imjonse

10 ms·

151.
by imjonse 2y ago
Noone said you cannot reuse the tailwind/nextjs template you used for crypto hustling if you genuinely feel you can move humanity forward.
152.
by imjonse 2y ago
Are you seriously suggesting Altman is a genius? Or Musk for that matter?
153.
by imjonse 2y ago
That and the NFSW finetunes that will inevitably follow; unlike the text-gen finetunes these could really cause trouble with deepfakes.
154.
by imjonse 2y ago
Something like this? https://hazyresearch.stanford.edu/blog/2024-07-01-jrt
155.
by imjonse 2y ago
yes, that is clearer indeed. However S4 and Mamba class models have also performed well at small scale and started lagging with larger models and larger context sizes, or at particular tasks.
156.
by imjonse 2y ago
To their credit, the authors (Y. Bengio among them) end the paper with the question, not suggesting they know the answer. These models are very small even by academic standards so any finding would not necessarily extend to current LLM scal
157.
by imjonse 2y ago
the author correctly intuits that a bunch of free users that may spread the word are worth more than N*0.95$ so the outrage on display here is not only unnecessary but counterproductive too
158.
by imjonse 2y ago
The author explicitly said to ask for one if needed, why would it be appalling?
159.
by imjonse 2y ago
Llama-3-V models do that, but are not published.
160.
by imjonse 2y ago
It is a family of multimodal models based on pretrained Qwen2-72B-Instruct LLM and InterViT vision encoder. There are three variants differentiated by the way the vision tokens are used: decoder-only (like the majority of existing VLM), usi
161.
by imjonse 2y ago
Remote US-only or Remote global?
162.
by imjonse 2y ago
One can start non-distruptively with uv by only using it as a pipx replacement at first ( uv tool). Nice and fast.
163.
by imjonse 2y ago
Thank you, my bad, you are right, as the other sibling comment.
164.
by imjonse 2y ago
Thank you. I was mislead by the date on the post and did not know the book was older. TIL.
165.
by imjonse 2y ago
The book's title is a nod to the Roger Waters album/song that deals with the same theme.
166.
by imjonse 2y ago
"next few thousand days" can we stick to years as a unit of measure and not spread Sam Altman's phrase :)
167.
by imjonse 2y ago
Warnings do not really work in practice. What if these activities are not simply self-harming but destroy the families of the addict and large parts of the fabric of society? Even you mention societal chaos. How does the libertarian world-v
168.
by imjonse 2y ago
Training data is crucial for performance and they do not (cannot) share that.
169.
by imjonse 2y ago
I am glad most people do not talk in real life using the same style this message was written in.
170.
by imjonse 2y ago
Apart from results on benchmarks, what sets Allenai models apart - Olmo/OlMoE/Molmo - is they are fully open, not just open-weights/free to use. The datasets used, a crucial ingredient, are also disclosed and open. UPDATE: th
171.
by imjonse 2y ago
Architectural innovations can definitely help with training/inference speed - that was the case with convolutional networks too - but for model performance (as in 'intelligence' not speed) dataset size and quality were always
172.
by imjonse 2y ago
It's a lot more comforting to work on AI (or any tech for that matter) if you believe or force yourself to believe it will be used for good. OpenAI's for-profit and gatekeeping approach is unlikely to be the path to the prosperity
173.
by imjonse 2y ago
maybe try out[local_i, local_j] ?
174.
by imjonse 2y ago
I see all these three (SD40/41, MH-Z19 and S8) on Aliexpress, but in yesterday's discussion on the sibling post it was said that the SD40 is likely a fake since Switzerland based Sensirion do not distribute there. Do you have a bl
175.
by imjonse 2y ago
auth generators are being done for 8.0
176.
by imjonse 2y ago
Aider will probably have some numbers at https://aider.chat/docs/leaderboards/
177.
by imjonse 2y ago
and was considered too dangerous to be released publicly.
178.
by imjonse 2y ago
Maybe it's a real person from Mechanical Turk who had a bad day?
179.
by imjonse 2y ago
They are too prestigious to try shumering it.
180.
by imjonse 2y ago
Hopefully not Cavendish, as those are too sugary for monkeys and you'll just get hallucinations.

More ›