Palmik

13 ms·

241.
by Palmik 2y ago
So does the frontpage of WSJ, though of course it leans in the other direction. Ability to block certain topics completely would be nice -- though it goes for all media. With ever smaller / faster (local) LLMs, I could imagine a nifty
242.
Zen: Firefox Based Browser(zen-browser.app)
72 points by Palmik 2y ago|43 comments
243.
by Palmik 2y ago
Did you know that many demos on events like Google IO are scripted or hardcoded? I'm saying that not to advocate for such practices, but rather to provide context. For what it's worth, it was quite clear to many that the robots ar
244.
by Palmik 2y ago
This is still four times more expensive than Cartesia ( https://cartesia.ai/ ) and three times more expensive than OpenAI's TTS API. In general, TTS APIs seem to run with much higher margins than LLMs from what I know.
245.
by Palmik 2y ago
Every time there's a thread about models from Meta, there's a flood of comments clarifying that they aren't really open source. So let's also set the record straight for FLUX: only one of the models released is open sour
246.
OpenFLUX.1(huggingface.co)
30 points by Palmik 2y ago|8 comments
247.
by Palmik 2y ago
The OpenFLUX project is derived from Schnell, as such it's also Apache 2.0. Unlike regular Schnell, it can be further fine-tuned and can use a proper CFG sampling. Context: FLUX is state of the art text-to-image generator. It has very
248.
by Palmik 2y ago
Really love the styles. Very tasteful and subtle. I understand you want it to be framework agnostic, but are you considering version of this for Radix UI (ala shadcn) / React Aria Components?
249.
by Palmik 2y ago
Saw the YouTube video when it was first posted, and it could be a great match for a new project I have in mind. Is there a practical implementation yet that supports not just strings, but also lists and maps? Would be great to see it integr
250.
by Palmik 2y ago
That's not necessarily my position. I think laws can evolve, but they need to be applied fairly. In this case, it's heading in a direction where only the blessed will be able to compete.
251.
by Palmik 2y ago
The only entities that will win with these lawsuits are the likes of Disney, large legacy news media companies, Reddit, Stack Overflow (who are selling content generated by their users), etc. Who will also win: Google, OpenAI and other corp
252.
by Palmik 2y ago
Did they help regular people defend themselves while saving on legal costs or not? Most of these cases wouldn't be defended at all otherwise.
253.
by Palmik 2y ago
The legal system has a great sense of self preservation. They will surely fight anything that possibly encroaches on their domain, especially things that give non lawyers the tools to defend themselves without feeding the machine.
254.
by Palmik 2y ago
So if they would say "you can only install it on cloud A" (like a lot of cloud software, such as Spanner), it would make it better?
255.
9.9 Linux CVE(threadreaderapp.com)
8 points by Palmik 2y ago|2 comments
256.
by Palmik 2y ago
This complaint seems hypocritical. Microsoft doesn't have to offer their software on other clouds at all. When will I be able to get Spanner on Azure or AWS or my own hardware?
257.
by Palmik 2y ago
YouTube is clearly boosting "positive" comments. I am not sure it's a good solution if you care about other people's opinions.
258.
by Palmik 2y ago
Is this article just a way to advertise and normalize their investments in this industry? They have done these kind of articles a few times in the past, so one has to wonder.
259.
by Palmik 2y ago
Except doing continued pre-training or fine tuning of the released model weights is the same process through which the original weights were created in the first place. There's no reverse engineering required. Meta engineers working o
260.
by Palmik 2y ago
That's a bad analogy. The weights are much closer to source code, because you can directly modify them (fine tune, merge or otherwise) using open source software that Meta released (torchtune, but there are tons of other libraries and
261.
by Palmik 2y ago
Given that the focus is performance, do you have any benchmarks to compare against the likes of TensoRT-LLM.
262.
by Palmik 2y ago
VRAM per GPU isn't such an interesting metric. If it was, everyone would be fine tuning on A100 80gb :) What matters is steps per $ and to some degree also speed (I'm happy to pay premium sometimes to get the fine tuning results f
263.
by Palmik 2y ago
Tell me more. At what scale were you able to get a significant discount and how much? Most people will be (full) fine tuning on 8xh100 or 16xh100 for few days at a time.
264.
by Palmik 2y ago
> For example, training and serving Llama 3.1 on Google TPUs is about 30% cheaper than NVIDIA GPUs When you say this, you should specify which Nvidia GPU you mean (I assume h100 SXM) and that price you are assuming for such GPU. One can&
265.
by Palmik 2y ago
For one time payments, like in e-commerce, it's relatively trivial for even a small business to implement. For subscriptions, achieving portability is much trickier.
266.
by Palmik 2y ago
The difference between (A) software engineers reacting to AI models and systems for programming and (B) artists (whether it's painters, musicians or otherwise) reacting to AI models for generating images, music, etc. is very interestin
267.
We have reached an agreement with the US AI Safety Institute(twitter.com)
1 points by Palmik 2y ago|0 comments
268.
by Palmik 2y ago
Because to self host it, you'd need to ensure you're PCI compliant. That works for established companies I guess, but not so much when starting out. Payment provider lock-in in a scary thing, when they can cancel your account at m
269.
by Palmik 2y ago
Curious to see if the same will apply to other materials like news, books, images, music, movies, etc.
270.
by Palmik 2y ago
The headlines on the ruling can be misleading: > Plaintiffs may have succeeded if they were instead seeking damages for past harms. But in her opinion, Justice Amy Coney Barrett wrote that partly because the Biden administration seemingl

More ›