AI context limits pose real safety risks

Summer Yue leads AI Safety and Alignment at Meta. She is doing everyone a favour when she talks through her experience with OpenClaw recently. Reading through the x.com thread (link below), she had been experimenting with a tame inbox (or 'toy inbox' as she puts it) and things had been working well. "

She then asked it to organise her much larger very much non-toy inbox. It appears that her instruction to 'Check this inbox too and suggest what you would archive or delete, don’t action until I tell you to' got compacted -- that is, as the system ran, it ran out of context and appeared to forget her original instruction.

She describes running to her Mac Mini to kill the processes. I can imagine the absolute horror. Lucky she was nearby the machine!

There are quite a few in her X.com thread criticising, especially given Summer's position and responsibilities. I think that's to miss the reality of what we're all collectively experiencing at the moment. She was testing -- and she's uncovered a key challenge with the technology stack at the moment. Irrespective of your view of AI at the moment, especially if you're a huge fan of OpenClaw, I think she does everyone a service by highlighting the challenges and risks with the likes of this technology.

"If you haven't tested OpenClaw, I'd strongly encourage you to do so in order to experience the potential yourself (but be very sensitive what you give it access to!)