wwarner

Thing is, it depends on whether llms + reinforcement can self-improve in principle. Learned recently that cognitive scientists, before the transformer & llms, were studying the possibility that thinking and learning might be based on some kind of prediction, i.e. something similar to token prediction, and I quite suddenly became less skeptical about the possibilities of llms. (Some will say I’m late to the party of course.) But if knowledge to date has been accumulated in a process quite like “chain of thought” in llms, then I don’t see any reason that computers won’t self-improve in the near future.

so well written

I long for the days when voting for outsiders was a fun way to fill out a ballot, back when election outcomes didn’t matter very much because the probable winners were all qualified to hold office.

Great discussion with a cognitive scientist on what LLMs might reveal about the role of prediction in human cognition.

It means that there is a spectrum of outcomes, and there is a band left of center which is anti-crime enough to support cameras with laws to govern them.

Uh, I guess you haven't compared my town's crime rate to London's. I'm for curtailing people's gun rights as well, for essentially the same reason as I'm for more cameras. There was an apparently random shooting at an intersection not far from me, and the shooter escaped in a car. The police pieced together some footage from the two private cameras that showed the car first getting some gas, and then parking before the shooting. They found the driver and he confessed to the shooting.

Unpopular opinion: I’m fine with cameras if they can be shown to solve crimes and lower the crime rate. No, I’m not ok with ICE surveillance, or secret police punishing free thinkers, but the extreme anti position simply cedes the technology to the authoritarians. So basically, if you’re promising London safety and the freedom and safeguards of the UK, I’m all for it!

Maybe, or maybe it was eaten into extinction.

warning: this shows a living mouse in a mouse trap.

Honestly i come to hn to escape true crime.

Expected a generator.

I've only ever been good at dancing to punk music, which has only cost me money so far.

Yes, and wondering why all the AI tooling is written in node.

Wouldn’t apache arrow serve the same purpose?

Haven't finished this but for me it's so refreshing to read some science on deep learning and not just weird predictions.

I feel the same, but then I remind myself of how I used to feel about dark matter (I really disliked it). Having an arena where no scientific question is out of bounds is great.

I run emacs without graphics from this docker image i maintain https://hub.docker.com/r/wwarner/emacs-native

I just use emacs for programming, not note taking or email reading or the other things people love it for. Strengths: magit, wgrep, vterm, buffer management, plugins & packaging. Weaknesses, irritations, embarrassments: yaml-mode, responsiveness, overlays & mini-buffers.

My usage of emacs is so vim-like that I’ve tried switching a few times. Vim is definitely faster, and overlays and cursor placement is much simpler and more intuitive. But there were still feature gaps and configuration issues that prevented full adoption.

Sane & interesting enough to have been disproven, by Boaz Barak iirc. Maybe not surprising since simulated annealing never achieved the results of gradient descent + backprop.

These drawbacks are all true, but sometimes storing directly to elastic is still the best way forward.

I feel like these are helpful, and I think the calculus oriented visualizations of convex surfaces and gradient descent help a lot as well.

I reluctantly agree. It’s like ebikes — yes it’s great that I don’t have to pedal up hill, but on the other hand the cyclists that did it the hard way deserved the praise and glory for their achievement while weak and distracted ebikers definitely do not.

The version manager approach feels like a giant step backward to me. Manage dependencies with containers.

holding my nose reading this. if scientific progress killed god, it seems unlikely that meaning would emerge from more ramblings of the same kind that gave rise to him in the first place. we have learned to disbelieve in miracles and to be skeptical of novelty, that change is excruciatingly slow and its cause is failure, pain and death. nature and the feelings that nature has given us should be our philosophical guide posts.

i run a basic emacs configuration within docker, so it has all the underlying executables & binaries installed where emacs looks for them. runs exactly the same on linux & macos. https://hub.docker.com/r/wwarner/emacs-native or https://github.com/wwarner/emacs-native-dockerfiles

M-x package-list-packages :)