Higgenbotham wrote:
> I was on the floor of the CBOT in 1994 talking to a trader who
> said he was using a neural network. If that's what you think
> programmers are going to lean on to beat human traders, something
> different will need to be devised because after 24 years of
> development it hasn't been successful. The billionaires are still
> humans and there's no Deep Blue equivalent trading computer that's
> amassed a billion dollars and there won't be.
In 1994, computers could beat humans at both chess and go.
Computers were not powerful enough to handle more than a few megabytes
of data to train a neural network, and the data had to be compiled and
fed in.
Today's computers are tens of thousands of times more powerful than
they were in 1994.
Today, a computer could trivially beat any human at chess using a
brute force minimax algorithm.
Today, Google's computer could beat anyone at go using an advanced
neural network algorithm augmented with their deep learning
algorithms.
Today, computers are powerful enough today for neural networks to be
trained on infinitely more data than Warren Buffett could learn from
in his entire lifetime. Not only that, it can learn on its own by
scanning the internet.
If humans are still better at some things today, that list of things
will be reduced with time.
By 2030, multiply that power by more tens of thousands, and humans
don't stand a chance. I won't be alive to see it, but maybe you will.
Enjoy!