cgyvbunji

Sorry for sharing facts.

Coal and nuclear cooling towers use a lot of fresh water, they try to reuse it but still between 2 to 3.5 billion gallons a year evaporates from each plant. Japanese nuclear plants don't have this problem for example because they are built on the coast and consume water directly from the ocean instead of using fresh water cooling towers.

When all you have is a hammer...

Export and import don't have to do with files? Glad they started with that one so I knew I didn't need to read the rest of it.

One sec let me ask an LLM..

Ok.

Maybe they took dried fruit

42

Not the first new F91W-ish model in 40 years but still neat, I hadn't heard about this one yet, thanks for sharing.

This is wrong mathematically and I can prove it. Bob has a gas car. Bob is bed ridden and never drives anywhere. 0 miles a year = 0 emissions. So there exists a threshold of miles driven per year below which it's greener to keep the gas car. QED.

3) the person who promoted them likes them.

I think that's projection.

Haha ok ^_^

It's so bad I don't pair it at all and crank the volume all the way up and hope for the best.

I don't care that it's Apple?

See how defensive people get about CarPlay? Why so emotionally invested? It's buggy and sucks when you are driving someone else's car or a rental. You proved my point too, which is that it's liked because it's better than everything else, not because it's objectively good.

I don't understand why anyone likes CarPlay. It has always seemed like people like it because it's better than nothing. It's just not a good system in an objective sense.

> I have no idea how they are pulling that off

Google has the most AI compute by far. I don't have the link handy or I would share it.

> The more application specific you get, the smaller the total volume of chips

Sure, and that is the definition of niche. FPGAs are niche.

You could have made a good point that an application specific design that already requires an FPGA might want to now also have a local LLM, so putting the LLM right on the FPGA might be the most expedient option in that case.

Other than that, I don't think people are reaching for FPGAs to do LLM training or inference in general because I don't think it can be cost effective vs other options.

It's a strange opinion to me, that because RCTs are too expensive, they are not done enough, so therefore we should rely on RCTs less. Can we somehow make RCTs cheaper and do more of them? The signal-to-noise ratio is so unbelievably low in medicine that anything that improves on that is indispensable.

They pull a lot of power.

Continue to exist in what form? Seamonkey aka Netscape Navigator still exists. Firefox can keep existing but not get updated enough to be a viable alternative to Chromium anymore.

How is firefox not essential to the open web?

It's the ASCII diagrams for me. Although I hesitate to point it out because they are usually helpful unlike the em dashes. Before LLMs, ASCII diagrams in readmes were a half baked mess, if one were present at all. Now every project has at least one perfectly made ASCII diagram - instant LLM tell.

Have you guys not figured out yet that the media justs runs random headlines with whatever they got paid to plug with a generation tacked onto it to sell more stuff by creating FOMO

"<Generation we want to sell more stuff to> has rediscovered the joy of <thing we want to sell more of>"

It's possible many such articles are actually truly random and they just add generations to them because it increases engagement, which would be somehow even more sad.

Either way they are definitely not legitimate and you can safely ignore almost any headline with a generation in it.

Wait, so anything that's not sold everywhere is personalized? Don't think so.

Yeah but then it's basically an AI ASIC with an FPGA block inside it. Basically the less FPGA-like an FPGA is, ie the more dedicated silicon in the FPGA for the task in question, the more power efficient it is, because custom logic in an FPGA is done in LUTs which is RAM and RAM is way way more power hungry than actual logic gates, and the fabric is apparently power hungry too. It's unfortunate to me because I like FPGAs and wish they weren't so niche, but they are inherently limited in this way.

FPGAs are not power efficient at all vs GPUs and ASICs anyway, which is going to be especially true when they are fully saturated by LLM inference.

False, hybrids reduce fuel consumption and therefore emissions substantially, Toyota Sienna went from 22 mpg combined to 35 mpg combined when they switched it from gas to hybrid. Several wild claims from you today.

> mechanically you inherit all the problems from both sides

Not really. The transmission is an e-CVT in hybrids which has no torque converter, the main part that breaks first in an automatic transmission. Unlike a regular CVT, the e-CVT has no belt, either. The e-CVT basically never breaks, so you basically never have a transmission go bad in a hybrid. On top of that, because the traction battery is so small, it's replaceable for $1-2k instead of $10k+ for a BEV. If anything, hybrids are the best of both worlds. Had you specified plugin hybrids you would have been correct.