It is incredibly fast, on that I agree, but even simple queries I tried got very inaccurate answers. Which makes sense, it's essentially a trade off of how much time you give it to "think", but if it's fast to the point where it has no accuracy, I'm not sure I see the appeal.
the hardwired model is Llama 3.1 8B, which is a lightweight model from two years ago. Unlike other models, it doesn't use "reasoning:" the time between question and answer is spent predicting the next tokens. It doesn't run faster because it uses less time to "think," It runs faster because its weights are hardwired into the chip rather than loaded from memory. A larger model running on a larger hardwired chip would run about as fast and get far more accurate results.
That's what this proof of concept shows
If it's incredibly fast at a 2022 state of the art level of accuracy, then surely it's only a matter of time until it's incredibly fast at a 2026 level of accuracy.
I think it might be pretty good for translation. Especially when fed with small chunks of the content at a time so it doesn't lose track on longer texts.
The Planck mass is just the square root of the quotient of dividing the product between the natural units of angular momentum and velocity, by the Newtonian constant of gravitation.
This Planck mass expresses a constant related to the conversion of the Newtonian constant of gravitation from the conventional system of units to a natural system of units, which is why it appears instead of the classic Newtonian constant inside a much more complex expression that computes the Chandrasekhar limit for black holes.
The Planck mass has absolutely no physical meaning (otherwise than expressing in a different system of units a constant equivalent with the Newtonian constant of gravitation), unlike some other true universal constants, like the so-called constant of fine structure (or constant of Sommerfeld), which is the ratio between the speed of an electron revolving around a nucleus of infinite mass in the state with the lowest total energy, and the speed of light (i.e. that electron speed measured in natural units). The constant of fine structure is a measure of the intensity of the electromagnetic interaction, like the Planck mass or the Newtonian constant of gravitation are measures of the intensity of the gravitational interaction.
The so-called "Planck units" have weird values because they are derived from the Newtonian constant of gravitation, which is extremely small. Planck has proposed them in 1899, immediately after computing for the first time what is now called as Planck's constant.
He realized that Planck's constant provides an additional value that would be suitable for a system of natural fundamental units, but his proposal was a complete failure because he did not understand the requirements for a system of fundamental units. He has started from the proposals made by Maxwell a quarter of century before him, but from 2 alternatives proposed by Maxwell for defining a unit of mass, Planck has chosen the bad alternative, of using the Newtonian constant of gravitation.
Any system of fundamental units where the Newtonian constant of gravitation is chosen by convention, instead of being measured, is impossible to use in practice. The reason is that this constant can be measured only with great uncertainties. Saying by law that it has a certain value does not make the uncertainties disappear, but it moves them into the values of almost all other physical quantities. In the Planck system of units, no absolute value is known with a precision good enough for modern technology. The only accurate values are relative, i.e. the ratios between 2 physical quantities of the same kind.
The Planck system of units is only good for showing how a system of fundamental units MUST NOT be defined.
Because the Planck units of length and time happen by chance to be very small, beyond the range of any experiments that have ever been done in the most powerful accelerators, absolutely nobody knows what can happen if a physical system could be that small, so claims that some particle could be that small and it would collapse in a black hole are more ridiculous than claiming to have seen the Monster of Loch Ness.
The Einsteinian theory of gravitation is based on averaging the distribution of matter, so we can be pretty sure that it cannot be valid in the same form at elementary particle level, where you must deal with instantaneous particle positions, not with their mass averaged over a great region of empty space.
It has become possible to use Planck's constant in a system of fundamental units only much later than 1899, i.e. after 1961, when the quantization of magnetic field was measured experimentally. However, next year, in 1962, an even better method was discovered, by the prediction of the Josephson effect. The Josephson effect would have been sufficient to make the standard kilogram unnecessary, but metrology has been further simplified by the discovery of the von Klitzing effect in 1980. Despite the fact that this would have been possible much earlier, only since 2019 the legal system of fundamental units depends on Planck's constant, but in a good way, not in that proposed by Planck.
