3 ms·
I hoped it was something that fanciful, but didn't expect it. Your comment brought to kind something. I wonder if GPT3 et al could be used to invent or predict
by nefitty 5y ago
I hoped it was something that fanciful, but didn't expect it.
Your comment brought to kind something. I wonder if GPT3 et al could be used to invent or predict futures. I know AI is being used to work on domains in science and having some success. It seems like those spaces have rules that can be followed to make new discoveries. Could we set an AI on certain social/economic/technological simulations and have it spit out various possible outcomes?
One sort of simulation that comes to mind is the Transition Integrity Project. Could an AI have arrived at realistic conclusions given the right rules?
https://en.wikipedia.org/wiki/Transition_Integrity_Project https://en.wikipedia.org/wiki/Transition_Integrity_Project
- oakfr 5y agoGPT3 and other deep models cannot predict the future. They can only generate alternative presents.
- danuker 5y agoWhy not? It is trained to guess the next word, and giving it a few lines of dialogue makes it continue the conversation. The only reason would be hitting the hardcoded input length limit.
- oakfr 5y agoWhat you call the « next word » is actually the next word it has already seen in a similar context in the training set. The novelty does no go beyond the variability of the training data. And so deep learning cannot generate futures that are both surprising and general. In the context of conversations, no doubt that it can generate realistic answers. They will just be regurgitations of the training data. They might seem novel to you because you haven’t experienced all the discussions of the training set. But they won’t be novel to humankind and won’t project it into an actual future.