AlphaGo goes 60-0 vs top GO players including Ke jie

youtube.com/watch?v=b_24iaUMRFs

AlphaGo is no the greatest GO playing entity on earth. Having gone 4-1 vs Lee Sedol, it is now 60-0 against the best players in the world.

Other urls found in this thread:

gogameguru.com/lee-sedol-defeats-alphago-masterful-comeback-game-4/
online-go.com/puzzles
twitter.com/NSFWRedditVideo

COMPUTERS MUST NOT BE SMARTER THAN HUMANS

>""""""""""""""""""""SMARTER""""""""""""""""""""

Reminder that computers are neither smart nor intelligent.

Is this like Deep Blue VS Kasparov so is there Google techs behind the scenes adjusting AlphaGo on the fly and consistently changing the rules to narrow the scope of the 'win''

What I consider "smart" really has nothing to do with being good at very specific tasks. We already know that computers are better at very specific tasks than humans, that is by no means a surprise.

Actual intelligence is figuring shit out on your own, without having someone program you. Even though DNNs are usually marketed as magic, they are not magic. Getting one to work requires a lot of work still, you need to carefully design the layout. The next step from DNNs would be self-organizing DNNs that figure out what kind of layout is best suited for a task. I'm not sure what the state there is, but I guess not very far. It should limit performance severely.

It is though. This success is a clear example of the overwhelming superiority of electric transistors as compared to human neurons. It is able to process an unimaginable number of games and communicate in a direct way via fast interfaces.

This is actually more impressive than being more intelligent or less intelligent. It is something even better, fundamentally superior in what seems to be every way.

A human brain with 250 IQ is not as impressive as what alphago can do in terms of processing speed and data transference. This should scare the fuck out of everyone

>This should scare the fuck out of everyone
It only scares the fuck out of people who don't know how computers work.
That's like saying a hammer should scare the fuck out of everyone because no human being can bang a nail as hard as it can. It's just a tool.

Oh yeah, can it explain the meaning of life or how a vagina feels like?

Checkmate atheists

go weeaboos btfo

their game has been reduced to tic tac toe.


same with chessfags, fuck them too

If an infinite number of monkeys were to slam their fists randomly on controllers playing GO against a computer which makes the MOST OPTIMAL move everytime, is it not a certainty that one of the monkey's would win eventually? Now extend that to best of 3. Eventually a monkey would randomly beat a computer at least 2/3 times, and therefore win the series.

Therefore the game is nothing more than a game of chance. You may as well roll dice and call it a day.

This computer doesn't understand shit at all, it's not intelligent.

>4-1
How did AlphaGo lose? Out of pity for puny humans?

Computers are still shit at pattern matching. Humans are still king when it comes to this.

What we need to do is figure out how to interface transistors with the human brain.

Okay buddy, lets see you out-perform grep.

I'm waiting.

All that stress from the match made his hair shrink. EITHER THAT OR AI SHRINKS HAIR

why do people act like this means that robots are better than humans at go?

all this means is that robots have become better than humans on small go boards.

Go can be played on many different sizes of go board.

Humans have a genuine understanding of Go and so would be able to use their heuristics to play on a 100x100 board if they really wanted to and they'd be somewhat worse than on 19x19 or smaller but alphago would perform dreadfully.

So really, this isn't a much bigger deal than deep blue

>but alphago would perform dreadfully
citation needed

see look at this idiot.

"omg the power of silicon defeaing neurons! the computer now literally understands the game better than humans!"

fuck off

read this blog
gogameguru.com/lee-sedol-defeats-alphago-masterful-comeback-game-4/

but if you don't know the rules of Go then you still won't really understand why lee won

>Computers are still shit at pattern matching. Humans are still king when it comes to this
not really, judging by the sheer number of retards that still walk the earth believing in absolute nonsense
people don't readily recognize patterns unless they are taught to recognize them

lphaGo has been trained on a standard full-size Go-board. Both it's move selection and position evaluation convolutional neural networks have been trained extensively on the basis of 130,000 amateur games (19x19) and millions of self-play games. Both neural networks depend heavily on the exact board position, especially the position evaluation. Both neural networks have been trained by reinforcement learning using the outcome of the game on a standard 19x19 board.

furthermore , part of alpha go still uses monte carlo tree search , which is totally dependent on board size.

you only need a few intelligent people for the species as a whole to advance.

Quick! How much is 466832 times 756432?
If this took you more than 0.01 seconds you must destroy your calculator at once.

You're in the lab and this guy laughs at your research paper. What do you do?

humans dont understand shit either

We just reproduce and survive

where do I go to learn Go strategies?

I get the game but I basically just make ladders with the CPU and it comes up with ways to eat my pieces

it scares people who don't understand what they're talking about, like you
>This success is a clear example of the overwhelming superiority of electric transistors as compared to human neurons.
complete and utter nonsense

online-go.com/puzzles

Go to the Puzzles tab. There's a lot of tutorial packs that are pretty good.

They're doing Starcraft 2 next which should be really cool.

Thank you

>the most highly optimized zerg rush ever performed

...

>switching from Go to starcraft
does google just really hate Koreans or something

some of us do not reproduce

>no human being can bang a nail as hard as it can
There is a "your mother" joke somewhere in there, but I can't find it (AlphaGo could, though).

It's *exactly* the same case for humans. Go is a game of memory and pattern recognition. No one really "understands" the game; strong players just acquire a database of millions of positions and learn their (sometimes minute) strategic differences. This is what we celebrate as "human intuition". Actual reading ahead is helped by this database, but still not fundamentally different from MC.

>computers are shit at pattern matching
Depends on the pattern and how the data is represented.