2 hours ago · Tech · hide · 0 comments

I think Open source LLM's will hit a ceiling for this one reason: safety guardrails. Today we see Mythos and GPT 5.6 Sol put under heavy scrutiny for the primary reason that it is too unsafe to release to the general public. The guardrails come in three layers - safety baked into the model itself, immediate flagging and offline batch analysis. Level 1: Baked into the modelHere's a strange example from Sonnet 4.5 Level 2: Immediate flaggingMythos was so cautious that you were not able to ask it a single question mentioning mitochondria. Level 3: Offline batch analysisI don't have an example ready but imagine FBI knocking on your door because you have repeatedly asked ChatGPT suspicious questions which were kinda valid individually - this is not hypothetical, something like this has happened in the past. Dimensions of unsafetyThe "unsafety" comes in a few dimensions cybersec capabilities that allow models to attack software in the wild and create space for hackers to do their thing…

No comments yet. Log in to reply on the Fediverse. Comments will appear here.