If you go far beyond nanoseconds, energy becomes a limiting factor. You can only achieve ultra-fast processing if you dedicate vast amounts of matter to heat dissipation and energy generation. Think on a galactic scale: you cannot have even have molecular reaction speeds occurring at femtosecond or attosecond speeds constantly and everywhere without overheating everything.
Maybe. It's not clear whether these are fundamental limits or merely technological ones. Reversible (i.e. infinitely efficient) computing is theoretically possible.
Reversible computing is not infinitely efficient, because irreversible operations, e.g. memory erasing, cannot be completely avoided.
However, the computing efficiency could be greatly increased by employing reversible operations whenever possible and there are chances that this will be done in the future, but the efficiency will remain far from infinite.
My favorite feature of the Touch Bar was that, if memory serves well, force push was right next to cancel in one of the IDEs, can't recall if Xcode or Intellij.
If your design language is “flat as we can make it” how can you visualise a third dimension? You have to already know which things are 3D touch ready.
I blame the software refresh of Apple after the 5-series UI language was removed. Minimal mechanical design with rich complex software is a beautiful contrast that strengthens how both feel.
I was gonna make a quip that someone should really get on the ":later" feature, but then I realized that the modern LLM craze more or less is that feature.
As far as I'm aware, this is not true at all. Adults actually learn considerably faster per hour spent, it's simply that most adults don't have as much time to dedicate to learning as kids do. This is confirmed both by studies as well as intuition — you get better at most things as you do more of it, studying is the same.
I wonder if it may depend on the type of learning.
Whether it's something completely new and alien - in which case children might be better, or whether it's learning through association - ie I understand that because I can connect it to previously learnt concepts ( which would favour people with more previous knowledge ).
For example, something like maths is often seen as a young persons game, but at the same time you probably don't want a 14 yr old running your company.
People keep parroting this point, but I don't think it actually applies, it's just one of those things that gets reposted a lot on the internet. When we're hiring a candidate, I generally don't go through their Github repos or blogs. I talk to them about what they've worked on and what they've done. Hobby projects can be a good starting point to talk about that, as can be blogs, but really you could start with anything. Most people start with their current day job and that's perfectly fine. You don't have to be coding both inside and outside of working hours do be a good applicant.
I'd go as far as saying it's counter-productive. I have a hobby-level project with actual users earning me some money on the side while requiring very little day-to-day involvement (roughly 2h per week) and there's no quicker way to get my door shut doing interviews than by mentioning it.
There's simply no way to package that which doesn't make the other side think that I'm gonna steal company's time at best and that I'm only looking for like a temporary gig until it takes off at worst.
I think motivation really is the key term here. Magnus is a five-time world chess champion, in a complete league of his own even when everyone else was literally only prepping to defeat him. He held the world champion title for ten years and eventually just declined to defend it. And that's relatable, if you're at the absolute top for ten years and no one manages to put a dent to it, what else is there? I think most people would look for new challenges and ways to fulfill themselves after that.
He declined to defend it because he disagreed with the way FIDE was organizing and managing the tournament. I believe this is around the time they threw him out of a tournament for wearing jeans, when he was not the only competitor present in jeans.
I think it's nearly universally accepted that his streak ended on a technicality rather than a legitimate decline/defeat.
I think it's more that he wanted to go out undefeated, rather than lacking motivation. Or rather the former driving the latter.
He made 5 title defenses. Two were against the previous generation of players, and he did extremely well. 2 were against players of his generation and were anything but compelling victories. He only won a total of 1 classical game in the 24 played, and that was in a must-win scenario because he had just lost for a final record in these matches of +1 -1 =22. And finally there was his match against Nepo which was looking to be another extremely close match until Nepo lost a critical game, and then went on monkey tilt, as is his reputation - proceeding to play horribly for the rest of the match and get wiped.
In an interview with Rogan, Carlsen stated he felt he peaked a bit before his match against Nepo, and so he probably did not view his chances of success in a world championship match as especially high. So he was going to have to spend months preparing for a match he could very well lose which would certainly tarnish his reputation as the GOAT of chess. I think this is why he couldn't find the motivation.
For instance there were new world records just around the corner. The most successful world title defenses is 6 and that was back in the early 20th century. With one more he could have surpassed Kasparov and at least tied the record.
