OpenAI's experiments in co-evolution (deep reinforcement learning and self-play) openai.com/blog/emergent-… reveals the emergent complexity of intentional agents competing and cooperating among themselves.
This kind of technology cannot be undersold. It's a paradigm shift not only in technological terms but also in scientific terms. Detractors who keep pitching expectation moderation are in fact performing a disservice to society as a whole.
Many investors lost their shirt when the dot-com burst. Yet society as a whole benefited wildly from the bust. Society discovered an extreme surplus of cheap fiber optics that accelerated progress in many other fronts.
Mass extinctions have beneficial effects for evolution. It's of course unfortunate for individuals but it does wonders for the collective whole. Thus, detractors of DL who worry more about an AI winter simply do not understand the dynamics of evolution.
It is bad enough that the general population is unaware of the progress of AI but "avoiding the AI winter" narrative is one way for the informed to shoot themselves in the foot by overthinking about the speed of progress. Nobody really understands the exact speed of innovation.
The value having precision about the velocity of progress is important when scarce resources are on the line.Humans have limited resources, more specifically our attention is a universally scarce resource.The velocity of progress is a guide to how much attention one should spare.
The AI Winter argument argues that there is sufficient uncertainty in AI progress that therefore one should not waste scarce resources on the activity. The flaw in the logic is that individuals scarcity can be someone else's abundance.
One can trace the acceleration of human innovation to the invention of credit. One can think of money as potential energy. By borrowing money from the future to fund the present, one essentially accelerates innovation.This borrowing happens when there is optimism.
In a world with pessimism where nothing ever makes progress, there is no incentive to borrow from the future to pay for the present. AI present velocity reveals that it is a no-brainer to borrow from the future to pay for the acceleration of the present.
@threadreaderapp unroll please
• • •
Missing some Tweet in this thread? You can try to
force a refresh
