Otterly99

7 ms·

91.
by Otterly99 7mo ago
But chess models aren't trained the same way LLMs are trained. If I am not mistaken, they are trained directly from chess moves using pure reinforcement learning, and it's definitely not trivial as for instance AlphaZero took 64 T
92.
by Otterly99 8mo ago
Thanks for sharing this. I also was under the impression that queries cost were mostly meaningless, but it seemed only is true for fresh sessions and short queries. I have to say, the result is less dramatic than I expected but still more s
93.
by Otterly99 8mo ago
Art in general is a bit weird like that. The value of a piece is definitely not completely tied to its physical attributes, but the story around it. The story is what creates its scarcity and generates the value. It is similar for collectib
94.
by Otterly99 8mo ago
Have you even read the article? It is literally citing farms in Kenya where workers look at user footage and annotate videos.
95.
by Otterly99 8mo ago
It is so frustrating indeed reading about these wildly exaggerated biological claims. The whole synthesis pipeline requires so much specific equipment and knowledge that at your kid in his/her basement would actually need a whole lab.
96.
by Otterly99 8mo ago
There has been a lot of research into discovering new physics (starting by discovering old physics) since the last 5-6 years and it always require: - A lot of high-quality data - Some careful design - (Not always) some external knowledge to
97.
by Otterly99 8mo ago
I was asking myself the same question this morning and kind of came to the same conclusion as you. It makes much more sense to design a automated drone piloting system built with a decision-making algorithm than use a LLM. For Mass surveill
98.
by Otterly99 8mo ago
If I'm not mistaken, BERT is a classifier (enters text, outputs labels) so it is not a "Language model", as it cannot be used for text generation.
99.
by Otterly99 8mo ago
There is this paper that proposed data compression as a way to judge the ability of a LLM to "understand" things correctly, training on older texts and trying to predict more recent articles: https://ar5iv.labs.arxiv.or
100.
by Otterly99 8mo ago
Do you have any good resource on explainable AI (XAI)? Most books I read on the subject tend to be a bit shallow and not very informative.
101.
Lab-grown organoids trained to solve CartPole(news.ucsc.edu)
2 points by Otterly99 8mo ago|0 comments
102.
by Otterly99 8mo ago
In general (not always, but it is mostly true) philantropy from billionaires and very profitable companies tend to be overshadowed by how much they profit from a system biased toward enriching them (see: The divide by Jason Hickel). A small
103.
by Otterly99 8mo ago
Maybe I am missing something obvious on the website, but where is the documentation? Where do you explain what each number mean, or at least a short overview of what the models are being tested on?
104.
by Otterly99 8mo ago
What if the car that you want to wash is already at the car wash, but you have a second car? That's still a dumb question nonetheless because you probably need to drive both cars back at some point.
105.
by Otterly99 8mo ago
Interesting article, and thanks for the introduction to "philosophy bro". I think the Copenhagen interpretation of Ethics is really a misnomer. In quantum physics, a particle can exist in a superposition of states until you observ
106.
by Otterly99 8mo ago
Hate to be that guy, but the Hazda live in Tanzania.
107.
by Otterly99 8mo ago
Counter-argument: what if LLMs can help alleviate a doctor's work by providing quick diagnostic for simple cases? How much time does a doctor spend writing prescriptions for cough-like symptoms? How much time does an ophthalmologist s
108.
by Otterly99 8mo ago
You might not get great musicians from using Suno, but an ad company might decide to just generate a jingle rather than hire a musician to do it. Same with images/videos. The result might not be great, but the companies does it in 3 mi
109.
by Otterly99 8mo ago
Just my opinion, but if you present it in a way that first explains the problematic, then explain what other similar tool fail to solve, and have a genuine understanding of the tool in the sense that you can understand and answer questions
110.
by Otterly99 8mo ago
To play devil's advocate, if you were serious about building a product, whether it was hand-coded or vibe-coded, you would iterate through the work and implement functionalities step-by-step. But with vibe-coding, you might not give en
111.
by Otterly99 8mo ago
The Fourier transform one is really cool. I loved the 5x oscilloscopes builder and the "draw your own wave" widget. Also, seeing the correlation coefficients of the 3Hz + 7Hz wave was really neat.
112.
by Otterly99 8mo ago
Well said. I use it the same way. Sometimes, a technical book will assume that you know a concept or will even use an acronym that is not explained (but obvious) or is just plainly not very explicit. I also use it to directly test my knowl
113.
by Otterly99 9mo ago
They do work to some extent. If doctors wanted to be thorough*, they would use two control groups. A placebo and a "don't do anything" group. Since the placebo effect is known to reduce pain, it is not ethical to not at least
114.
by Otterly99 9mo ago
Exactly. Once you start using it intelligently, the results can be really satisfying and helpful. People complaining about 1000 lines of codes being generated? Ask it to generate functions one at a time and make small implementations. Peopl
115.
by Otterly99 9mo ago
I completely agree. In my lab, we have been struggling with automated image segmentation for years. 3 years ago, I started learning ML and the task is pretty standard, so there are a lot of solution. In 3 months, I managed to get a working
116.
by Otterly99 9mo ago
Thank you for bringing this nuanced view. I also am not sure why so many people are vehemently against this. I would bet that at least 90% of researchers would agree that the writing up is definitely not the part of the work they prefer (to
117.
by Otterly99 9mo ago
Probably because: - Banning things for non-adults is easier, they don't vote. - In the eyes of the law, social media is now seen as cigarettes and alcohol. Something teenagers should not be allowed to have access to, but is fair game o
118.
by Otterly99 9mo ago
Thank you for pointing that out. It irritates me every time I see the "water waste" argument. A toilet tank contains 5-13L of water, so if everybody started flushing only half the time they piss, that's an average of 4.5L of
119.
by Otterly99 9mo ago
Commenting on the link you provided: is using a LLM the right approach for this? Why not train a specified model to generate accurate CAD drawings, rather than offloading the task to some LLM that, while it might have some knowledge of mach
120.
by Otterly99 9mo ago
Thanks for the paper, I just read it and loved the approach. I hope the concept of using data compression as a benchmark will take off. In a sense it is kind of similar to the maxim "If you cannot explain something in simple terms, you

More ›