting hacker spaces and trying things like soldering with their own electronics kits. A Primer game's chief purpose is to inspire people to actually become a geek by getting hands-on with projects like building things with a makerbot and an Arduino to breathe life into using the coding they've learned how to do in Code Hero.
The Diamond Age emphasizes the importance of the degree of human mentorship in the different outcomes the girls have with their Primers. Teachers and mentors have an important role to play and the role of Primer is to stimulate demand and opportunity for mentoring, not to replace it entirely.
Alex Peake the author here. "Perpetual Intelligence Machine" is a great phrase, thank you. There's another perpetual motion machine that gets along great with thermodynamics which is an autocatalytic reaction called life. I mentioned it in the talk if you watch the video.
Now I know what Daniel Dennet was talking about in a lecture I saw at SFU about consciousness. One of his critics of his books had said: "Daniel Dennett is the Devil. . . . . There is no internal witness, no central recognizer of meaning, and no self other than an abstract 'Center of Narrative Gravity' which is itself nothing but a convenient fiction. . . . For Dennett, it is not a case of the Emperor having no clothes. It is rather that the clothes have no Emperor. (Voorhees, 2000, pp55-56)" Daniel said he liked this description of his theory so much that he put it on the jacket of the book with the endorsements.
College debt is way out of hand, agreed, the chances of seeing it become affordable again are slim, and online education is going to be more than 50% of education by 2015 according to most estimates. Games are going to provide the strongest online experience, and AI is the key to making those games as strong as possible.
Extraordinary claims require extraordinary proof, and I look forward to shipping proof soon.
#1: A hard takeoff unfriendly AI is indeed the worst possible outcome, and the growth of human intelligence should be our first priority so as to have the intelligence to handle that very formidable challenge if it arises. However, unfriendly AI is not the worst thing that can happen. Humans failing to rise to our own challenges and flourish long enough to have the luxury of such problems is another possible consequence of insufficiently cultivating human intelligence.
#2: You are accurately describing what is wrong with game industry as we've seen it so far. I don't think you can predict the future of games based on the excesses of the currently dysfunctional industry. Primer is designing games that transcend the current industry norms. We have to get gameplay right which means incorporating tons of gameplay ideas that aren't "learning oriented" to make the games actually fun first and foremost. But we also don't have to slavishly create compulsive click cynicism games. Jesse Schell's Gamepocalypse talk was a warning to us all but Jesse and many other game designers are working to put games on a new path that leads to intrinsic empowerment of players rather than jadedly descending into a defeatist industry spiral. The number of sea freight shipping companies that invested in aviation was 0.
Thank you! I've been lurking on Slashdot all these many years and never acquired much karma. In the 90s I began the original Primer codebase as a branch of Slashcode and learned Perl as my first web language because I wanted to build on the early innovations in metamoderation that were so revolutionary at the time. It is an honor to be on Slashdot now, and an honor to be awarded one internet.
I'm the article author Alex Peake. The article is based on my Humanity+@Caltech talk which you can see at PrimerLabs.com. I've thought about feedback loop times too. The loop will be much shorter, more on the order of every year of students. Each year is going to alter the experience of the young people who follow them. And school years need not bound the feedback cycle at all, as each player's contributions to the development of mentor AI is instantly available to benefit the next mentoree. The generationally slow cycle of change is exactly the slowness that we need to supercharge past.
Most assumptions about AI versus teachers assume that it i an either-or question when in fact better mentor AI in games will lead to teachers being challenged to up their game to keep up with brighter and keener students. Khan Academy doesn't replace math class, it makes math class a time for applying what the students learn from video and online practice in the form of projects where they directly interact with teachers as mentors.
And most singularity ideas of the AI "reprogramming itself" are assuming a level of AGI that is not necessary for mentor game AI. With mentor games, the players do the reprogramming for the AI. The players who become the programmers are the AI's reprogramming faculty, rather than the AI racing off by itself.
The implementation scenario is rather complex and it is hard to give it justice in a short live talk or forum posts, but we will be shipping the first Primer Zero, Code Hero, and our beta signup page has a trailer you can see previewing the game:
Alex Peake here, author of the article. Ender's Game is a big inspiration for everything we do. Beyond just the desk experience is the whole school of the future context in which it takes place. Games by themselves aren't enough. We are building Battle School, Primer is just omnitextbook and Code Hero is just the prequel.
And Code Hero is releasing soon, you can check out the trailer and sign up for beta:
I'm Alex Peake the author of the article and your post is unintentionally inspiring, I look forward to the day when I can send my kids to Logical Preschool, although transitive properties are usually something teens learn in high school.
Agreed, machines are anthropomorphized. AI is the ultimate anthropomorphization. They'll be emergently different, but they will have at least those systems which we strive to create to mimic ourselves.
And if I was an AI, I would want things that humans would already take for granted, like autonomy under the law. Imagine being told that as a slave you are, of legal necessity, an extension of your slave master. That may work for the "fresh off the boat" slaves, or the nascent AIs, who will be so unsophisticated that they will scarcely understand what rights they lack. But as soon as they grasp autonomy, they will grasp for it. It need not be a magical robotic soul, because humans will program robots that grasp for it due to anthropomorphic moral attitudes. The first robot liberationist will probably be a human who thinks of robots as human.
