make3

35 ms·

451.
by make3 3y ago
that's why the huggingface llm arena exists
452.
by make3 3y ago
it's trained with a lora adaptor, so it's either an error or they also count the adaptor. they use a 16 param inner lora dimension however, so it's unlikely that that's the reason (too small) an important point to keep i
453.
by make3 3y ago
66k USD is very achievable for ppl in the western world
454.
by make3 3y ago
maybe a naive question, but do international credit cards not just work? if you aren't getting paid locally, it shouldn't matter too much?
455.
by make3 3y ago
this is truely evil :)
456.
by make3 3y ago
I hate that this will likely be decided by a 75 years old judge that hasn't been close to a computer since getting his 15 years old grandson to fix his patience game
457.
by make3 3y ago
"Why can't AI at least cite its source" each article seen alters the weights a tiny, non-human understandable amount. it doesn't have a source, unless you think of the whole humongous corpus that it is trained on
458.
by make3 3y ago
that's definitely not what I meant or implied; it makes sense that the immune system is a classifier of sorts
459.
by make3 3y ago
huh that sounds exactly like machine learning overfitting vs generalization
460.
by make3 3y ago
I would assume that you would get a big performance drop, because knowledge is split in a very non human understandable way between the experts
461.
by make3 3y ago
>>> The problem is that "normal" people are much more "morally flexible" and thus share an implicit understanding of what rules are important (i.e. actual rules) and which ones you're supposed to say you fo
462.
by make3 3y ago
That wouldn't work, you could never trust him at all, as he's expressed zero remorse and actually intent to continue to do crime, and is apparently very strongly autistic & doesn't understand the complex nuances of what i
463.
by make3 3y ago
OP's code is much easier to understand, though, which is the main (only) purpose of their code
464.
by make3 3y ago
the numbers are not at all comparable, because Gemini uses 34 shot and variable shot vs 5 for gpt 4. this is very deceptive of them.
465.
by make3 3y ago
why not? Pepsi is basically a marketing company as they're not exactly developing new space faring vehicles, or new flavors, even
466.
by make3 3y ago
profit is not the only morality
467.
by make3 3y ago
? it's a normal capitalist corporation, what exactly did you expect
468.
by make3 3y ago
really ? "gpt4 beats our tiny model", that's your paper?
469.
by make3 3y ago
stable diffusion video is not that good
470.
by make3 3y ago
isn't it a low-fi slideshow presentation jam? how is that better
471.
by make3 3y ago
I agree with Cillian Murphy for Altman, they both have the deep blue eyes
472.
by make3 3y ago
There will also be a Hollywood movie, for sure. My friend suggested Michael Cera as both Ilya and Altman
473.
by make3 3y ago
how did Xerox invent so many things, that's crazy
474.
by make3 3y ago
there like literally infinite amounts of new culture
475.
by make3 3y ago
I agree. this theory sounds ludicrous to me except for people who maybe have absolutely zero technical know-how
476.
by make3 3y ago
tool use is already that.
477.
by make3 3y ago
quaternions? for rotations
478.
by make3 3y ago
true. it does make sense however, you can't expect the ai model to reside in the wearable device
479.
by make3 3y ago
People have proved that they're willing to go far to make tech lighter. Wearing glasses when you can just wear nothing seems like a big progress to me. The only thing that remains is to make it look better imho
480.
by make3 3y ago
You can even make the joke that you're guiding the LionRover from Space, their virtual desktop solution

More ›