I still find this more impressive than chatgpt. Mostly for the fact that it knows what it's doing and why.
Al Faggo
>>101585604>Mostly for the fact that it knows what it's doing and why.it doesnt and neither does chatgpt
>>101585816prove that it doesnt.
>>101585869prove that it does.
>>101586045prove that it doesnt.
>>101585816For starters it knows what the purpose of the game is. Rather than just guessing what it's likely to be
>>101587137>For starters it knows what the purpose of the game is
>>101587179It is saying that because others have already said this. Alpha go and alpha zero after it were making strategies nobody has ever seen before.
>>101587202>It is saying that because others have already said this.
>>101585604>chatgptI find chatGPT infinitely more impressive because it demonstrates that robowaifus are on the horizon
>>101587202they were making things that looked like strategies but werent
>>101585604I still have no idea how to play this shit.
>>101589429It's pretty easy to start, but strategy is opaque to learn. It's a very abstract game so a lot of lessons and explanations are also abstract or proverbial. It's a great, great game. Learn to play on OGS.
>>101589388????????????
when will ai stop falling for the obvious (human detectable) tricks that exploit its assumptions about the other player's strategy? pretty sure all of these still have outlier failures if the opposing player does strategies no ordinary go player would do, explicitly intending to trick the network.
>>101585604this. ChatGPT is a good chatbot
>>101589786!!!!!!!!
>>101585604chatgpt is disappointing not impressive...
>>101585869>>101586087Prove that it does please
>>101586087>>101585869if it actually played by selecting a strategy, then it would need magnitudes fewer computing cycles to select its moves.
>>101589789These failures modes were discovered by an adversarial AI. Potentially you could use this adversary to generate training data to fix these failure modes. Problem is that requires work and katago the most popular open source AI is worked on by 1 programmer. Fineart is worked on by a team but I don't know if they still work on it.