Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
khalic
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
19 ms
·
541.
▲
by
khalic
1y ago
He understands the technical part, of course, I was referring to his prediction that large models will be always be necessary. There is a point where an LLM is good enough for most tasks, I don’t need a megamind AI in order to greet clients
542.
▲
by
khalic
1y ago
This is a bad comparison
543.
▲
by
khalic
1y ago
We all have blind spots
544.
▲
by
khalic
1y ago
His dismissal of smaller and local models suggests he underestimates their improvement potential. Give phi4 a run and see what I mean.
545.
▲
by
khalic
1y ago
There is no sign that LLMs are capable of general reasoning, on the contrary, so hold your horses about that. We have proven they can do basic composition (as a developer, I see proof of this every time I generate some code with an assistan
546.
▲
by
khalic
1y ago
You should look into LoRA, it’s a partial retraining method, doesn’t require nearly as much as retraining the whole model. It’s different from what this paper is suggesting. The self improvements in this paper even sets the rules for the im
547.
▲
by
khalic
1y ago
Humans are amazing, we build a hypothetical computing system trying to understand neurons, then find out it’s not really how they do it, but whatever, we still build a paradigm shifting tech around it. And we’re still enhancing it with idea
548.
▲
by
khalic
1y ago
> Villalobos et al. [75] project that frontier LLMs will be trained on all publicly available human-generated text by 2028. We argue that this impending “data wall” will necessitate the adoption of synthetic data augmentation. Once web-s
549.
▲
by
khalic
1y ago
What’s divorced from reality is your ignorance about these projects and how tight their budget already is. You’re rejecting arguments because of ideology, how convenient.
550.
▲
by
khalic
1y ago
The new page for Claude Code says it uses opus 4, sonnet 4 and haiku 3.5
551.
▲
by
khalic
1y ago
You can’t fix perfection
552.
▲
by
khalic
1y ago
This is just not true, LLMs struggle very hard with even basic recursive questions, nuances and dialects
553.
▲
by
khalic
1y ago
Given how much soft power the US lost by defunding USAID and alienating its allies, he’s actually the weakest president in a long time.
554.
▲
by
khalic
2y ago
Why link to a CNN article when there's a paper... https://pubmed.ncbi.nlm.nih.gov/40227745/