Emerging reasoning with reinforcement learning | Hacker News
news.ycombinator.com
Emerging reasoning with reinforcement learning | Hacker News
Large language models sometimes pick up skills like logical reasoning or translation even though nobody wrote rules for them.
This paper says those surprise skills pop up because the network itself is a huge tangled dynamic system, so fresh abilities surface when inner interactions cross certain tipping points.___LINEBR... See more
you will have to do something no current machine can do reliably: reason and act on something new and unexpected, based not on some immense database of prior experience, but on a powerful and flexible understanding of the world.
Ernest Davis • Rebooting AI: Building Artificial Intelligence We Can Trust
AI’s ability to identify patterns, make predictions based on them, and continually learn from outcomes may not create human-like intelligence, but it creates something arguably more valuable and practical: the ability to adapt in response to uncertainty.