But how much time does that 0.3 watt hour query take to run? They imply that an individual ChatGPT query takes 0.3-3 watt hours, but most queries come back in seconds, so we need to scale that over a whole hour of processing.
Edit: Scrolling down: "one second of H100-time per query, 1500 watts per H100, and a 70% factor for power utilization gets us 1050 watt-seconds of energy", which is how they get down to 0.3 = 1050/60/60.
OK, so if they run if for a full hour it's 1050*60*60 = 3.8 MW? That can't be right.
Edit Edit: Wait, no, it's just 1050 Watt Hours, right (though let's be honest, the 70% power utilization is a bit goofy - the power is still used)? So it's 3x the power to solve the same question?
Amphetamine is actually a very effective weight loss drug. And it's sort of orthogonal to the fact that it's a stimulant. Stimulants in general can cause an acute reduction in appetite and temporary weight loss. This tends to stabilise with tolerance, however. As someone with obesity and ADHD, thus was my experience with methylphenidate treatment. And I used to think the weight loss effects of amphetamine were analogous until recently.
Amphetamine and methyphenidate(MPH) have very different ways of acting as stimulants. MPH is an inhibitor of the dopamine transporter(DAT) and the norepinehrine transporter(NET). These cross-membrane proteins essentially "suck up" the dopamine or norepinehrine after neurotransmission, thus regulating the effect. MPH inhibits this process, increasing the effect. This is called a norepinephrine/dopamine reuptake inhibitor(NDRI). Cocaine also works like this, as well as the antidepressant wellbutrin(bupropion).
Amphetamine on the other hand, is a bit more complicated. It interacts with DAT/NET as well, as a substrate, actually passing through them into the neuron. Inside the neuron, it has a complex series of interactions with TAAR1, VMAT2, and ion concentrations, causing signaling cascades that lead to DAT reversal. Essentially, enzymes are activated that modify DAT in such a way that it pumps dopamine out of the neuron instead of sucking it up. How that happens is very complicated and beyond the scope of this comment, but amphetamine's activity at TAAR1 is an important contributor. As such, amphetamine is a norepinephrine-dopamine releasing agent(NDRA). Methamphetamine, MDMA, and cathinone(from khat) also work like this.
Anyway, recently I was reading about TAAR1 and learned something new, namely that TAAR1, besides being and internal receptor in monoaminergic neurons, is also expressed in the pancreas, the duodenum, the stomach, and intestines and in these tissues, TAAR1 activation will increase release of GLP-1, PYY, and insulin, as well as slow gastric emptying.
So in essence, there may be some pharmacological overlap between ozempic and amphetamine(I'm still looking for data on how significantly amphetamine reaches TAAR1 in these tissues, so unclear what the relevance is. But amohetamine is known to diffuse across cellular membranes, so it's likely there is an effect).
Also interesting, amphetamine was recently approved as a treatment for binge eating disorder. Not only because it causes weight loss, but because it improves functioning in the prefrontal cortex(crucial to its efficacy in ADHD), which is apparently implicated in the neuropsychological aspects of BED as well.
There is a mixed picture on this. I see a lot reports of reports of it causing binging in the evenings despite no prior issues.
The issue is that therapeutic doses are not the multi-day bender of a speed-freak that forgoes sleep to keep their blood-concentration permanently high. Instead it's a medicated window of 6-12 hours with a third or more of their waking hours remaining for rebound effects to unleash stimulation-seeking demons that run wilder than ever.
Stockfish being so strong is not merely a result of scaling of computation with search and learning. Basic alpha-beta search doesn't really scale all that well with compute. The number of nodes visited grows exponentionally with the number of plies you look ahead. Additionally alpha-beta search is not embarassingly parallel. The reason Stockfish is so strong is that it includes pretty much every heuristic improvement to alpha-beta that's been thought of in the history of computer chess, somehow combining all of them while avoiding bugs and performance regressions. Many of these heuristics are based on chess knowledge. As well as a lot of very clever optimisation of data structures(transposition tables, bitboards) to facilitate parallel search and shave off every bit of overhead.