Magnus has always been unhappy with the format of the WCC cycle. He first skipped it in 2011, when he was already the top-rated player but not yet champion (https://www.chess.com/news/view/carlsen-quits-world-champion...), and very nearly skipped it again in 2013.
Actually he recently stated that he IS still disappointed about that whole incident because nothing changed and is currently backing up hikaru on drama around similar issues.
The motivation issues can stem from poor management :)
From what I recall, he automatically lost that one game but was not thrown out of the tournament. Eventually he just stopped playing the world championship altogether, which is when he lost his title.
I don't really follow human chess, but I wonder what the new nr 1 player thought of themselves after essentially becoming the "best player in the world who doesn't wear jeans." Must be so frustrating to know there is something left to achieve but your league's shenanigans will prevent you from achieving it in an official and prestigious manner.
The jean controversy was a couple of years after Magnus stopped defending the title. It has nothing to do with it. Magnus just doesn't care about the format of the world title.
I think something broke for him while playing Caruana in 2018. The classical games were a snooze fest of defensive plays after defensive plays and everything was settled in the rapid tie break in a fairly unsatisfying manner.
He is not the first to complain about that by the way. Fischer hated the format too.
The freestyle championship was better in pretty much every way.
There was no snooze fest though in 2018 WCC. The games were extremely exciting, with unbalanced pawn structures. They all ended in draws only because of their strong defensive skills and a touch of luck in a few games.
Unbalanced pawn structure is a feature of the Sveshnikov Sicilian but Caruana had done a lot of prep and it was obvious. Carlsen quicky left the main line for the boring 7. Nd5. Plus, Carlsen missed a lot of good moves because he had to play it safe. To me, it was boring chess of the highest level.
A lot of it felt like watching engines by proxy. One prepared well on a very complex opening. The other found the best meta counterplay and held until he reached the tie break.
Game 12 is a travesty. It was clear he just wanted to move to rapid.
You are clearly confused.
1. It was Magnus who played Sveshnikov, and Fabi who played the 7. Nd5 line.
2. 7. Nd5 line is not boring, it's one of the sharpest things to play against Svehsnikov, with pawns marching on both sides of the board. Mainline Sveshnikov on the other hand can be very drawish, with maneuvering play, where no side has a pawn break in sight.
3. It was the beginning of the Leela era, with many new opening concepts. It was a short-lived golden age, with players excited to try new ideas. After this match Magnus had a second peak, incorporating new knowledge, and won quite a lot of sharp games in Sveshnikov.
4. Well, game 12 was interesting, but EV of extra time to relax and prepare for rapid games with Caruana in 2018 was quite high. Fabi wasn't a very good rapid/blitz player back then.
5. Carlsen-Nepo WCC was a massacre, while the games were quite boring, except a few at the beginning. That match or any of the next matches would be a much better example for a case against classical chess.
I myself quite like chess960, and it's clearly the right direction. Opening prep is absurd now, especially at super GM level.
But that's it. There was no win because the opportunity to even compete was taken away. Imagine you train your whole life and finally win the Olympic Gold medal, but everyone knows it's only because the true nr 1 ignored to compete in this format.
Winning a title is never about facing the greatest possible opponent. Even the people who show up aren’t at their absolute best, but consider everyone who doesn’t devote their lives to the sport. The greatest potential chess player of all time likely does something else with their lives.
Without prep Magnus would be vastly less likely to win, and he’s not doing the prep because he’s not competing. How exactly is that different than someone not devoting themselves to the sport 20 years ago?
I make no claims about what’s going on in people’s heads here just the underlying reality.
I was pissed I didn’t go to nationals in high school largely because I got no sleep the night before due to a crappy hotel stay. Losing a close game to board 1 while trying to stay awake sucked, but I was hardly the only person off my game, so that’s just how things workout.
This is plain wrong for anybody that has been actually following chess.
He said like a year before declining that he would only defend against Alireza because he was the younger generation. Nobody believed him, Alireza didn't win the Candidates and he said "no thanks" and then everybody was surprised pikachu.
The Jean thing was way later and in an unrelated event.
reply