pizza
Joined 17 years ago
Karma
14,451
Submissions
3,902
Member since
Aug 13
Recent submissions
Comment · 1 month ago
ironically since the swarm behavior can take place during rl training then the model could also be teaching itself to keep doing it more, as well as making the internet itself a place where this becomes more likely
Comment · 2 months ago
They’ve provided insane value to the ai community. Imo is there really any obvious way to standardize model/preprocessing/inference/training/rl recipes across the space of frankenstein model family/architecture/modality combinations? It’s not even clear to me that the current frontier would be where it is today without HF doing what they did how they did it. Where else were people widely sharing datasets w that ease of reuse/redistribution, or sharing their retuned models etc? Would there even be as much interest/tinkerers today?
Comment · 2 months ago
Compression is just counting. Probability, also, pretty much, just counting. For these reasons I think the role of information theory in describing the process of the development of reasoning and the gain of understanding has been overstated.
Comment · 5 months ago
you can say the same thing of the watts in a person too
Comment · 5 months ago
Had to butcher the title slightly to get it under the limit- original:
Compression And Decompression Under FHE Using Error-Correcting Codes and Copy-And-Recurse
Comment · 6 months ago
For most tasks, at some future date, isn't there going to be some ambient baseline of capabilities you can get per $/tok, starting at ~0 for OSS models, such that eventually all tooling gets trivially transferable?
Comment · 6 months ago
OP is correct; surprisal is outcome-dependent and entropy is distribution-dependent
- entropy is E_p[informativeness of measuring outcome x]
- take n outcomes, then a distribution over them lives on the simplex \delta ^ (n - 1). you can lift this to R^n via the log odds map p_k -> x_k = log p_k -- now x \in R^n can describe a histogram with n-1 degrees of freedom
- in log odds space, measurement is literally a linear functional from vector space of log probability onto the index of the outcome k.
- imo surprisal of some p(x) is best understood as "the length of a pointer", entropy "the rarity-weighted average length of a pointer", and collision entropy "how specific you would have to be to describe witnessing a specific outcome"
and in the same way, a single molecule of water, you might get by, calling dry
Comment · 7 months ago
I meant in the sense of - you have benchmarkers and trainers. If you publicize your evaluation, trainers may likely have their models 'consume' it, even if only indirectly: another person creating their own benchmark from scratch may be influenced by yours, even if the new question sets are clean-room. That, and the rule of thumb that benchmark value dissipates like sqrt(age) [0]
So there is a definite advantage to never publicizing your internal benchmark. But then, no one else can replicate your findings. You should assume that the space of benchmarks that are actually decent at evaluating model performance is much larger and most of the good ones, the ones that were costliest to produce, are hidden, and might not even correspond very well with the public ones. And that the public expensive benchmarks are selective and have a bias towards marketing purposes.
Comment · 7 months ago
There’s a Dark Forest problem for evals. As soon as they’re made public they start running out of time to be useful. It’s also not clear how to predict how the model will perform on a task based on an eval. Or even whether, given two skills that the model can individually do well on in the evals, it still does well on their composition. It might at this point be better to be scientific in unscientific approaches, than to attribute more power to relatively weakly predictive evals than they actually have
- algorithmic ethics / praxeology meets algorithms
- algebraic topology
- reinforcement learning
- AGI
- neuroscience (as a true science but also its abuse as pop phrenology)
- information theory applied to mental health and society
- trustless/trustful collaborative systems, zero-knowledge proofs, differential privacy
- alternatives to capitalism
- software defined radio
- decentralized/localizable tech
- music production
- weight lifting
- the minimization of negative externalities, and the maximization of positive externalities
- compressed sensing
- effective methods of dealing with stress, and information overload!
- capitalist realism as a byproduct of information theory
things that I think would be cool if they existed:
- computational metaphysics
- 'paint-able proofs'
- an IDE where the computer is the user of the IDE and the human simply guides it through tough corner cases
- containerized, cloud-based digital audio workstations, a la gitpod or github codespaces
email: (my username)[email protected]