4 ms·
Every time I see something about trying to control an LLM by sending instructions to the LLM, I wonder: have we really learned nothing of the pitfalls of in-ban
by thenewwazoo 8mo ago
Every time I see something about trying to control an LLM by sending instructions to the LLM, I wonder: have we really learned nothing of the pitfalls of in-band signaling since the days of phreaking?
- Mountain_Skies 8mo agoIt'll be a sad day for Little Bobby Tables if in-band signaling ever goes out of fashion.
- quadrature 8mo agoSure but the exploit here isn’t prompt injection, it is an edge case in their billing that isn’t attributing agent calls correctly.
- thenewwazoo 8mo agoThat's fair - I suppose the agent is making a call with a model parameter that isn't being attributed, as you say.
- cpa 8mo agoIt reminds me of when I used to write lisp, where code is data. You can abuse reflection (and macros) to great effect, but you never feel safe. See also: string interpolation and SQL injection, (unhygienic) C macros
- direwolf20 8mo agoAllowing phreaking was an intentional decision, because otherwise they could have carried half as many channels on each link.