3 ms·
LLMs suffer the same problems as any Von Neumann architecture machine, It's called "key vulnerability". None of our normal control tools work on LLMs like ASLR,
by normalaccess 1y ago
LLMs suffer the same problems as any Von Neumann architecture machine, It's called "key vulnerability". None of our normal control tools work on LLMs like ASLR, NX-Bits/DEP, CFI, ect.. It's like working on a foreign CPU with a completely unknown architecture and undocumented instructions. All of our current controls for LLMs are probabilistic and can't fundamentally solve the problem.
What we really need is a completely separate "control language" (Harvard Architecture) to query the latent space but how to do that is beyond me.
https://en.wikipedia.org/wiki/Von_Neumann_architecture
https://en.wikipedia.org/wiki/Harvard_architecture
AI SLOP TLDR:
LLMs are “Turing-complete” interpreters of language, and when language is both the program and the data, any input has the potential to reprogram the system—just like how data in a Von Neumann system can mutate into executable code.
- fc417fc802 1y agoIsn't it more akin to SQL injection? And would a hypothetical control language not work in much the same way as parameterized queries?
- normalaccess 1y agoThe more I looked into it it's not just the control language itself we need but a way of querying the model that is completely orthogonal to human language. But I think that would be impossible because as the newer models grow they would soon understand the control language re-blurring the line between the control language and output language. "Speaking" in a new way will not be outside it's ability to pattern match. When your fundamental compute block is language itself (and not a subset) you bounce into the limits of our understanding of language and cognition. It's a new Tower of Babel we are building by pouring all of humanities records into a mold and hoping a tower to heaven pops out the other side.