Is the next Einstein an AI System?
Technically, AI can now do many cognitive tasks much better than humans, molecular modeling is one of them, NLP and image recognition are among the most well-known. But will AI become truly creative? Or in other words, can AI predict unknown unknowns. We already know it is great in discovering and solving known unknowns?
Philosophically speaking the question is if the next Einstein will be an AI system. The answer is most likely yes. The saying is that humans are unbelievably complex. I have found that humans are extremely simple, always applying simple heuristics. The point is that we are not aware of these heuristics and come up with super complex models and explanations, leading at times to wild conspiracy theories. Mr.Trump is embodying this with his wild theories, ranging from vaccination theories to Obamagate, while his simple heuristic is “do what will most likely get me re-elected”, unimpeded by any ethical constraints and restrictions.
We just need to apply Occam’s razor “the simplest solution is most likely the right one”.
Gerd Gigerenzer has written a series of books, the most well known is “Simple Heuristics That Make Us Smart”, where he explains that many complex problems can be much better solved with simple heuristics,
In my own research over the last twenty years I have been trying to uncover as many of these really simple heuristics as possible. The honest signals of collaboration are some of them, digital virtual tribes is another one.
Once we have found all of these simple heuristics, we will teach them to the AI system, which then can take over and be the next Einstein. But this is still a long process, as it is awfully hard to discover these simple heuristics.
In the meantime I think we still need to try to obtain a deep understanding of each problem, to identify the underlying simple heuristic that solves it. In the case of Einstein himself, he was using his famous thought experiments to come up with simple and intuitive solutions to fiendishly complex problem that other physicists had been unable to crack. For instance, Einstein was famously riding an elevator in the patent office of Berne, when eureka struck him and he saw the principle of relativity, leading in the end to the simple equation E=m*c^2.
Once this library of simple heuristics is large enough, they can be combined into the next Einstein.
When will we have it? I don’t know, I am not Einstein.
Philosophically speaking the question is if the next Einstein will be an AI system. The answer is most likely yes. The saying is that humans are unbelievably complex. I have found that humans are extremely simple, always applying simple heuristics. The point is that we are not aware of these heuristics and come up with super complex models and explanations, leading at times to wild conspiracy theories. Mr.Trump is embodying this with his wild theories, ranging from vaccination theories to Obamagate, while his simple heuristic is “do what will most likely get me re-elected”, unimpeded by any ethical constraints and restrictions.
We just need to apply Occam’s razor “the simplest solution is most likely the right one”.
Gerd Gigerenzer has written a series of books, the most well known is “Simple Heuristics That Make Us Smart”, where he explains that many complex problems can be much better solved with simple heuristics,
In my own research over the last twenty years I have been trying to uncover as many of these really simple heuristics as possible. The honest signals of collaboration are some of them, digital virtual tribes is another one.
Once we have found all of these simple heuristics, we will teach them to the AI system, which then can take over and be the next Einstein. But this is still a long process, as it is awfully hard to discover these simple heuristics.
In the meantime I think we still need to try to obtain a deep understanding of each problem, to identify the underlying simple heuristic that solves it. In the case of Einstein himself, he was using his famous thought experiments to come up with simple and intuitive solutions to fiendishly complex problem that other physicists had been unable to crack. For instance, Einstein was famously riding an elevator in the patent office of Berne, when eureka struck him and he saw the principle of relativity, leading in the end to the simple equation E=m*c^2.
Once this library of simple heuristics is large enough, they can be combined into the next Einstein.
When will we have it? I don’t know, I am not Einstein.
Comments
Post a Comment