tomthe

9 ms·

61.
The state of Computer Vision and AI: we are far away (2012)(karpathy.github.io)
2 points by tomthe 4y ago|0 comments
62.
by tomthe 4y ago
He suggests that Microsoft should buy all the startups it can get its hands on and then lock them away from Redmond... > I already know what the reaction to this essay will be. Half the readers will say that Microsoft is still an enormou
63.
Microsoft Is Dead (2007)(paulgraham.com)
3 points by tomthe 4y ago|1 comments
64.
by tomthe 4y ago
This would of course be even more fun with ChatGPT, but it is a nice and funny demo of their whisper.cpp library. The second video is worth watching: https://user-images.githubusercontent.com/1991296/202914175-...
65.
Talk = GPT-2 and Whisper and WASM(github.com)
189 points by tomthe 4y ago|50 comments
66.
by tomthe 4y ago
Pen and paper is quite good for this. Really!
67.
by tomthe 4y ago
There are no weights and no data, only some code to create a pytorch character based network and train it. Will you provide weights or data in the future? Do you have any benchmark over Nominatim or Google maps? I think something like this
68.
by tomthe 4y ago
Very nicely done! Although it is a bit short on features (the only interaction is to reveal something after a button press?). I like the way this is monetized, but is it also possible to export a course (as an author)? If not, I would proba
69.
by tomthe 4y ago
For both Reddit and HN there are complete dumps that you can download easily within much less than a day. I have worked with both, the reddit dump from pushshift ist quite big ( https://files.pushshift.io/reddit/ severa
70.
by tomthe 4y ago
Here are two more pictures: https://newatlas.com/good-thinking/ifff-construction-inflata... But without reinforcement, this is only a toy. But I find it interesting nonetheless.
71.
by tomthe 4y ago
Can anyone compare this to DataLad [1], which someone introduced to me as "git for data"? [ https://www.datalad.org/ ]
72.
by tomthe 4y ago
Yes, I mean Pythran ( https://github.com/serge-sans-paille/pythran ). Thank you. Was Nuitka better? Pythran is quite simple to install and use in Jupyter.
73.
by tomthe 4y ago
Unfortunately, phytran is missing in the comparison. Phytran works in a lot of cases and it easy to use by just using python types. I would like to see a comparison with taichi, as taichi also seems to be interesting.
74.
by tomthe 4y ago
They all look great! But the usual customers of stock photos are not looking for dragons or cavemen taking a group selfie. And all the images are not the result of a beginner trying their first prompt, they all took many tries to be generat
75.
by tomthe 4y ago
From the cherry-picked example-images on that page, it seems like Imagen more closely follows the prompt than the open Stable Diffusion model[0]. Stable Diffusion needs a lot of hints before it makes out of the ordinary pictures. In general
76.
by tomthe 4y ago
nice idea, but one can not actually see the products unless you sign up...
77.
by tomthe 4y ago
Correct. Expect regression to the mean.
78.
by tomthe 4y ago
Thank you! It is a nice article. When I saw it on the frontpage I went to my comments to check whether it was submitted by you or someone else.
79.
by tomthe 4y ago
Please post this gif as a show hn, so it gets some attention and maybe someone builds on top of this idea!
80.
by tomthe 4y ago
Is there an easy way to do this? Are you telling me that I could have shorted UST?
81.
by tomthe 4y ago
Yes, you are of course right. One shouldn't dismiss good work because of some flaws. But anyway... this is a rather silly pie chart and I wonder how this was able to pass through multiple authors and peer-review.
82.
by tomthe 4y ago
I can't take anyone seriously who publishes a pie-chart like this: https://ars.els-cdn.com/content/image/1-s2.0-S03043894203199...
83.
by tomthe 4y ago
OpenAlex seems to be a really useful alternative to Scopus and Web of Science. Although I found that their author-disambiguation is not as good as the one from Scopus. I would like to see that they make it easier to download database dumps.
84.
by tomthe 4y ago
Hi, I work in an academic setting and I found datasette always interesting to publish datasets, but it is hard to host a python web application on institutional servers (would have to convince several people who would have to maintain this)
85.
by tomthe 4y ago
I am of course sceptical whether this notation system will gain a bigger following, but I really appreciate their nice website and how they clearly communicate their idea. They even have a nice interactive online tutorial: https:/&#x2
86.
by tomthe 4y ago
What a great random list of random websites! I liked: https://www.summerwind.eu/home-english/ "the number one horse solariums in the world!"
87.
by tomthe 4y ago
Yes, you can save more than 30 pound per year by switching of standby devices. I just calculated it for my TV: https://www.wolframalpha.com/input?i=0.01+kwh+*+24+*365*++0.... which draws 10 W in standby (which I measured).
88.
by tomthe 5y ago
Not zero coding, but idyll tries to shine in that space. You write markdown + variables and include some visualizations and out comes a scrollytelling site. https://idyll-lang.org/docs
89.
by tomthe 5y ago
"All models are wrong, some are useful" describes it very well for me. It still amazes me, how late I really understood this and how many intelligent people not fully understand it.
90.
by tomthe 5y ago
Agree. I used wallabag for years, but now I just use joplin and its webclipper for the same purpose and it works really well, including the sync over syncthing! The only downside is that Joplin does not save the position where I stopped rea

More ›