valine

20 ms·

211.
by valine 3y ago
Those gestures that have to be memorized allow you to efficiently navigate your phone. I don't want a phone that sacrifices efficiency for ease of use. If you're actually struggling to use your phone please just turn on assistive
212.
by valine 3y ago
You can split your AI models between two 4090s and have access to 48GB of VRAM for less money. You do take a slight performance hit compared to running on one card, but the 4090 also has more CUDA cores which help mitigate the slowdown.
213.
by valine 3y ago
It’s not a direct apples to apples comparison, but compared to a RTX 4090 this isn’t compelling. Other than the 32GB of VRAM there’s not much going for this card. For $4k most people would be better off with dual 4090s.
214.
by valine 3y ago
I’ve been imagining the future will be smart glasses paired with mini pocket keyboards, like the bottom half of a blackberry. Or who knows maybe everyone will lean heavily into dictation. I think there’s also a future where hand tracking ge
215.
by valine 3y ago
Unless you're a professional athlete no one is going to pay you to work out. It has no value for society.
216.
by valine 3y ago
Does it? If I pay you to move rocks back and forth in a field does that work have dignity? Making someone work a job that could be done cheaper and faster by a robot doesn’t benefit anyone. You’re destroying economic value and wasting the w
217.
by valine 3y ago
There’s no real dignity in work that can easily be done by a robot. A lot of these jobs make people miserable anyway, maybe we shouldn’t be fighting so hard to keep them.
218.
by valine 3y ago
If you fill the search results with enough SEO garbage eventually ads will look better by comparison. If ads are making your user experience better you have a serious problem.
219.
by valine 3y ago
It’s disappointing to see google lacking basic self awareness like this. Ten years ago they seemed highly competent and nearly unstoppable.
220.
by valine 3y ago
The launch was scheduled for 4/17, 4/20 was a backup window. If anything they delayed the launch to hit 4/20, not rushed it.
221.
by valine 3y ago
This is cringy. I’m no fan of Musk but SpaceX is making consistent progress on their fully reusable rocket. There’s plenty of reasons to criticize Musk without pretending SpaceX is a failure.
222.
by valine 3y ago
Regen breaking should in theory help mitigate that. A lot of that energy is recaptured when you slow down. Heavier vehicles have more energy to recapture, assuming the drive train is up to it.
223.
by valine 3y ago
It means general support for hardware acceleration in linux. OpenGL is used more places than you would think. Switch to software rendering and you’ll immediately see the difference just navigating the OS.
224.
by valine 3y ago
For what it’s worth M2 Ultra is the newest chip in the Apple silicon lineup. It was only introduced last June.
225.
by valine 3y ago
I’m surprised we weren’t doing this already. I’d like to see what happens if you train a small language model on preschool level reading material, and ramp up both the model size and training data complexity as you go. My hope would be you’
226.
by valine 3y ago
Good science takes time. Anyone making definitive claims, including this article, is full of it. LK-99 is probably not a super conductor, key word probably. It’ll be definitive when the original samples have been independently tested.
227.
by valine 3y ago
Theft takes the original, piracy makes a copy, AI art remixes the original. I’m not sure how to classify AI art but it definitely isn’t theft.
228.
by valine 3y ago
If this were true ChatGPT would be useless for any topic that doesn't have a vast Q&A dataset. The purpose of instruct fine tuning is to illicit the model to output in a question answer format. The bulk of the model's knowledg
229.
by valine 3y ago
For complex questions GPT4 works much better than 3.5.
230.
by valine 3y ago
You’re assuming the bulk of ChatGPT’s coding knowledge comes from stack overflow. I suspect training directly on code bases and documentation is more important.
231.
by valine 3y ago
Reusability helps with long term reliability. You would probably want to avoid launching nuclear material on a new rocket, but a human rated rocket like Falcon 9 would be totally fine.
232.
by valine 3y ago
> Today's AI models are missing the ability to reason abstractly, including asking and answering questions of "Why?" and "How?" I’m convinced the author has never used an LLM before.
233.
by valine 3y ago
Seems reasonable, then again we're dealing with a potential superconductor where sudden discontinuity is the expected result.
234.
by valine 3y ago
Original team had 20 years to play the synthesis lottery. Maybe you tweak the setting just enough and that low resistance drops to zero. Who knows, I’m still feeling optimistic. He said in the video that their sample was more pure than the
235.
by valine 3y ago
That’s what I was thinking. In the video he brushes it off as a potential equipment malfunction, but perhaps it’s the effect everyone is looking for.
236.
by valine 3y ago
What’s the explanation for the sudden drop between 230K and 250K? It’s not dropping to zero but something is happening there.
237.
by valine 3y ago
There's probably other similar chemistries that also super conduct at room temperature. Call it the marching tetrahedra of super conductors.
238.
by valine 3y ago
Is an H100 clocked at 100GHz too much to hope for?
239.
by valine 3y ago
The thermodynamic limits of computation are so far below what we have today it would be a game changer to approach them.
240.
by valine 3y ago
It’s not over until multiple labs fail to reproduce it. A definitive no will take months.

More ›