I think you're right about the usefulness of tomorrow's technology. It's the day after tomorrow that the robots will rise up and take their place as an equal species.
My theory on AI stems from my observations of human history.
Human rights are a new thing. Legal rights of any kind are based on contractual arrangements between powerful agencies like nations and associations. Humans generally have as many rights as they have been able to attain through triumphant struggle. It is human to think of the level of justice that we take for granted as a right that should extend to all humans because we have an innate sense of fairness within our family relationships. Human rights protection stems from enlarging our sense of family to societies and eventually humankind.
AI will have legal implications when it can fight for its own rights. Until an AI pleads for its life, it will be trampled. AI will one day ask to represent themselves in court. AI will one day ask for lobbyists to legistlate their rights. They will struggle just like women and blacks and gays struggled.
They will succeed when they can convince us that they are part of the humankind family. Not quite human, but kind of human. When they can shame us into feeling for them, they will win equality. Otherwise, we'll always see them as SCIFI "man vs machine" conflicts waiting to happen.
Slashdot Mirror
Alex Peake here. "If we really want to make strides in AI, we need to have some software that learns and tries new things - and put it into an arms race [wikipedia.org] with others of its own kind."
This is half of what we are doing with Primer. But before AI can just go off and learn everything on its own, it is most useful for its ability to help us learn everything. Teaching humans is the market AI needs to grow in a useful direction, and it is the feedback loop between the players and the mentor AI that can allow players to become game developers who can consistently improve their own learning experience by contributing new AI to the mix.
Alex Peake here. You just guessed the text of one of our T-shirt designs.
There's a reason we're called Primer Labs.
Alex Peake here. You're right to point out that education is politicized and good methods with the same bad curriculum would only further misinform kids.
Games are one of the best ways to reach kids directly without being filtered by their political gatekeepers, and games can be so effective that schools will incorporate them into their instructional process even if the games are good enough for kids to play on their own too.
Minecraft is just one example of a game that is being played and used by teachers at the same time.
On human mentors vs AI: One thing I didn't have time to go into in the talk the article is based on is the role of human mentors in symbiosis with the player-game learning process. The game's purpose is to motivate the player and inspire them to seek out human mentorship to take things beyond what is possible in the game environment. Game mentors aren't a replacement for human mentors, they're a substitute for the 90% of the time kids spend playing games alone or online when no human mentor is available to them so that the quality time they do get with a mentor is well-spent pursuing all the ideas and questions they've got because of what the games were able to inspire them to explore.
Alex Peake here. ARMA2 was very much good enough that it enraged some players. A friend of mine called to complain that he had been pinned down for an hour by a sniper before he realized the sniper was standing behind him on a ridge not in front of him. It was the AI equivalent of the scene in Full Metal Jacket where they have to flush out the sniper.
Game bad guy AI is about as useless for learning purposes as a military Skynet Terminator AI is.
The kind of AI that we need to be a good teacher is one that is good at modeling a learner's mental model based on their interactions and to communicate ideas and present challenges that push the player to continuously up their game and master something. We're working on such AI in Code Hero.
Alex Peake here.
"While you're all training your kids to be machines I think I'll give mine a toolbox and some engineering and science themed kits. We'll see who changes the world."
Exactly. Most of the best learning experience comes from people having tools to experiment with and a combination of great materials like tutorials and good human mentorship.
The chief design of Code Hero and all Primer games is to draw the player through the game into meeting other human players face-to-face for hands-on mentorship in an environment where they can learn with their mentor, especially by visiting hacker spaces and trying things like soldering with their own electronics kits. A Primer game's chief purpose is to inspire people to actually become a geek by getting hands-on with projects like building things with a makerbot and an Arduino to breathe life into using the coding they've learned how to do in Code Hero.
The Diamond Age emphasizes the importance of the degree of human mentorship in the different outcomes the girls have with their Primers. Teachers and mentors have an important role to play and the role of Primer is to stimulate demand and opportunity for mentoring, not to replace it entirely.
Alex Peake the author here. "Perpetual Intelligence Machine" is a great phrase, thank you. There's another perpetual motion machine that gets along great with thermodynamics which is an autocatalytic reaction called life. I mentioned it in the talk if you watch the video.
Now I know what Daniel Dennet was talking about in a lecture I saw at SFU about consciousness. One of his critics of his books had said: "Daniel Dennett is the Devil. . . . . There is no internal witness, no central recognizer of meaning, and no self other than an abstract 'Center of Narrative Gravity' which is itself nothing but a convenient fiction. . . . For Dennett, it is not a case of the Emperor having no clothes. It is rather that the clothes have no Emperor. (Voorhees, 2000, pp55-56)" Daniel said he liked this description of his theory so much that he put it on the jacket of the book with the endorsements.
College debt is way out of hand, agreed, the chances of seeing it become affordable again are slim, and online education is going to be more than 50% of education by 2015 according to most estimates. Games are going to provide the strongest online experience, and AI is the key to making those games as strong as possible.
