POPA ReconIf the 🧠 training data runs out, will the development of AI end? source
If the 🧠 training data runs out, will the development of AI end? source 1. Fixed assumptions - The argument that AI needs to collect endless human work records in order to learn everything it does leans on the assumption that the current learning method will continue - Tom Davidson argues that if AI improves the learning algorithms themselves, the same capabilities can be learned with much less data. 2. Improve the learning method -The key doesn't end with having more of the right answers memorized - AI directly researching how to learn new tasks with fewer examples - Once this is possible,…
The excerpt is quoted from a public Telegram channel, which owns its text.