Stockfish is a culmination of a lot of computer science research, chess knowledge and clever, meticulous design.
While what you mention is true, I'm not sure how it undermines the bitter lesson. Optimizing the use of hardware (which is what NNUE essentially does) is one way of "increasing compute." Also, NNUE was not a chess specific technique, it was originally developed for Shogi.
Not sure why this is downvoted, it's just factually true.
Police and politicians talking about outlawing things that help criminals as though it will somehow affect the criminals, will never cease to amaze and amuse me. It's such an elementary error of logic.
The fact is that in a reasonably free society it's quite feasible to get away with lots of crime, if you're smart enough. There is no stopping this. Especially if it's a crime which doesn't leave a whole lot behind in terms of physical evidence. Downloading an OS is one such thing. Sure, if you seize my phone, you could prove it runs Graphene. But in a free society, you need probable cause for that, sorry. And if I am some major criminal, and Graphene stops my criminal enterprise from being proven, in a free society that's always preferable to getting busted, because the punishment for using graphene is gonna be meaningless compared to the punishment I'm avoiding by using it. Because a free society includes a protection against disproportionate punishments for minor crimes. Sure I'll pay your $500 fine to avoid 20 years in prison. Cost of doing business.
Once you realise this, you realise the only way to tackle crime is by legalising as many of them as possible, as long as they're not actively and unambiguously violating people's rights. Murder and other violent acts, obviously stay illegal. Drugs, prostitution etc? Legalise them. That's most of the crime right there, because these classes of crime actually provide something that's in wide popular demand. Demand + black market pricing + lack of taxes means lots of money, and money means power to create strong criminal organisations that can do whatever they want with impunity, including influencing politics. With all that out the window, all you have left is a bunch of individuals going at it alone; murdering psychopaths, desperate poor people, the mentally ill, crimes of passion, sex crimes, etc. And you just freed up a ton of societal resources to channel into those vestiges, both via targeted, intelligent policing and broader societal reforms that target the sociological processes that cause these kinds of crime(like wealth inequality, to name one).
Instead, what we get is a never ending arms race towards a totalitarian society. Oh well, see you after the next revolution, I guess.
Legalizing drugs had a pretty bad outcome for Portland, which is why they re-criminalized some drugs.
Prostitution leads to trafficking, a word I absolutely despise, particularly when it is used as a past-tense verb: "trafficked." Ugh! What poorly educated government hacks do to our language should be criminal! Regardless, human trafficking is terrible and if that part could be fixed, then maybe prostitution wouldn't be so horrible, but it is because it is pretty much never a voluntary situation for the women, but always some kind of coercion.
I spent some time* working on the firmware side of developing custom electronics based on various AVR chips, ATmega328 among them. Arduinos are not good for much more than babby's first microcontroller project. They're not even that great for prototyping. Besides the aforementioned hardware design issues, the "arduino" language(really just C++) and core library had several problems both in terms of code quality and abstracting over things that shouldn't be abstracted over when working with such a limited chip(8bit, 2k SRAM...), like significant memory allocations and interactions with SREG.
My EE partner in crime ended up designing a prototyping board himself, with various creature comforts included that we needed shields for with Arduino, and I ended up writing just C with avr-libc instead of using any of the arduino library/tooling, developing a set of core modules to use the things we added to our boards, in a more flexible manner than the Arduino library. It took some time, but it saved us a lot of time and friction in our future prototyping efforts.
All that being said, there's nothing wrong with Arduino as a platform for learning and personal tinkering. I do think they could've done a better job bridging the gap between that and prototyping though.
* Ten years ago, so my memory of specifics is very fuzzy and only reflects the state of things back then.
> Arduinos are not good for much more than babby's first microcontroller project.
Baby’s first microcontroller project is exactly what they excel at and, by doing so, they made hobbyist microcontroller development vastly more accessible.
