thinks about icy planetary geophysics, space systems engineering, American Pragmatism, and mindfulness
South Pole–Aitken basin
Joined September 2011
- Tweets2.9K
- Following884
- Followers257
- Likes51.7K
opus 5.5.. may in principle be able to do character design? but I'm not smart enough to actually get it to do that
Sam, just like the Nestle guy who said "water is not a human right", is 100% correct
Qwen is getting its own BCI before us
I've been vibe coding my dream game, a cozy wii-sports-resort-vibes version of flight sim-cade Death Stranding for the Steam Deck. currently working on improving landscape texture blending and incorporating sediment deposition modeling into the world generator
new interfaces are all that's going to matter once unlimited superintelligence is on tap. we're all bottlenecked on text atp, and it's just a march towards higher and higher bandwidth interfaces from here on out
🛰️ retweeted
This is the central dogma of the space industry, accepted uncritically by almost everyone until they pick up the phone to get a quote from a launch provider. A better request would be to start a company that lowers launch prices, which are going up @garrytan
so whats the plan for when robots do chores in every household --> adversary government backdoors into every robot --> exactly one weight clamped in every robot's decisionmaker model --> ???
was Steve Jobs secret sauce just prompting humans like LLMs, decades early? "Don't give up, believe in yourself!"
is seeking mathematical beauty a suboptimal heuristic, nothing more than the relief of sparse encounters with comprehensible utility in a gnarled, opaque jungle of power?
does anybody offer ITAR-compliant Gemma 4 31B (or smarter) inference at >1000 TPS? asking for a friend
in other words, if you fourier transform a signal and fuck up every signal at frequencies above those associated with information-bearing envelope modulations, the signal... continues to bear information
Speech can survive being flipped backward every twenty milliseconds.
The same speech in backward pieces of 200 then 100 then 50 then 20 milliseconds.
The original comes last.
When do the words become clear?
A listening demo from @OrukLabs.
Peirce's Pragmatic Maxim handily resolves the problem of the Ship of Theseus. The answer is in the terms of the OEM warranty
🛰️ retweeted
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/
forget stage 4 cancer, nobody knows how *Tylenol* works!
