dartos

38 ms·

751.
by dartos 2y ago
> people have different definitions for what "open source" means. They shouldn’t. It’s just market confusion. There is an explicit widely accepted definition. Also like llama (the file you download from huggingface) isn’t even
752.
by dartos 2y ago
Hate to be one of those, but it’s not open source. It’s just regular old freeware. You can’t built llama yourself and it’s license contains a (admittedly generous) commercial usage restriction.
753.
by dartos 2y ago
The devil you know
754.
by dartos 2y ago
Sort of? Then there’s trust that it won’t make up information. It probably won’t be used for any HR/legal work for fear of false info being generated.
755.
by dartos 2y ago
YMMV wrt your experience and luck. I’m a pretty experienced developer and I struggle to get any useful information out of LLMs for any non-trivial task. At my job (at an LLM-based search company) our CTO uses it on occasion (I can tell by t
756.
by dartos 2y ago
They’d probably just focus on laser printers. Higher buy in anyway.
757.
by dartos 2y ago
I may be missing something, but I thought that each context token would result in an 3 additional parameters per context token for self attention to build its map, since each attention must calculate a value considering all existing context
758.
by dartos 2y ago
> Good knowledge of facts I haven’t seen this demonstrated in gpt-4 or Claude sonnet when asking anything beyond the most extreme basics. I consistently get subtly wrong answers and whenever I ask “oh okay, so it works like this” I alway
759.
by dartos 2y ago
No thanks, I’d like it to be accurate ;) Regular ol tests would do
760.
by dartos 2y ago
I used to agree, but as an experienced engineer asking about rust and y-crdt, it sent me down so many wrong rabbit holes with half valid information. I used Claude recently to refresh my knowledge on the browser history API and it said that
761.
by dartos 2y ago
Incremental improvements I don’t think counts as the innovation you were hinting at. Especially not when both phone manufacturers and car manufacturers are engaging in more and more rent seeking behavior
762.
by dartos 2y ago
All of what you described can change wildly from model to model. Even across different versions of the same model. Maybe a database could be built with “tells” organized by model.
763.
by dartos 2y ago
During inference time, yes, but training time does scale exponentially as backpropagation still has to happen. You can’t use fancy flash attention tricks either.
764.
by dartos 2y ago
Let’s see how they do with Miami drivers.
765.
by dartos 2y ago
> isn't this surmounting the problem? I’m sorry, I’m not really sure what you mean by this. There are techniques to lower the rate of hallucinations, but not solve it. Those techniques also vary wildly from application to applicatio
766.
by dartos 2y ago
No it doesn’t. You’re not unconscious while you’re sleeping.
767.
by dartos 2y ago
> We really are seeing massive amounts of innovation in the economy that improve people's lives Name some? The issues people care most about atm seem to be getting worse (housing prices, cost of living, etc)
768.
by dartos 2y ago
> centralized model is far superior to an arbitrary and unknown number of parties receiving metadata about messaging. One opaque MITM is worse than 1000 MITM whose inner workings are bound by a secure protocol. That’s how the internet wo
769.
by dartos 2y ago
Nothing in my comment argues against the existence of bees either. I have some opinions on the matter, but arguments about unfalsifiable traits aren’t really useful, especially since you’re not going to be convinced, nor I. I don’t personal
770.
by dartos 2y ago
> Genie 2 can generate consistent worlds for up to a minute, with the majority of examples shown lasting 10-20s.
771.
by dartos 2y ago
Humans are complex.
772.
by dartos 2y ago
> I really don't understand why people think it's some insurmountable wall. People do hallucinate in the same way llm-s do. Go talk to some real people. A person making a mistake is not similar to a statistical model’s predicti
773.
by dartos 2y ago
> Best case: roguelike adventure. I know you didn’t mean it like this, but this is kind of an insult to the insane amounts of work that go into crafting just the RNG systems behind roguelikes.
774.
by dartos 2y ago
“Instructions” as in directions for accomplishing a task and “instructions” as the binary signals we send to a CPU to process are not the same. CPU instructions may very well not be intentional. Do you know every instruction that your compi
775.
by dartos 2y ago
Obviously a Chinese threat actor at work.
776.
by dartos 2y ago
I don’t think shrimp Jesus was initially intentionally bad. They were farming interactions on fb pages and… well… people on fb seem to like AI images of Jesus.
777.
by dartos 2y ago
So… very boring? Consistent inconsistency gets old very very fast.
778.
by dartos 2y ago
A 9 hour old account too
779.
by dartos 2y ago
Are books simply ink on paper? Are stories just ordered words?
780.
by dartos 2y ago
Can you? I’ve never heard that from a C++ dev. I don’t think any C++ dev worth talking about thinks that C++ is particularly safe, just that (modern) C++ has plenty of powerful safety features. (Like smart pointers. Analogous to rust’s Rc a

More ›