The Arduino value comes from the ease-of-starting and they made that a lot easier than the then-extant state of the art.
>So ... exactly for what the device is being sold as? Weird complaint: "I purchased an apple, and all I got was an apple that's only good as an apple."
Like I said:
>>All that being said, there's nothing wrong with Arduino as a platform for learning and personal tinkering.
I was just adding my 2 cents on Arduinos based on personal experience. That is all.
>Then you would know that ATmegas are in a lot of successful commercial products from the past.
Yes. What led you to believe I was suggesting otherwise? I made no criticism of the ATmega328, any other ATmega chip, or the AVR ISA for that matter. I could make some if I wanted to, but it doesn't seem relevant. The topic was Arduino boards, which typically contain an AVR chip, but is in fact not a chip but a dev board.
Yeah, I can't wait for this slop generation hype circlejerk to end either. But in terms of being used by people who don't care about quality, like scammers, spammers, blogspam grifters, people trying to affect elections by poisoning the narrative, people shitting out crappy phone apps, videos, music, "art" to grift some ad revenue, gen AI is already the perfect product. Once the people who do care wake up and realise gen AI is basically useless to them, the internet will already be dead, we'll be in a post-truth, post-art, post-skill, post-democracy world and the only people whose lives will have meaningfully improved are some billionaires in california who added some billions to their net worth.
It's so depressing to watch so many smart people spend their considerable talents on the generation of utter garbage and the erosion of the social fabric of society.
In my experience(currently about 15kg into a 40kg weight-loss program), eating enough fat can also be very helpful for losing weight. It seems counter-intuitive, but it works for me. Fat contributes a great deal to satiety. My diet setup has been to have breakfast and dinner only, no lunch on most days. This way I can make both meals quite calorific, filling and plenty tasty. Crucial for maintaining adherence to the setup, which is by far the hardest part of weight loss.
When you have to go 7 to 8 hours without eating before dinner you want plenty of slow-burning calories. Long chain fats, protein, slow carbs, with plenty of fiber.
My typical breakfast ends up being one slice of bread with liver pate and cheese, another with peanut butter and either nutella(if I'm doing morning cardio or some other exercise mid-day. Lots of sugar in nutella, which gets used up immediately by the exercise anyway) or various kinds of jam with no added sugar(usually pear and apple, since they're not so tart and are pretty sweet without added sugar), and a protein pudding cup(20g protein). The bread needs to be whole-grain, of course. Ideally 100% whole grain.
This ends up being about 700 calories, which is a pretty substantial breakfast. And most importantly, it includes a lot of protein(from liver, peanut butter, cheese, the bread and the pudding), a good mix of saturated fats with plenty of SCFA and MCT from the cheese and liver, mono- and polyunsaturated fat from the peanut butter, and tons of soluble and insoluble fiber from the bread and peanut butter.
This tends to keep me full until dinner time, at which point I can typically eat up to 1300 kcal depending on how active I've been.
On extremely active days, I might either add another slice of bread to breakfast, or have a protein snack and some fruit after exercise, as well as electrolyte drink with sugar in it during(important both for energy and fluid uptake).
Anyway, I'm rarely hungry except for just before eating, which is the idea. I think this would be much harder on a low-fat diet.
Yes, some people are fine, and some people are not fine at all. Even Terence Mckenna had at least one extremely bad trip(that we know of) on mushrooms that drove him to quit taking them for the rest of his life. That was with a lifetime of experience taking mushrooms and other psychedelics, and at a dose several orders of magntiude lower than the dose suggested by this paper.
And this is not a one-off thing, but a monthly regimen. So we're talking about taking a dose that's way beyond heroic, every month, for the rest of your life. That's so far outside the realm of responsible psychedelic use, I don't think any human has even come close to attempting it. The only thing that's for certain is there's no way of knowing what the psychological effects would be, but I have a hard time believing it wouldn't get extremely ugly.
