SoylentNews is people
SoylentNews
from the that's-what-they-want-us-humans-to-think dept.
From TFA:
If you thought artificial intelligence was already overhyped to death, this week will have given you a heart attack. On Monday, excitement levels among hacks hit the roof amid claims Facebook had scrambled to shut down its chatbots after they started inventing their own language.
Several publications called the programs "creepy." Some journalists implied Facebook yanked the plug before, presumably, some kind of super-intelligence reared its head. The UK's Sun newspaper demanded to know: "Are machines taking over?" Australian telly channel Seven News even went as far as to call it an "artificial intelligence emergency." Newsflash: it isn't.
[...] Zachary Lipton, an incoming assistant professor of machine learning at Carnegie Mellon University in the US, told The Register this week: "The work is interesting. But these are just statistical models, the same as those that Google uses to play board games or that your phone uses to make predictions about what word you're saying in order to transcribe your messages. They are no more sentient than a bowl of noodles, or your shoes."
Previously:
AI is Inventing Languages Humans Can't Understand. Should We Stop It?
(Score: 0) by Anonymous Coward on Friday August 04 2017, @02:02PM (1 child)
Yeah, that is what I meant by "progression mechanisms." In Atari games the progression mechanism is very easy to quantify - it's your score in the game. But that's still something that even the most sophisticated AIs need to be told. Whereas on the other hand I think if you took a human, even one that had never once played or had any knowledge a digital game in their life, they would likely eventually come to understand that progression means increasing your score. For now no AI is really capable of developing their own progression mechanisms. When that is possible... well, things will get very interesting.
(Score: 0) by Anonymous Coward on Friday August 04 2017, @02:36PM
That makes me think: Progression is achieving something that takes effort. Something that doesn't take any effort quickly gets boring. On the other hand, something that doesn't show progress, even after a lot of effort, sooner or later will make you give up.
So basically the key here seems to be an expected-effort curve: Too little expected effort makes it too boring to do. Too much expected effort makes you give up. All that assuming purely intrinsic motivation, of course; if you (expect to) get paid for it (be it in money, or in reputation), it may considerably change the equation.
Of course the optimum point of expected effort is individually different. People who set their optimum unusually low we call lazy: They only do things that take very little effort. People who set their optimum unusually high may be considered stubborn — they likely invest much effort for little chance to succeed, but if they are lucky, they might achieve something great that others wouldn't.
Indeed, there's probably also learning at the meta level involved: We learn where we should set our optimum effort point to maximize our results.
Actually stated that way, the only hard part seems to be estimating the required effort to reach a goal. But then, that sounds like just another learning task to me. Just have positive feedback for achieving something, proportional to effort, and negative feedback for effort that didn't achieve anything.
Thinking again about that: It also means that also the ability to be frustrated probably is inherent in general intelligence. Maybe Douglas Adams was on the right track with the idea that the prototype robot with Genuine People Personality would be constantly depressed …