New Research Tests Whether AI Models Know Their Own Thoughts | Quanta Magazine
1 directory member surfaced this signal.
“Some AI models can detect injected “thoughts” without being trained to do it. Only sometimes, but how did that ability emerge? Via @quantamagazine.org @benbenbrubaker.bsky.social I found this to be an interesting read.” evidence ↗
“Some AI models can detect injected “thoughts” without being trained to do it. Only sometimes, but how did that ability emerge? Via @quantamagazine.org @benbenbrubaker.bsky.social I found this to be an interesting read.”