Subtitles section Play video
-
Greetings Professor Falken, Shall we play a game?
-
Machine learning, neural networks and deep learning are all buzzwords in AI circles.
-
But essentially they’re all referring to the attempt to teach a computer to recognize
-
a dog in any photo, or translate a language in real time, or… lots of other tasks
-
To do that, engineers present a computer with a lot of data, and teach it by repetition;
-
similar to teaching an animal a trick... encourage it when it does well, and discourage it when
-
it’s wrong.
-
Eventually, the AI will get better at facial recognition or predictive text, or… whatever!
-
But the thing is, even if it can spot the dog, the computer has to do the calculation
-
every time.
-
It has to scan every pixel and check everything.
-
It can’t anticipate what you ask, it can’t realize that usually the dog is here or there
-
in the photo…
-
They don’t work that way.
-
Until now.
-
Engineers at ‘DeepMind’ -- Google’s AI system -- wanted to give their computer
-
a memory.
-
As in, the ability to remember the tasks it had done before… and learn from its success
-
and failures on those tasks.
-
They did this in… sort of a funny way.
-
They taught the system to play Atari, and published a paper on their findings.
-
I love science.
-
It makes sense!
-
Video games are about learning new skills through repetition and failure.
-
You have to build on what you learned!
-
As the game progresses, requirements usually grow in complexity, and have to be combined
-
with other skills…
-
All that requires memory, and the understanding of the tasks you’ve completed before.
-
But AI still needs lots of repetition and data to learn.
-
So, they made the computer play each game 20 million times.
-
Then, because they added the ability of the machine to remember what it did before...
-
Deepmind could apply what it learned playing Pong to playing Fishing Derby, or Kung-Fu
-
Master.
-
AI is already good at playing other games, like Go and Chess -- but it looks at the board,
-
and then calculates the best options for each move.
-
Each time.
-
Imagine if a computer could keep things in mind when learning to solve a new problem,
-
like realizing opponents trained by certain teachers would always try some special move.
-
Or more practically, realizing you always take the same route home, every day.
-
It gets a little freaky…
-
Do we want machines to be able to learn?
-
To anticipate actions?
-
To be able to get set in their ways over time?
-
Bae Stephen Hawking said artificial intelligence will be ”either the best, or the worst thing,
-
ever to happen to humanity.”
-
But as of now…
-
AI ain’t ready.
-
It had to play those video games 20 million times.
-
Even the most hardcore Atari gamers never approached that and they’ve mastered all
-
those games.
-
Over time, AI learns and then it just… does it, but unlike life it doesn’t have the
-
ultimate motivator…
-
It wasn’t programmed to die if it fails at gathering resources.
-
It wasn’t programmed to reproduce…
-
It’s just a machine doing a simple, single job.
-
Even though now, AI might acquire memories, we probably shouldn’t read too much into
-
it.
-
Jerry Kaplan, a futurist and computer scientist wrote in the MIT Technology Review…
-
“The robots may be coming, but they are not coming for us—because there is no “they.”
-
Machines are not people, and there’s no persuasive evidence that they are on a path
-
toward sentience.”
-
Are you still freaked out by AI?
-
Cause you shouldn’t be.
-
Watch this video right here to learn why and also let us know in the comments what you
-
think about all this stuff and come back for more episodes.
