4 ms·

Imagine sprinting to build a state of the art LLM only to have the AI safety team severely cripple the model's usefulness before launch. I wouldn't be surprised

Imagine sprinting to build a state of the art LLM only to have the AI safety team severely cripple the model's usefulness before launch. I wouldn't be surprised if there was some resentment among these teams within Google DeepMind.

kenjackson 3y ago
Much easier to loosen the rails than to add guardrails later.
- lordswork 3y ago
  Sometimes the easy path is not the best path.
hiAndrewQuinn 3y ago
that's actually the uhh Non-Human Resources department thank you,
bsdpufferfish 3y ago
Imagine hiring “ai safety experts” to make a list of grep keywords.
- lobocinza 3y ago
  "AI safety" is humans censoring humans usage of such tools.
  - vacuity 3y ago
    And we seem to be lacking in "human safety", which I find more egregious.
phatfish 3y ago
It gets lumped under "safety", but I bet it is also due to perceived reputational damage. The powers that be at Google don't want it generating insecure code (or to look stupid in general), so it is super conservative. Either it ends up with people comparing it to ChatGPT and saying it generates worse code, or someone actually uses said code and moans when their side project gets hacked. I get the feeling they are touchy after Bard was soundly beaten by ChatGPT 4.