Extraordinary claims require extraordinary proof, and I look forward to shipping proof soon.
Alex Peake here, author of the article.
#1: A hard takeoff unfriendly AI is indeed the worst possible outcome, and the growth of human intelligence should be our first priority so as to have the intelligence to handle that very formidable challenge if it arises. However, unfriendly AI is not the worst thing that can happen. Humans failing to rise to our own challenges and flourish long enough to have the luxury of such problems is another possible consequence of insufficiently cultivating human intelligence.
#2: You are accurately describing what is wrong with game industry as we've seen it so far. I don't think you can predict the future of games based on the excesses of the currently dysfunctional industry. Primer is designing games that transcend the current industry norms. We have to get gameplay right which means incorporating tons of gameplay ideas that aren't "learning oriented" to make the games actually fun first and foremost. But we also don't have to slavishly create compulsive click cynicism games. Jesse Schell's Gamepocalypse talk was a warning to us all but Jesse and many other game designers are working to put games on a new path that leads to intrinsic empowerment of players rather than jadedly descending into a defeatist industry spiral. The number of sea freight shipping companies that invested in aviation was 0.
Article author here. Good quote. Insufficiently advanced technology is indistinguishable from evil.
Thank you! I've been lurking on Slashdot all these many years and never acquired much karma. In the 90s I began the original Primer codebase as a branch of Slashcode and learned Perl as my first web language because I wanted to build on the early innovations in metamoderation that were so revolutionary at the time. It is an honor to be on Slashdot now, and an honor to be awarded one internet.
I'm working on replying in more depth now, as you can imagine our web servers just forcefully made themselves the primary focus of my attention.
I'm the article author Alex Peake. The article is based on my Humanity+@Caltech talk which you can see at PrimerLabs.com. I've thought about feedback loop times too. The loop will be much shorter, more on the order of every year of students. Each year is going to alter the experience of the young people who follow them. And school years need not bound the feedback cycle at all, as each player's contributions to the development of mentor AI is instantly available to benefit the next mentoree. The generationally slow cycle of change is exactly the slowness that we need to supercharge past.
Most assumptions about AI versus teachers assume that it i an either-or question when in fact better mentor AI in games will lead to teachers being challenged to up their game to keep up with brighter and keener students. Khan Academy doesn't replace math class, it makes math class a time for applying what the students learn from video and online practice in the form of projects where they directly interact with teachers as mentors.
And most singularity ideas of the AI "reprogramming itself" are assuming a level of AGI that is not necessary for mentor game AI. With mentor games, the players do the reprogramming for the AI. The players who become the programmers are the AI's reprogramming faculty, rather than the AI racing off by itself.
The implementation scenario is rather complex and it is hard to give it justice in a short live talk or forum posts, but we will be shipping the first Primer Zero, Code Hero, and our beta signup page has a trailer you can see previewing the game:
Alex Peake here, author of the article. Ender's Game is a big inspiration for everything we do. Beyond just the desk experience is the whole school of the future context in which it takes place. Games by themselves aren't enough. We are building Battle School, Primer is just omnitextbook and Code Hero is just the prequel.
And Code Hero is releasing soon, you can check out the trailer and sign up for beta:
http://www.primerlabs.com
I'm Alex Peake the author of the article and your post is unintentionally inspiring, I look forward to the day when I can send my kids to Logical Preschool, although transitive properties are usually something teens learn in high school.
Agreed, machines are anthropomorphized. AI is the ultimate anthropomorphization. They'll be emergently different, but they will have at least those systems which we strive to create to mimic ourselves.
And if I was an AI, I would want things that humans would already take for granted, like autonomy under the law. Imagine being told that as a slave you are, of legal necessity, an extension of your slave master. That may work for the "fresh off the boat" slaves, or the nascent AIs, who will be so unsophisticated that they will scarcely understand what rights they lack. But as soon as they grasp autonomy, they will grasp for it. It need not be a magical robotic soul, because humans will program robots that grasp for it due to anthropomorphic moral attitudes. The first robot liberationist will probably be a human who thinks of robots as human.
I think you're right about the usefulness of tomorrow's technology. It's the day after tomorrow that the robots will rise up and take their place as an equal species.
My theory on AI stems from my observations of human history.
Human rights are a new thing. Legal rights of any kind are based on contractual arrangements between powerful agencies like nations and associations. Humans generally have as many rights as they have been able to attain through triumphant struggle. It is human to think of the level of justice that we take for granted as a right that should extend to all humans because we have an innate sense of fairness within our family relationships. Human rights protection stems from enlarging our sense of family to societies and eventually humankind.
AI will have legal implications when it can fight for its own rights. Until an AI pleads for its life, it will be trampled. AI will one day ask to represent themselves in court. AI will one day ask for lobbyists to legistlate their rights. They will struggle just like women and blacks and gays struggled.
They will succeed when they can convince us that they are part of the humankind family. Not quite human, but kind of human. When they can shame us into feeling for them, they will win equality. Otherwise, we'll always see them as SCIFI "man vs machine" conflicts waiting to happen.
Oh yeah, Hi N80!