ynniv

37 ms·

31.
by ynniv 2y ago
For a second I thought it had been ported to wasm! This is still a browser based VNC
32.
by ynniv 2y ago
Everyone should be interested in Qubes OS.
33.
by ynniv 2y ago
Is the clipboard also a "data leak"?
34.
by ynniv 2y ago
"you may have concerns about privacy" Way to FUD: these models are local unless you enable ChatGPT integration.
35.
by ynniv 2y ago
Umm, two 3090's? Additional cards scale as long as you have enough PCIe channels.
36.
by ynniv 2y ago
FWIW, I asked Claude to translate and compare its translation to this as well as Google Translate. I then pasted top threads from this page, and asked it to retranslate the original. First: Hello, today I'll be teaching you about t
37.
by ynniv 2y ago
I'm not sure that this disproves my original point that self driving cars and flying cars don't belong in the same list because they are fundamentally different engineering problems.
38.
by ynniv 2y ago
Engineering is about focusing on what matters. There's no point in talking about flying cars: they will exist when portable fusion exists, so just talk about that.
39.
by ynniv 2y ago
It's long, so I'm skimming a little and... flying cars. If you don't know why we don't have flying cars, you're not a good engineer. It really doesn't matter what prestigious lab you ran, as that apparently did
40.
by ynniv 2y ago
This is honestly the real "hacker news". $250 for what was likely millions of dollars of visual design and global household name recognition
41.
by ynniv 2y ago
The people have no will at the Fed
42.
by ynniv 2y ago
These are 128GB each . Also, Nvidias inference speed is much higher than Apple's. I do appreciate that my MBP can run models though!
43.
by ynniv 2y ago
I'm pretty frugal, but my first thought is to get two to run 405B models. Building out 128GB of VRAM isn't easy, and will likely cost twice this.
44.
by ynniv 2y ago
Either they would have actually "gone around" and executed the original approach (which would have been aborted when the gear didn't lock), or they had had mechanical problems and wouldn't have been able to make this 180
45.
by ynniv 2y ago
According to the timeline I saw they were cleared before declaring an emergency. This means at the time of emergency the standard checklist required their gear had been locked and flaps set. Either they were not operating within regulation
46.
by ynniv 2y ago
Cause of death was landing halfway down a runway at high speed with no gear. Why were they cleared to land if they hadn't set flaps and locked gear? There was ocean in front of them both before and after their turn: landing there would
47.
by ynniv 2y ago
Lol, that's awesome. Yeah, it's pretty easy to push Claude around. I think that's okay because you're only fooling yourself, but it should really have a bit more of a spine.
48.
by ynniv 2y ago
Glad you checked it out! I generally avoid ChatGPT at this point. Do you have a good example of incorrect responses? The easy ones tend to be trained away in the next model update
49.
by ynniv 2y ago
I think o1 and o3 are OpenAI stretching the usefulness of 4o. Anecdotally the latest Claude model is naturally better than 4o, and having it use inference tokens to think through or verify work results in similar o1/o3 gains. Are we su
50.
by ynniv 2y ago
Yes, but that isn't their value, and they can use other tools to provide these answers reliably.
51.
by ynniv 2y ago
Honestly, finding something that Claude confidently gives me the wrong answer to is harder than getting it to clearly explain its reasoning: Me: Suzzie has three brothers. Each brother has two sisters. How many sisters does Suzzie have? Cl
52.
by ynniv 2y ago
It's certainly a common complaint, but I'm not sure that tokenizer tricks are insurmountable problems: How many "r"s are in s r t r a w b e r r y s t r a w b e r r y? Let me count the "r"s one by one: s (r[1])
53.
by ynniv 2y ago
A healthcare professional could give you terrible advice, which is why we have legal incentives to reduce harm. Now that AI companies are worried about similar regulation, the best LLMs are very reluctant to give medical advice.
54.
by ynniv 2y ago
I'm not convinced AI is as hamstrung as people seem to think. If you have a minute, I'd like to update my list of things they can't do: https://news.ycombinator.com/item?id=42523273
55.
by ynniv 2y ago
Me: How many squirrels are in my backyard? Claude: I'm unable to see your backyard or count any squirrels there. I would need either a photo of your backyard or specific information from you about the squirrels you've observed to
56.
by ynniv 2y ago
This isn't true, and especially not true of Claude. Could you say this in a more testable way?
57.
by ynniv 2y ago
This is true of people as well. And yes, there is acceptable inaccuracy in legal, compliance, healthcare, safety, etc, it's just very low. There's even a saying for dealing with it: "get a second opinion".
58.
by ynniv 2y ago
Which... checks notes... we don't have. In 2012 the idea of self-driving cars was almost ridiculous, let alone dedicated infrastructure for them. We've had some research papers, and some initiatives have been proposed. I agree t
59.
by ynniv 2y ago
Did you provide the same reference material that a developer would use?
60.
by ynniv 2y ago
Hallucination is an interesting problem for both LLMs and people. This is why we like syntax and unit tests. I think there may be ways to work around it.

More ›