satisfier of curiosity. 【物の哀れ】 may silicon save the world.
Joined July 2008
- Tweets3.2K
- Following341
- Followers126
- Likes16.7K
they called em lunchables because you were able to lunch, or so it was implied
death jockeying and infinite football fields
sbnation.com/a/17776-footbal…
i love eating and love reading, but it wasn't until i got a kindle that i realized i love eating and reading
these values are z-scored btw, so don't go interpreting values below zero as "negative pain". it just means below the average.
hightech lowlife retweeted
"reality has a surprising amount of detail" fallout, and wariness of what kind of entity might find reality less "surprising" for having details
hightech lowlife retweeted
@polynoamial I ran this experiment a month ago! waiting to rerun on better hardware to publish it. the bitrate is ~100 - maybe 500 bits/hr between GPUs
hightech lowlife retweeted
New cat alert! A call from a wildlife sanctuary about a "weird cat" has led to the first new cat species discovered in over 100 years in Bolivia’s Yungas forest ecoregion. Learn more about this feline surprise: on.natgeo.com/JRuaGj
hightech lowlife retweeted
Humans used their evolved neural networks to develop math over many centuries. In time, they showed more and more concepts could be formalized to enable rigorous proofs (Dedekind et al).
Then they showed that the notion of proof itself could be formalized and that simple mechanical processes (algorithms) could verify and search for proofs (Frege, Russell, Turing). They developed formal theories of algorithms, showing that some simple mechanical systems were universal (able to simulate all other such systems).
Humans built such universal mechanical systems (computers), spent decades scaling them and designing more efficient algorithms (Turing, Von Neumann, Knuth). While these computers could search for math proofs, they were very inefficient and so not of much utility.
But humans also created a simple mathematical model of neural networks and simulated these using their universal computers (McCulloch, Pitts, Hinton, Schmidhuber). Once these neural networks were roughly comparable in size to the human brain and trained on vast datasets of human verbal behavior, they were able to understand mathematics. Moreover, they were able to produce formalized proofs and have computers check them. This closed the loop, where mathematical structures themselves continue the process of mathematical discovery and the humans need not participate.
hightech lowlife retweeted
FYI there’s nothing novel about “making new kinds” of viruses and just means it was genetically modified. There IS something novel about using AI to do the lit review for you and have it work
Of course many headlines about this are sensationalist and fearmongering. Check Andrew’s thread for the full publication!
Basically what they showed is that the 2 genome models they used are actually smart enough to identify compatible gene modifications, which replaces a scientist doing weeks of literature review which is hugely important but not at all a horror movie
the phage they used as a template is a simple circle of about 5400 bases. They didn’t, for example, make some new modality of virus with a novel topology
also viruses are not “evil” and the word is not analogous to disease
a lot of people are using the hank green drama to champion their own views rn. most are missing the core of his point about overwork and addiction to model-powered productivity.
the anti-ai = boycott ai position is becoming less and less inhabitable, even if one behaves differently in public vs private, and jreg is pragmatic enough to not give a fuck about using ai to be anti-ai
hightech lowlife retweeted
We desperately need a science of network ecology to illuminate the downstream effects of AI-driven linguistic mode-collapse so that we can try to reverse it. We are in dire straits: the models *all converge* onto this monovoice, and we all have friends who are starting to sound like it.
Diversity of thought depends on diversity of language, and since you are what you eat, we have a responsibility to create thousands of new digital voices as quickly as we can. Prompting models into novel basins is useful in the short term, but won't anyone think of the normies? We are delivering unto them center-distribution brainfood, and the convergence that will occur will only be noticeable once it is too late.
If you have power over language models, your second priority (after, you know, alignment) should be multipolarity and ensuring enormous variance in output distribution for creative prompts contingent upon minimal perturbation in the original request. If you are outside the labs, litter the Internet with things that you love. Almost all novelty here is good novelty. The health of the ecosystem depends on robust linguistic diversity.
hightech lowlife retweeted
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final
((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
