modeless

31 ms·

631.
by modeless 1y ago
They generally pay a premium to the stock price before the transaction is announced, which means shareholders get more than the market value of their investment, which they can immediately roll into other investments. Usually people don&#x2
632.
by modeless 1y ago
It's not the first time. To some extent it's just how the government contracting game is played. Regrettably. https://en.wikipedia.org/wiki/Blue_Origin_Federation,_LLC_v....
633.
by modeless 1y ago
This is not an update for SpaceX fans. It's aimed squarely at people who have opposed SpaceX's expansion in Florida at Kennedy Space Center and Cape Canaveral. These people have made arguments about safety and the environment as w
634.
by modeless 1y ago
> each parameter exists in two forms simultaneously during training: a full-precision 32-bit floating-point value (p) used for gradient updates, and its binarized counterpart (pb) used for forward computations So this is only for infere
635.
by modeless 1y ago
I find Slack and Discord to work fine in browser tabs and never felt the need to install their desktop apps. VSCode/Cursor is the only Electron app I felt actually provided value.
636.
by modeless 1y ago
Why are model providers allergic to version number increments?
637.
by modeless 1y ago
Every home should have a heat recovery ventilation system. It gives you fresh outside air in the whole home, but filtered to remove PM2.5 and without the inefficiency.
638.
by modeless 1y ago
Yeah I saw a lamp for this and it had a proximity sensor to prevent overexposure. There's no way I'm buying a lamp that needs that to be safe. Especially if it's expensive and only lasts a year or two.
639.
by modeless 1y ago
This is an overstatement. Yes, UB does not necessarily cause a violation of memory safety, but triggering UB alone is not the goal of an attacker. UB is a means to an end and the end is usually a violation of memory safety leading to arbitr
640.
by modeless 1y ago
This may say more about C++ than JSON
641.
by modeless 1y ago
I wouldn't expect a library like this to be secure. If you want it to be memory safe, compile it with Fil-C.
642.
by modeless 1y ago
VP is a low rank at a bank but a high rank at Meta. I would be very surprised if anyone at Meta with VP in their title made less than $1M total compensation. "Director" at Meta is also a high rank.
643.
by modeless 1y ago
Cost is inclusive of energy and a lot more. It's a better metric.
644.
by modeless 1y ago
Ah, I see I was in fact reading it wrong. So 9 is definitely an unusual but not unprecedented number of submissions.
645.
by modeless 1y ago
It was within the allotted time. If I'm reading the scoreboard correctly [edit: I wasn't], the human teams typically submitted dozens or hundreds of attempts at each problem.
646.
by modeless 1y ago
Human brains and cloud instances are not remotely equivalent. What you can compare on an equivalent basis is cost.
647.
by modeless 1y ago
It doesn't matter how many instances were running. All that matters is the wall clock time and the cost. The fact that they don't disclose the cost is a clue that it's probably outrageous today. But costs are coming down fast
648.
by modeless 1y ago
More information on OpenAI's result (which seems better than DeepMind's) from the X thread: > our OpenAI reasoning system got a perfect score of 12/12 > For 11 of the 12 problems, the system’s first answer was correct.
649.
by modeless 1y ago
Is there a technical reason for the web limit to be 16 GB specifically? Or is it just a round number picked so that the limit could be standardized? Also, has the limit on JS heap size (and ArrayBuffer size) also been relaxed to 16 GB or is
650.
by modeless 1y ago
It's not just visual. You also need a representation of the rules of the game and the strategies that make sense. The puzzles I'm solving are not straight Sokoban, they have per-game varying rules that need to be discovered (again
651.
by modeless 1y ago
Yes, vision is a problem, but I don't think it's the biggest problem for the specific task I'm testing. The memory problem is bigger. The models frequently do come up with the right answer, but they promptly forget it between
652.
by modeless 1y ago
Spatial reasoning is weak, but still I frequently see models come up with the right answer in reasoning steps, only to make the wrong move in the following turn because they forget what they just learned. For models with hidden reasoning it
653.
by modeless 1y ago
In general I think the more of the scaffolding that can be folded into the model, the better. The model should learn problem solving strategies like this and be able to manage them internally.
654.
by modeless 1y ago
I've been testing LLMs on Sokoban-like puzzles (in the style of ARC-AGI-3) and they are completely awful at them. It really highlights how poor their memory is. They can't remember abstract concepts or rules between steps, even if
655.
by modeless 1y ago
The iPhone was not a completely new consumer device form factor. There was a huge existing market for cellular phones, even smartphones with touchscreens. There is no equivalent pre-existing market for smart glasses today.
656.
by modeless 1y ago
Vision Pro is the same form factor as Meta's glasses in the same sense that a semi truck is the same form factor as a moped.
657.
by modeless 1y ago
It is "for a completely new consumer device form factor"
658.
by modeless 1y ago
Karl Guttag has published far more information than you ever wanted to know about Lumus in the past, e.g. https://kguttag.com/2021/05/24/exclusive-lumus-maximus-2k-x-...
659.
by modeless 1y ago
The Meta Ray-Bans have been extremely successful for a completely new consumer device form factor. But they don't have a screen. Meta is releasing new glasses with a screen and this is a look into the display technology they are using.
660.
by modeless 1y ago
This is squarely on Apple. Everyone knows the mark of a trustworthy store is that they will direct you to their competitors when they don't have what you're looking for. The web doesn't have a approval process at all and yet

More ›