cubefox

45 ms·

901.
by cubefox 8mo ago
An Arc B580 should easily be enough to play most games with Lumen and Nanite at reasonable resolution and frame rate, especially if the game also supports XeSS or even XeSS 3.
902.
by cubefox 8mo ago
> we also know that companies will use it to pad benchmarks. Unlike Nvidia, Intel explicitly doesn't use it to pad benchmarks.
903.
by cubefox 8mo ago
So if your algorithm counts apples and counts pears, those wouldn't both have the type "integer". Far from it. They would have the types "number of apples" and "number of pears".
904.
by cubefox 8mo ago
Banning targeted advertising doesn't help against pornography.
905.
by cubefox 8mo ago
Interesting that GPT-5.1 and 5.2 (0 of 10 correct) are a lot worse than the older GPT-5 (7 of 10 correct). But unfortunately the article doesn't mention whether they used the reasoning model or not. Even more interesting: Gemini 2.0 Fl
906.
by cubefox 8mo ago
To avoid your proposed punishment, they will implement things like ... ID verification.
907.
by cubefox 8mo ago
There is a new way to train diffusion models to insert tokens between existing tokens rather than unmasking <mask> tokens: https://openreview.net/forum?id=VbvXjs5f72 However, I believe this would "only" be a
908.
by cubefox 8mo ago
Another picture: https://www.reddit.com/media?url=https%3A%2F%2Fi.redd.it%2Fh...
909.
by cubefox 9mo ago
This doesn't mention the drawback of diffusion language models, the main reason why nobody is using them: they have significantly lower performance on benchmarks than autoregressive models at similar size.
910.
by cubefox 9mo ago
The early version of Gemini 2.5 did initially show the actual CoT in AI Studio, and it was pretty interesting in some cases.
911.
by cubefox 9mo ago
Yeah, E Ink Gallery is the only real exception to the rule with full color support. (The store signage can also change colors, but they don't support color mixing, so there are just three or four colors in total.) Unfortunately, even a
912.
by cubefox 9mo ago
> Isn't that how color images printed paper works, too? No. When you print a piece of paper some color, e.g. red, it will be completely red. But most e-paper screens will only be 33% red (optimistically) and 66% black. This is becau
913.
by cubefox 9mo ago
Gemma≠Gemini
914.
by cubefox 9mo ago
> Isn’t e-paper the general category of low power displays? Yes, or more precisely: reflective displays without backlight. There were many such display technologies a while ago (when the Kindle took off and various companies tried to com
915.
by cubefox 9mo ago
So Veo (Google), Sora (OpenAI) or Seedance (ByteDance)?
916.
Why I'm Worried About Job Loss and Thoughts on Comparative Advantage(lesswrong.com)
85 points by cubefox 9mo ago|124 comments
917.
by cubefox 9mo ago
I'm confused, that doesn't make sense to me: > They largely come from hyperscalers who want hard drives for their AI data centers, for example to store training data on them. What type of training data? LLMs need relatively lit
918.
by cubefox 9mo ago
> According to Mosley, Seagate is not expanding its production capacities for now. Growth is to come only from higher-capacity hard drives, not from additional unit numbers.
919.
by cubefox 9mo ago
I tested it, I think it's super inaccurate.
920.
by cubefox 9mo ago
I just tested originality.ai and it claimed 100% probability that the editors note on the Ars retraction [1] was itself AI generated. For the Gemini article on Benji Edwards it was "only" 56%. I think your tools need a lot more ev
921.
by cubefox 9mo ago
Again, their SpaceX reporting might be "positive", but that doesn't mean it has positive bias. Apart from that, I previously said the non-SpaceX reporters are strongly biased. Anything about the SpaceX reporting isn't re
922.
by cubefox 9mo ago
No, I didn't say that at all. I said they are biased, that "they make things much more negative (= less positive) than they are", which is a very different thing. Basically, I'm saying they report x-10 rather than x (tha
923.
by cubefox 9mo ago
> immediately flagged by AI text checkers as LLM-generate Proof? Which one? I would like to test a few other articles with your checker to test its accuracy.
924.
by cubefox 9mo ago
I have said that they have a strong negative bias. Whether the underlying news is positive or negative is completely irrelevant. Relevant is that they make things much more negative (= less positive) than they are.
925.
by cubefox 9mo ago
Would that excuse being extremely negative about anything that is much less than extremely negative?
926.
by cubefox 9mo ago
Apart from articles by the two space reporters, any news about Musk tend to be biased towards being extremely negative.
927.
by cubefox 9mo ago
> his writing is terrible and extremely AI-enthusiastic I disagree, his writings are generally quite good. For example, in a recent article [1] on a hostile Gemini distillation attempt, he gives a significant amount of background, includ
928.
by cubefox 9mo ago
Only Windows 11 had rounded windows though.
929.
by cubefox 9mo ago
It seems unlikely "model" was ever equivalent in meaning to "architecture". Otherwise there would be just one "CNN model" or just one "transformer model" insofar there is a single architecture involve
930.
by cubefox 9mo ago
I'm pretty sure only the second is properly called a model, and "BERT models" are simply models with the BERT architecture.

More ›