What do you think guys? Can we learn to beat big rigs?
P.S. I always knew everyone posting here was a robot.
I tried very hard to come up with something witty to say about this, but failed.
2 get.
Is it the future now?
>>4
The future was yesterday, welcome to our postmodern dystopia.
DQN DQN LOL
I would get better at a game if I played it for two weeks straight too. But anyway, an advanced AI is ought to put some people out of job. And perhaps make this world a better place by make politicians and their struggle for influence and power outdated and irrelevant.
Which is probably why all those scientific turbonerds signed a petition to limit AI research. "Ooh, it's the end of humanity, " — they whine forgetting that there wasn't much of it to begin with.
These are not new concepts and I get frustrated when pop news is so late to the party.
>>7
I personally think it's quite the breakthough in machine learning
>>8
It's not as impressive as WATSON or any of the AVIDA stuff that the BEACON coalition put out, I really think google's lagging behind in this one.
>>9
Actually, WATSON is definitely less impressive than DQN. The ability to comprehend and learn rules on the go based on loose visual information is much more impressive and closer to human beings. WATSON was specialized on two tasks, querying a huge database of information really quick and find and understanding written english. Google leads one of these areas already, I'll let you guess which one. It boils down to the fact that WATSON can't learn new shit by itself, whereas DQN can.
>>8
We're working on different timescales I think. I heard about this some time last year.
>>11
Yeah, that timespan seems pretty short to me, but if you are complaining about breaking news on tech magazines not being that much of breaking news, then I guess your complain is sensible.
>>12
Yeah, that's what I was complaining about.
As an aside, Watson's approach seems to be that of "take AI ideas from the 80s and throw lots of 21st century hardware at it" so it's not particularly impressive, even compared to other things.
>>13
I thought Watson had NLP far, far beyond the 80s, ignoring hardware. Can you provide information on the design of this approach to handling language, especially compared to what was available in the 80s?
>>14
I can't really, but NLP is making huge advances in general because people are in love with it. You can get impressive results in various forms of semantic analysis even with a (relatively) small data set. Watson even has an offline copy of Wikipedia in its 4TB data store, which in my mind is kind of cheating -- all else being equal considering the state of the art in NLP, it's a glorified expert system.