Anyone talking about reward functions is not talking about AGI. This is the disconnect between brute force results and an actual cognitive architecture. DL + Time is the reward function for the posts doing well, BTW, because it fits the existing model.
Anyone talking about reward functions is not talking about AGI. This is the disconnect between brute force results and an actual cognitive architecture. DL + Time is the reward function for the posts doing well, BTW, because it fits the existing model.