atemerev

46 ms·

541.
by atemerev 2y ago
Asking for the information about non-public individuals, including myself. RAG-assisted GPT-4 easily provides such information. GPT2 output is consistent with a good model without RAG (it tries to speculate, but says it doesn’t have such in
542.
by atemerev 3y ago
No, I tested it, this is a pure model without RAG (at least no internet search)
543.
by atemerev 3y ago
I think someone is flagging this (hi @sama!)
544.
by atemerev 3y ago
Indeed, all my test prompts are giving much better results than gpt4-turbo and Claude Opus. And yet, the OpenAI style is clearly recognizable.
545.
GPT-4.5 or GPT-5 being tested on LMSYS?(rentry.co)
543 points by atemerev 3y ago|360 comments
546.
by atemerev 3y ago
The law is not there yet, the document was fake. But still it is possible. I can’t return, and I won’t be able to get a citizenship of the country I currently live in. Guess I will live as a stateless person.
547.
by atemerev 3y ago
"clever"
548.
by atemerev 3y ago
I wonder what will happen if one starts developing, say, a RTOS for cruise missiles in Rust. Will this be controversial enough to destroy the Rust community?
549.
by atemerev 3y ago
Mediocre jobs are mediocre even when remote work is allowed. And Australia has their time zone problem, which is important if you plan to work for US companies.
550.
by atemerev 3y ago
Aggression is not excusable. I am fond of American values, but unprovoked aggression is not one of them.
551.
by atemerev 3y ago
I am saying this as the "other". I am Russian (though not living in the country anymore). My country has started an unforgivable war of aggression. It is my imperative to do anything I can to remedy this. Some things are quite mor
552.
by atemerev 3y ago
I fully agree. Defense is the moral imperative of free nations.
553.
by atemerev 3y ago
Defense is not morally questionable. Particularly in the presence of adversaries who _are_ clearly morally questionable.
554.
by atemerev 3y ago
If we are interested in AGI safety, we should experiment with slightly unsafe things before they become hugely unsafe, instead of trying to fix known unknowns while ignoring unknown unknowns. We should open source current small models and o
555.
by atemerev 3y ago
1) There is not enough data in the entire world to train a trillion parameters large language model. And there is limited use for synthetic data, yet (we need an internal adversarial/evolutionary model like AlphaZero for that, which I
556.
by atemerev 3y ago
At my workplace, we are now completing the multiscale simulation of a full mouse brain (neuron/synaptic activity, cellular metabolism, etc). It fits into our small supercomputer first deployed in 2019, and there's a large potentia
557.
London Defense Tech Hackathon(twitter.com)
6 points by atemerev 3y ago|1 comments
558.
by atemerev 3y ago
This is the video that produced the same effect on me. Apoptosis (2006): https://www.youtube.com/watch?v=DR80Huxp4y8 We are all made of nanobots.
559.
by atemerev 3y ago
Actor model with preemption solves the entire thing like it was never a problem. Three axioms, one mechanism, puff, gone. But for some strange reason, bona fide preemptive actors exist only in Erlang and derived languages (Elixir, Gleam, et
560.
by atemerev 3y ago
I prefer plain actor model, it is so simple that it happily abstracts the entire problem away. CSPs require a lot of reasoning beforehand.
561.
by atemerev 3y ago
Vector displays are fascinating. I am trying to emulate the UI occasionally.
562.
by atemerev 3y ago
Still, email and usenet sort of worked. I have sent my first international email in 1993, from a school lab in my home town in Siberia. In larger cities, I bet there were ways to do it even earlier.
563.
by atemerev 3y ago
And that is how people invented blockchain (it is slightly more complicated, particularly that each transfer should include the source address in the output, but the idea is the same). As mentioned elsewhere in the thread, we also need mult
564.
by atemerev 3y ago
APFS is really sensitive to this situation, more than other file systems. I don’t remember how I managed to resolve this situation, but it involved booting into the single user mode and some magic not accessible to a regular Mac user.
565.
by atemerev 3y ago
Reasoned decisions (if you think that empire building is reasonable) without morality and empathy _are_ evil. This is how Putin operates. Also, raping and torturing are very "evil for the sake of evil", if you ask me.
566.
by atemerev 3y ago
Nvidia now works more or less fine with Wayland.
567.
by atemerev 3y ago
Because it uses some American tech or patents, and there is leverage and pressure point over these.
568.
by atemerev 3y ago
It is really counterproductive policy to keep highly qualified immigrant workers from coming into the country. And sadly common everywhere.
569.
by atemerev 3y ago
Exactly, I don't know why you are being downvoted. There is data that has better to be secret, even though it is funded by public money. No controversy in that.
570.
by atemerev 3y ago
But that’s good, isn’t it? Finally I could afford to buy something instead of renting.

More ›