sangnoir

37 ms·

751.
by sangnoir 2y ago
> Well, the arguments out there aren’t that LLM’s are too brash, or discourteous or, insensitive. People are saying they’re “dangerous”. I didn't say that... > None of your examples speak to danger. Why should they have support
752.
by sangnoir 2y ago
I wasn't thinking of HAL (which was operating according to its directives). I was extrapolating on how occasional hallucinations during self-training may impact future model behavior, and I think it would be psychotic (in the clinica
753.
by sangnoir 2y ago
Rebuking, shunning and ostracism are key levers for societal self-regulation, and social cohesion. Pick any society, at any point in time, amd you will find people/ideas that were rejected for not confirming enough. There are limits to
754.
by sangnoir 2y ago
> “Do they know things?” The answer to this is yes but they also think they know things that are completely false Thought experiment: should a machine with those structural faults be allowed to bootstrap itself towards greater capabiliti
755.
by sangnoir 2y ago
...but can you game out how one might achieve this in way that the victim won't immediately die, and the organizers are not criminally liable? As a thought experiment, of course.
756.
by sangnoir 2y ago
> Shifting to zero-trust requires a lot of migration work across all of this, "training" ie figuring out how to get stubborn IT people to buy in to the new way of doing things[...] You're right, most times organizations ne
757.
by sangnoir 2y ago
No, no, no - the trends are clear: it was obviously crypto first, then AI next, both brought to you by the same benevolent taste-making thought-leadership organized in elite private chat groups. /s
758.
by sangnoir 2y ago
> For some reason, I would assume that if you have connection, it is faster than 2G these days. That assumption does not hold for a sizable chunk of Meta's 3.98B-strong userbase. The list of counties that switched off 2G is surprisi
759.
by sangnoir 2y ago
Why not both?
760.
by sangnoir 2y ago
> Price tells a story. If AMD performance would be in par with Nvidia they would not sell their cards for 1/4 price What were your thoughts on Zen (1) vs Intel's offerings then? AMD offered more back for the buck then too.
761.
by sangnoir 2y ago
The bar for success for Google and Meta is much lower than Nvidia - at least for internal usage. Any dollar amount that Google saves on CapEx or OpEx by using custom silicon instead of buying Nvidia helps bring down the cost of revenue. T
762.
by sangnoir 2y ago
If only Meta had a way to monetize engagement with generative AI content in a way that scales with quantity of generated content.
763.
by sangnoir 2y ago
Are you suggesting Meta or Google - who stand to save billions - won't be able to get top performance from their custom chips because their tooling/hardware won't support CUDA?
764.
by sangnoir 2y ago
> Post-M series I hear this from time to time and always ask people to show me something in the same weight class with equivalent battery life, performance, and screen quality Law of diminishing returns. If I only need 8 hours of batter
765.
by sangnoir 2y ago
Who would have guessed that the Great Filter preventing interstellar civilizations is greed?
766.
by sangnoir 2y ago
> Fun fact: turns out I'm really good at mentally adding and multiplying 2 and 3-digit numbers while unconscious. You'll be fully conscious - you just don't retain the memory. Its like driving on a route you've driven
767.
by sangnoir 2y ago
> Ultimately maintaining things isn't free. This is a hypocritical stance for any project that accepts free contributions from the public. Besides rent-seeking, what reason could justify preventing the community from maintaining fea
768.
by sangnoir 2y ago
> They aren't all that scared of a small subset of geeks running Linux directly on their iPad and using that as a regular Linux-ey thing, I don't think. I don't think they are worried about a few geeks running bash or dock
769.
by sangnoir 2y ago
All I need is a Cloud-to-Butt-like browser plugin that replaces all instances of "GenAI" and "generative AI" with "AI slop"
770.
by sangnoir 2y ago
> In my opinion, OpenAI is building a platform business [1]. OpenAI is primarily an innovation platform. It provides tools, models, and APIs that enable developers and businesses to create new applications and services, fostering a wide
771.
by sangnoir 2y ago
> Any systems level team will prefer this form of tightly integrated solution to the systems layer problem if they are responsible for the highly built out specialized distributed application we call netflix If any team would prefer th
772.
by sangnoir 2y ago
Setting aside stock appreciation, are your refreshers greater than new hire FAANG RSUs for a similar position? Most FAANG total compensation dips in year 5 due to 4-year vesting schedules, and refreshers for existing employees being lower t
773.
by sangnoir 2y ago
> Altman isn't an Effective Altruist Or any type of small-a altruist, really. Which is in direct conflict of the image he wishes/wished to portray by "only" officially getting a bay-area-intern-level salary with no eq
774.
by sangnoir 2y ago
"Timelapse video" is the usual term for such videos.
775.
by sangnoir 2y ago
> Yes, they're much more _capable_, but usually even less responsive than e.g. in Windows 95 days Hard disagree: all applications back then needed splash screens to reassure the user something was happening while the application w
776.
by sangnoir 2y ago
People who are consistently unable to distinguish fiction from reality make terribly witnesses; if an obviously high crackhead would fare better than an LLM on the witness stand.
777.
by sangnoir 2y ago
> Google have been trying that thesis out for years and yet TPUs aren't flying off the shelves in the way H100s are Google does not sell TPUs to 3rd parties at all[0]. Or do you mean cloud customers prefer H100s to TPUs - if so, I&#
778.
by sangnoir 2y ago
Apple is only sticky to those who buy into the ecosystem. Those whose a single Apple device (or 2) without paying for any Apple service are not trapped in the walled garden
779.
by sangnoir 2y ago
They hyperscalers can bypass CUDA if it is profitable. Most AI practitioners use Torch rather than CUDA directly, so it's effectively "under the hood". If some director at Meta figures they could reduce Meta's capex by $
780.
by sangnoir 2y ago
> You really think investors like sequoia and a16z are dumb enough to fall for Elon hyping things up? a16z invested $350 in Adam Neumann's real estate venture - after WeWork. VCs will absolutely knowingly invest on hype if they th

More ›