fooker

10 ms·

241.
by fooker 5mo ago
Google search and Wikipedia both started out being fairly reliable to their source of truth. Google pretty much guaranteed that their top results were relevant to the search query. And wikipedia had an army of people making sure everything
242.
by fooker 5mo ago
Imagine the dystopian future where your freedom depends on convincing a panel of AI judges that you are innocent. I assume you'd have access to AI lawyers too, better ones if you can pay for larger/newer models! Meanwhile the judg
243.
by fooker 5mo ago
Search and fact checking are different problems though. LLMs are pretty decent at 'search' given the inherent knowledge compression, and some amount of inaccuracy is fine.
244.
by fooker 5mo ago
Just because it is important for the use case does not mean we can make it work. It's a pretty well known fundamental limitation of the technology. No amount of elbow grease will get it there. There's an interesting tradeoff here,
245.
by fooker 5mo ago
It's a marketing failure (or success, depending on how you see it). AI is pretty useful for a great many things, but to really attract more and more investment the current technique seems to be convincing people that AI is useful for e
246.
by fooker 5mo ago
I don't get why everyone is hellbent on getting LLMs to perform fact checking. This is not the technology for it. Sure it might sorta kinda work in some circumstances. That doesn't make it a good fit. Think of it like buying a ref
247.
by fooker 5mo ago
> Instead of 0-60 in 3 seconds, ... The theoretical limit is 1.7s, which is already basically achieved by multiple gas and electric cars.
248.
by fooker 5mo ago
Fortran compilers will happily tile your loops. C compilers don't really do this.
249.
by fooker 5mo ago
You missed the point. Those two representations are equivalent, yes. But that's not what flattening loops mean.
250.
by fooker 5mo ago
> So you would be equally critical of overloading [] for maps? No, I'm not sure how you got that impression. Overloading is great. It's also confusing when it does something completely different from what you intuitively expect
251.
by fooker 5mo ago
And just in case you have not come across this, C++ allows you you overload all the relevant operators here: [], *, -> So, you really can't tell what's going on behind the scenes. I wanted to pull my hair out seeing some &#x2
252.
by fooker 5mo ago
> The compiler may also flatten a loop. http://c2.com/cgi/wiki?SufficientlySmartCompiler In practice, C compilers are still notoriously bad at loop optimizations. Polyhedral optimizations provided some hope, but no
253.
by fooker 5mo ago
C array types are weird because C doesn't really need arrays. It's not what C was about. But if you designed a language in the era where Fortran, THE array language, reigned supreme, nobody would use your language. The mindshare F
254.
by fooker 5mo ago
They'll say whatever is needed for investors to get in line. Computers, 1998. AI data centers, smells like 2027.
255.
by fooker 5mo ago
> what is being discussed are 100 MW space data centers Thats... not how engineering anything works. 20kw gives you a reasonable amount of compute, about a GB200 rack. That's pretty good for serving inference workloads. Starship is
256.
Is It Impossible to Cool a Datacenter in Space? (Scott Manley) [video](youtube.com)
4 points by fooker 5mo ago|9 comments
257.
by fooker 5mo ago
TLDW - Starlink V3 design should be able to do 20kw of computers. It's a complex engineering challenge to scale it up.
258.
by fooker 5mo ago
> new and fascinating form of consciousness > cold, calculating technology devoid of any empathy or care I don't see why these statements are contradictory. AI seems to be both of these in my opinion. Unless you can only accept t
259.
by fooker 5mo ago
Brave's MITM for ads (supposedly) after client side decryption has killed the brand for me. They seem to have walked back some of this, but it seems pretty unclear what and why, and I'd rather not try to find out given there'
260.
by fooker 5mo ago
I have no qualms about people writing AI assisted stuff, the points being made matter more. Of course it's annoying if a single sentence is blown up into a page of prose by AI and an AI summarizes it into a different sentence on the ot
261.
by fooker 5mo ago
This is a surprisingly nuanced and technically literate take on this topic. Kudos. I wonder if this sort of thing got this dude elected, to navigate the changing times. (duplicating my comment from the other thread as this seems to have mor
262.
by fooker 5mo ago
This is a surprisingly nuanced and technically literate take on this topic. Kudos. I wonder if this sort of thing got this dude elected, to navigate the changing times.
263.
by fooker 5mo ago
AWS was launched around 2006 (2002 internally at Amazon I think). Where could you rent VMs in 2006? IIRC there were two ways to run stuff, get your own server or get an account on a big shared computer.
264.
by fooker 5mo ago
Sure, if you can stop being sanctimonious.
265.
by fooker 5mo ago
Thank you
266.
by fooker 5mo ago
It’s not that far in the future that you’ll start seeing quite a few ‘intelligent Americans’ in European and Chinese PhD programs :)
267.
by fooker 5mo ago
When you say world, do you also include the Middle East and South America and various Asian countries that were invaded, bombed, or couped by America?
268.
by fooker 5mo ago
Whenever I see something this obviously false on a forum, it’s always a head scratcher. Perpetrator of misinformation or victim? Ignorant or malevolent?
269.
by fooker 5mo ago
And all these banks were bailed out by big brother. So the people were right.
270.
by fooker 5mo ago
This approach is basically crying wolf, only works once or twice. After that ‘national security issue’ becomes a running joke. A bit like calling everyone a terrorist, or (very fashionable in the 1960s) communist to achieve your goals.

More ›