lelag

10 ms·

31.
by lelag 2y ago
I've noticed quite a lot of Prolog content lately. I don't know if it's just a case of the Baader–Meinhof phenomenon, but it looks to me that recent article[1] about Prolog being able to improve LLM reasoning abilities renewe
32.
by lelag 2y ago
Pretty weird choice of TTS engines. None of them are anywhere near state of the art as far as open TTS system goes. XTTSv2 or the new F5-TTS would have been much better choices.
33.
by lelag 2y ago
It is hard. Lichess admins are highly skilled and I'm sure they already have a well designed infrastructure. You can see what they use at https://docs.google.com/spreadsheets/d/1Si3PMUJGR9KrpE5lngSk... The is
34.
by lelag 2y ago
It's not the same product, even if you consider just virtual machines rather than higher level services that others commenters are referring to. Sure public cloud is more expensive but you pay for the reliability of not being bound to
35.
by lelag 2y ago
It actually uses less than 3 GB of VRAM. One issue is that the research code is actually loading multiple models instead of one, which is why it was initially reported you need 8 GB if VRAM. However, it cannot be used for the same use case
36.
by lelag 2y ago
OP is probably using an AI accelerator like this: https://coral.ai/products/accelerator which works great on a PI and uses very little power. It will do the Yolo part, but you can't really expect it to do the mult
37.
by lelag 2y ago
There would not be much point. Colmap is already very capable in reconstructing a 3D scene from images from unknown poses if you have the camera intrinsics. Besides processing speed, this project (and the underlying dust3r model) strength i
38.
by lelag 2y ago
For me, the big multiplier is written communication: writing email, documentation, instant messaging—these are tasks where touch typing will make an even bigger difference. Unlike programming or using a CLI, there is no autocomplete to assi
39.
by lelag 2y ago
Honestly, maybe don’t. You love Clojure, that’s awesome. But introducing it at your company if it does not have a lisp culture already might not go so well. You say so yourself: lisp dialects like Clojure come with a steep learning curve, a
40.
by lelag 2y ago
Previous discussion: https://news.ycombinator.com/item?id=41339163
41.
by lelag 2y ago
Yes, there is: Lulu at https://objective-see.org/products/lulu.html .
42.
by lelag 2y ago
In the past, I might have used a python library like asciitable to do that. This time, I just copy pasted the raw metrics I found and asked an LLM to format it as an ASCII table.
43.
by lelag 2y ago
Possibly because it is hard to make a video streaming service that requires over 300 Mbits of bandwidth commercially viable.
44.
by lelag 2y ago
The 405b model is actually competitive against closed source frontier models. Quick comparison with GPT-4o: +----------------+-------+-------+ | Metric | GPT-4o| Llama | | | | 3.1 | |
45.
by lelag 2y ago
I don't know why I was expecting something relating to feature engineering and/or feature extraction in the context of AI research before I clicked... Now I'm there, wondering why we need full blown frameworks to handle boole
46.
by lelag 2y ago
It supports XTTSv2 which is currently the open-weight state of the art. So, pretty damn good ( https://huggingface.co/coqui/XTTS-v2/blob/main/samples/en_sa... ). Too bad that the project is in limbo a
47.
by lelag 2y ago
Or does it ? Drawing on a parallel from my experience in NLP during the early 2000s, the field was largely dominated by linguists trying to grok language and structuring rules manually. However, the most significant advancements came when w
48.
by lelag 2y ago
The way I see the current crop of AI is that it has mastery of language and a very wide set of knowledge on many different areas but no deep understanding. Ask it about anything you are not very familiar with, and you will get something tha
49.
by lelag 2y ago
You are right, I forgot they would go around during the yellow.
50.
by lelag 2y ago
NASCAR is not meant to be logical, it's meant to be an exciting spectacle. Just look at the "lucky dog" rule: it allows the first of the lapped drivers to be put in the main lap during a yellow. The goal is to create a race w
51.
by lelag 2y ago
All link shorteners, even those that use a readable URL, keep the target URL in a database. It's not possible to decode any of them without the service being available. The novelty here is that the service deliberately makes itself unu
52.
by lelag 2y ago
I would say that it misses one of the main drawback of using Htmx: it does not force you to create a proper API. When you create a standard single page application, you generally start by building a JSON API on the server side, and a client
53.
by lelag 3y ago
Given how good the newly released available-weight VoiceCraft[1] project seem to be, I would say the cat is already out of the bag. People will soon have to learn that they should not trust voiced content. In regards to scam prevention, the
54.
by lelag 3y ago
Can you share what update you are referring to ? I've played with Bark quite extensively a few month ago and I'm on the fence regarding that model: when it works, it's the best, but I found it to be pretty useless for most us
55.
by lelag 3y ago
Interesting. Thanks for the clarification.
56.
by lelag 3y ago
Some exciting projects from the last months: - 3d scene reconstruction from a few images: https://dust3r.europe.naverlabs.com/ - gaussian avatars: https://shenhanqian.github.io/gaussian-avatars - relightabl
57.
by lelag 3y ago
Yes, I think Gaussian Splats are were all the rage is. My limited understanding is that Nerfs are compute-heavy because each cloud point is essentially a small neural network that can compute its value from a specific camera angle. Gaussian
58.
by lelag 3y ago
Really cool. I wonder if it would work to use the DUSt3R project[1] to get the camera intrinsic and extrinsic parameters so you could create a scene that can be turned into a nerf from any pictures without the need for special apps to use f
59.
by lelag 3y ago
Can’t argue with this logic, more data points certainly helps. I was arguing about vision vs lidar, vision + lidar is certainly better than vision alone.
60.
by lelag 3y ago
I don't quite agree on this one. While I think that Musk choices to go full vision when he did was foulish because he made his product worse, his main point is not wrong: human do drive well while using mostly vision. Assuming you can

More ›