lossolo

35 ms·

601.
by lossolo 2y ago
> The paper i linked is not addressed by the paper you linked. The paper you linked attempts to give LLMs the same benchmarks in a format they aren't best suited for. I don't know how you can call that "addressed". Wh
602.
by lossolo 2y ago
> Oh i've read them. It's hard to believe that after reading all the papers and the blog I linked, along with the references there, any reasonable person would come to such strong conclusions as you did. This makes it hard for
603.
by lossolo 2y ago
You clearly didn't read any of these papers. Quote from one of them "Our results support the hypothesis that GPT-4, perhaps the most capable “general” LLM currenly available, is still not able to robustly form abstractions and rea
604.
by lossolo 2y ago
> You can get them to solve unseen problems just fine Prove that the problem wasn't seen by them in other form. > Specify a grammar in BNF notation and tell it to generate or parse sentences for you. You can produce a more than r
605.
by lossolo 2y ago
No, none of the Millennium Problems or other math problems (unsolved by humans for decades or centuries) have been solved solely by LLMs, even though they possess all the knowledge in the world.
606.
by lossolo 2y ago
> "generalize to its dataset" is a contradiction Not "to" but over, example the same code written in one language over the other language. > if you think being able to generalize in ways similar to the whole of the
607.
by lossolo 2y ago
Generalize over their training data—they cannot generalize out of distribution. If they could, they would have already solved most human problems. So no, they do not generalize on unseen text. They will produce what is most statistically pr
608.
by lossolo 2y ago
> I come from former communist country and you are absolutely right - everybody had to work, both men and women, not having a job was a straight ticket to long and nasty jail. Of course that's artificial, so lower classes had absolu
609.
by lossolo 2y ago
Does anyone else have a problem with it on Mac when trying to open videos using the 'Open With' dialog in Finder? I see it start for a second, and then it just quits. It works properly from the command line
610.
by lossolo 2y ago
> Community Notes seems to be highly effective at combatting misinformation, at least of the most popular tweets. It doesn't, millions of people see it before community notes show up. Musk is sharing fake news himself[1]. This is da
611.
by lossolo 2y ago
Megaupload was not deleting content. They only deleted links to the content. So, if you had 10 links to the same video, and a copyright holder contacted them to delete their content under link #1, they would only remove that link but leave
612.
by lossolo 2y ago
A lot better quality of training data and instruction tuning (data again). There is no other secret sauce.
613.
by lossolo 2y ago
Use this https://livebench.ai It's a better benchmark.
614.
by lossolo 2y ago
Based on english doc it seems it's not dubbing but voice over.
615.
by lossolo 2y ago
> The render nodes used for Interstellar had 156GB of RAM, and a decade later the biggest GPUs still don't have that much memory NVIDIA H200 GPU has 141 GB of memory, and the AMD Instinct MI300X GPU has 192 GB of HBM3 memory.
616.
by lossolo 2y ago
> Software RAID and dmcrypt will burn plenty of cycles, enough to prove as exceptions to the "no FPU instructions in the kernel" guideline. LUKS has a negligible impact on I/O bandwidth, and the same is true for software
617.
by lossolo 2y ago
> Unlimited human life expectancy vs limited resources? How would that work? Just like today but controlling the population would be necessary. Do you want to live forever? Fine, but no children until you die. Since people would still st
618.
by lossolo 2y ago
Not the same, it will get you worse output and is harder to do right in practice.
619.
by lossolo 2y ago
Writing tests won't help you here, this problem is the same as other generation tasks. If the test passes, everything seems okay, right? Consider this: you now have a 50-line function just to display 'hello world'. It outputs
620.
by lossolo 2y ago
There is a thesis in your question. I'm not a terminated Twitter employee. I follow him on X and see exactly what he writes, the amount of disinformation he retweets, his manipulation tactics and what he does. That's enough for me
621.
Japan stocks plunge as much as 7% as Asia shares extend sell-off(cnbc.com)
138 points by lossolo 2y ago|126 comments
622.
by lossolo 2y ago
How could it not be? What if there is a buffer overflow that could grant you GOD mode? It would be so fun for the first few millennia.
623.
The Reason People Aren't Having Kids(theatlantic.com)
10 points by lossolo 2y ago|6 comments
624.
Europe's €10T Gamble(politico.eu)
2 points by lossolo 2y ago|1 comments
625.
by lossolo 2y ago
https://shane.ai/posts/cgo-performance-in-go1.21/
626.
The Autopilot Data That Reveals Why Teslas Crash(wsj.com)
2 points by lossolo 2y ago|0 comments
627.
U.S. Commercial Real Estate Is Headed Toward a Crisis(hbr.org)
38 points by lossolo 2y ago|62 comments
628.
by lossolo 2y ago
> Whatever happened to that new voice interface they "announced" back in May? Yeah, and it was suddenly made just the day before Google announced their models... and do you remember Sora? That was in February.
629.
by lossolo 2y ago
For some time, we have been at a plateau because everyone has caught up, which essentially means that everyone now has good training datasets and uses similar tweaks to the architecture. It seems that, besides new modalities, transformers m
630.
by lossolo 2y ago
Indeed, this quote hits hard: "The past five years have felt like navigating the valley of death. Franklin and I, like many of you in the Asian community, came to the USA to pursue our dream through intelligence and hard work. We wante

More ›