Gonna be a lot more money to make timing the software short than chasing the semi softness imo, but def needs a wingier mindset.
For what it’s worth if u do want a permabear’s view, this isn’t even that scary. Again, bull ends when the final bear loses hope, but wow are things orderly.
Something has to explode over the weekend. I would guess the reality game over would be if ANTH lied, and hence the rush
Tbh as a lay person reading this, I was fairly surprised at the conclusion as well. Like the jump to “killing” superposition (with “scale”) because an earlier interpretation couldn’t see a middle ground seems weird.
Seems to be coming from a place of “I realized”.
As hopeful as stuff like this (& funnily enough a strongly indicative that large scale agents = AGI), there’s still little to show that test cycles (even simple ones) are anywhere close in time to training runs
Good news is that progress is so quick maybe we can spare some time
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
Loathe retweeted
Mrs. Buff is drunk texting me pics on girls night out. I’m exactly where I wanna be #WifeGuy
I honestly just don’t believe in such absurd lows of internal correlation. I get it: the S&P is the greatest invention of modern human civilization but still there are limits.
