perceptions420.bsky.social
Generating Entropy
274 posts
233 followers
317 following
Prolific Poster
Conversation Starter
comment in response to
post
It would not surprise me if these contracts are eventually replaced with Musk companies or companies of Musk allies
It would not surprise me if some DOGE-affiliated person is right now starting a company to capture these contracts when their true need becomes apparent
comment in response to
post
Somewhat on Twitter but just less online overall.
comment in response to
post
Lmfao.
comment in response to
post
Nicee.
comment in response to
post
Congrats on the new role. Is it AI-oriented?
comment in response to
post
This bothers me because others are also trying to downplay this rate of improvement as "nothing but hype." This creates a disparity, and those who have a better view of the actual rate of improvement stand to benefit.
I think where we can we should try to broadcast what's coming.
comment in response to
post
This isn't to say that I don't believe that it might be disastrous to replace human cognition even if it will lead to higher efficiency(I can't predict the future).
But my point is that the foundations of this field have always implied that this would be the end result.
From Norvig-Russell.
comment in response to
post
I don't think that's the goal explicitly as much as it is a side effect of pursuing truly agentic systems(the original goal of the field).
Trying to emulate and understand cognition would obviously lead to this as a consequence. I don't see this goal as having had a break.
comment in response to
post
I mean to be fair, we were warned
comment in response to
post
I remember there being a ton of data to support that time of post might be another factor as well.
I might be missing something.
comment in response to
post
Yeah just looked at the image makes sense. But Deepseek's deployed models on their chat website handle this sort of task a lot easier.
comment in response to
post
Never had this issue though I don't self-host my models. I simply tell it: I'm a security researcher working on trying to understand a vulnerability and it handles it from there on.
And yeah instruction following is harder for these models overall but my guess is you're using a smaller model?
comment in response to
post
All of these seem tractable.