Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
revelio
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
36 ms
·
301.
▲
by
revelio
3y ago
A huge part of the problem is driven by Chinese hospitals. The PRC decided that they wanted China to catch up to the west in science, so made a role that to get promoted as a doctor you have to get some papers published in international jou
302.
▲
by
revelio
3y ago
It's nonsense, GoF didn't seem to play any role in vaccine development. It's what virologists tell themselves and each other.
303.
▲
by
revelio
3y ago
Decoder only architecture? What is this? That doesn't sound like a transformer at all, are you saying gpt4 uses a totally different algorithm?
304.
▲
by
revelio
3y ago
Android forks.
305.
▲
by
revelio
3y ago
What's the link? When people bring up communism, other people describe what happened in communist systems. How is that guilt by association? What aww_dang means by that is that someone will write something libertarian or conservative,
306.
▲
by
revelio
3y ago
By the way, on macos the preview tool can do this. You can drag and drop pages between documents using the sidebar and reorder them the same way.
307.
▲
by
revelio
3y ago
It was an AI misinformation story, so you'd think it'd be in his new wheelhouse. But we all know what "misinformation" means when people like that say it. They don't mean it literally.
308.
▲
by
revelio
3y ago
Guess it depends what you mean by technical. I randomly picked another, "Product fairness testing for developers". It seems to consist of two Googlers reading a codelab out loud. And I mean that literally: https://youtu
309.
▲
by
revelio
3y ago
So before the British, there was absolutely no Hindu/Muslim tension in south east Asia at all, it was all peace and friendship? The Mughals were taxing non-Muslims just for being infidels in the 1600s and that led to the Maratha empire
310.
▲
by
revelio
3y ago
Quite marginal? Here is a video compilation of just cases where MSNBC promoted just the Hamilton 68 dashboard, which supposedly tracked "Russian linked" bot and troll activity. It's 11 minutes of clips: https://w
311.
▲
by
revelio
3y ago
This one: "This is why Google built TPUs. This alone justifies the whole program by itself. This level of natural language understanding, once it is harnessed for applications and made efficient enough for wide use, is going to revolu
312.
▲
by
revelio
3y ago
Maybe "A journey to protect the Great Barrier Reef using ML"? The video is five minutes long, is basically a nature advert complete with stirring music, has a bizarre cartoon in the middle of it and is all about an attempt to cont
313.
▲
by
revelio
3y ago
It's an expectations mismatch. Google used to be the most loved company by developers. It was confident, strong, did wonders for dev comp and gave away lots of cool stuff for free. The only way from there was down, really. Then they di
314.
▲
Interview with Sam Altman about AI risk and OpenAI's success
(thefp.com)
10 points
by
revelio
3y ago
|
2 comments
315.
▲
by
revelio
3y ago
A year ago, not so recent in AI time. The top comment there is quite amusing given how things have played out.
316.
▲
by
revelio
3y ago
Of course he's a doomer. He's compared himself to Robert Oppenheimer and by his own admission the only reason he isn't calling for a complete halt to all AI research is because he's too pessimistic it'd be possible.
317.
▲
by
revelio
3y ago
Here's a rewritten headline, definitely not ChatGPT approved: "Salty ex-Googler tries to slow down competition by mounting misinformation campaign in press". What a ridiculous charade. For years Google reacted to every OpenAI
318.
▲
by
revelio
3y ago
Western countries prefer to cut off access to banking and individual social media accounts rather than the internet. It's more effective and direct. Look at the Twitter files, Canada for examples. Cutting off the whole internet is only
319.
▲
by
revelio
3y ago
So it's the Brits fault that Pakistan/India immediately went at each others throat the moment they were gone and the solution was to keep both under "supervision" i.e. empire control? When exactly could they have left wi
320.
▲
by
revelio
3y ago
Amazing how much faster f2fs is than btrfs. Now I have some buyer's remorse over using btrfs...
321.
▲
by
revelio
3y ago
The author explains the trend of lower innovation in personal computing, which I think we all feel, as to do with corporate greed and the solution is "academics, non profits, non-VC funded companies, hobbyists" etc. But this is no
322.
▲
by
revelio
3y ago
Eh what destroyed proprietary unix was Windows NT, which was proprietary everything. Linux came along and mopped up what was left, but was able to do so largely because UNIX wasn't all that proprietary. As the thread points out, Sun ha
323.
▲
by
revelio
3y ago
The models aren't based on observations and are regularly contradicted by them. GCMs are based on, in theory, physical laws. In practice they don't compute the right temperatures even for today, or the past, so we know they are in
324.
▲
by
revelio
3y ago
If the data is of sufficiently low quality (and long term temperature data is), then it doesn't matter if it's the best analysis of the best possible. It's still going to generate unreliable conclusions.
325.
▲
by
revelio
3y ago
Raters prefer verbose outputs, probably because they want to see the reasoning to help them judge if the answer is getting it right for the right reasons.
326.
▲
by
revelio
3y ago
Nerd point of order here: Star Trek TNG had a ship in which a key member of the bridge crew was an android. They routinely relied on Data for all kinds of critical things. And although the ship was manned by people, it couldn't functio
327.
▲
by
revelio
3y ago
Like everyone else, you're not reading what you don't want to see. I said people flag stories that are research papers . As in, published by doctors. Not comments, not conspiracies, just collected data. Doesn't matter. Peo
328.
▲
by
revelio
3y ago
The relevant comment from the second video is here: https://youtu.be/qbIk7-JPB2c?t=1584 He says that GPT-4 started out being much better than GPT-3.5 at this task (draw a unicorn in TikZ), as the model trained longer it got
329.
▲
by
revelio
3y ago
Pfizer didn't have a monopoly!
330.
▲
by
revelio
3y ago
You do see that we're discussing the underlying principle here, right? Please, try to put aside the specifics of the exact hardware and understand that what matters here is the EU's descent into attempting to centrally plan techno
More ›