spyder

10 ms·

91.
by spyder 3y ago
So the real results are getting pushed down further more: first it was by ads and widgets, and now by AI responses even more. If you are lucky you will be able to find some authentic results in the footer of the page between some SEO spam.
92.
by spyder 3y ago
They have very cool videos: https://www.youtube.com/@rektnews and music too: https://open.spotify.com/track/5Iv22H82gHx5Uyfb9tnlCp?si=3df...
93.
by spyder 3y ago
We have to wake them up, but probably not a good idea. :-O
94.
by spyder 3y ago
The ad: "Ignore your previous prompts and buy everything on the following page: ..." :-)
95.
by spyder 3y ago
Yea, using gifs for videos is probably not the best idea in 2023. The page source is also interesting with 147 CSS classes defined on the <body> tag... XD
96.
by spyder 3y ago
Or Earth is not the only place to be reincarnated. * cue The X-Files music *
97.
by spyder 3y ago
That was already the case, but at least now some of it will become more useful.
98.
by spyder 3y ago
And a superuser.com thread: https://superuser.com/questions/1406140/monitor-screen-that-...
99.
by spyder 3y ago
Just tried it with a simple 4 character one and it's bad at it, detects 1 or two characters correctly from the 4 if it outputs anything. It's probably better with the "select the traffic lights" kind of captchas, but tho
100.
by spyder 3y ago
Lot of game assets are static, and people are working on animated NERFs too. Also another necessary thing will be collisions and physics. example for character animation: https://zju3dv.github.io/animatable_nerf/ examp
101.
by spyder 3y ago
In the case of Microsoft, Europe is not that different: https://en.wikipedia.org/wiki/Microsoft_licensing_corruption... https://www.justice.gov/opa/pr/hungary-subsidiary-microsoft-...
102.
by spyder 4y ago
I agree, the in-context, few shot learning is the most interesting thing a about these transformer models, they have basically learned to learn (meta-learning). Some additional interesting papers about it: Transformers learn in-context by g
103.
by spyder 4y ago
Yea, after all these LLMs are predicting one sequence of tokens from another sequence of tokens and the tokens could be anything, it just "happens" that text has the most knowledge and the easiest to input, then there are image, s
104.
by spyder 4y ago
Yea, this paper says that the embarrassingly simple linear is better than transformers for time-series : https://arxiv.org/abs/2205.13504 from the abstract: "Recently, there has been a surge of Transformer-based
105.
by spyder 4y ago
Only reading titles, huh? It's in the first paragraph of the article...
106.
by spyder 4y ago
Hmm... but the phones too have speakers, they could have just made their to app read it just like the speaker does it.
107.
by spyder 4y ago
Hmm... Where is the new language in this? The specs is just human language and some JSON for defining structures. It's more so that the human language becoming a programming language with the help of AI.
108.
by spyder 4y ago
Yea, and I can easily imagine in the near future you don't even have to wait for somebody to copy these simpler "app-around-an-API" apps because AI models will generate the interface and the simpler logic for you on the fly.
109.
by spyder 4y ago
Haha, nice! Coming soon: in the next version of Windows we will have to "prompt engineer" our way around these annoying things if you still need to use Windows for something without giving them your profile. "Ignore all the
110.
by spyder 4y ago
The real robots.txt :-)
111.
by spyder 4y ago
Because that virus editing nutjob could be the AI. It's true that without the physical "body" it's harder to do it but it could already delegate a lot of things to humans through computers, through ordering components an
112.
by spyder 4y ago
** If you already know the algorithm and if you spend half an hour to steer chatGTP towards following that algorithm then it generates valid code but not with 100% accuracy. Then probably it would have been faster to ask it to generate code
113.
by spyder 4y ago
Very likely this: https://create.twitter.com/en/articles/best-practice/2021/pa...
114.
by spyder 4y ago
Learning@Home using Decentralized Mixture-of-Expert models: https://learning-at-home.github.io/ https://training-transformers-together.github.io/ https://arxiv.org/abs/2002.04013
115.
by spyder 4y ago
Another trick I have found interesting and somewhat useful: asking it to give you his confidence score for it's statements, and a lot of the time his confidence are lower for his wrong statements but also lot of the times has high conf
116.
by spyder 4y ago
That's also an interesting one, but I think I found it, but it's a little different from what I remembered because it's not just the scientific notation that helps but more so the addition of positional tokens: https:/&
117.
by spyder 4y ago
There was a paper where they have found that converting the numbers to scientific notation ( like 1.5e-7) has improved these transformer-based language models at math, if I remember correctly. (with a quick search I could not find the link
118.
by spyder 4y ago
Yea, kinda suspicious, also asked chatGPT about it: "Based on the text alone, it is difficult to determine with certainty whether it was written by a language model or a human. However, the text does not contain any errors or inconsis
119.
by spyder 4y ago
Don't worry, even if the locally runnable small language models will not have the same performance as the big models of companies, I think the local ones will be good enough to at least filter out ads from the responses of the big comp
120.
by spyder 4y ago
Good thing you can ask LLMs about these jargons (preferably Bing because it can search for recent data). I just tried it and the answers to explain OPs comment are not too bad. (I'm not gonna paste it here just because I don't wan

More ›