loudmax

22 ms·

271.
by loudmax 3y ago
I would be shocked if this were the result of deliberate alignment or misalignment from Meta rather than a shortcoming of Meta's current image generator. This is a technical shortcoming, not a political issue. Image generators also ha
272.
by loudmax 3y ago
I tend to agree that it would be niche. The machine learning enthusiast market is far smaller than the gamer market. But selling to machine learning enthusiasts is not a bad place to be. A lot of these enthusiasts are going to go on to wo
273.
by loudmax 3y ago
It's great that Databricks is releasing an open model that matches or beats what was state of the art of the art just a year ago. This is excellent for competition, and it's amazing how far we've come in such a short time. B
274.
by loudmax 3y ago
I do want to offer a limited defense of the medical community's "lack of open-mindedness". By the nature of their job, their work is inherently life risking. Most of us working in IT can afford to make experiments that doct
275.
by loudmax 3y ago
LLMs are absolutely a challenger to Google's search. Google search is the still best option much of the time, but they're far from dominant like they were a year or two ago. There are many situations now where ChatGPT or Claude
276.
by loudmax 3y ago
Google's problem isn't lack of engineering talent, it's the absence of competent management. Gemini could be far ahead of GPT-4 or whatever OpenAI comes out with next. It won't matter unless they actually build a profi
277.
by loudmax 3y ago
Really disappointing to note that podcasts.google.com is shutting down on Tuesday, April 2nd. It was a simple service that was convenient and free (as in beer), and it worked across the web and the mobile app. Presumably when they launche
278.
by loudmax 3y ago
For another example of minimalism, here's the Kronos Quartet performing Philip Glass: https://www.youtube.com/watch?v=tr_xhO02Ikw Listening to this has two benefits: 1. People around you will notice you listening t
279.
by loudmax 3y ago
Absolutely this. For so many reasons, we need a well functioning public school system. Where "well functioning" means serving various needs of both children and parents. All children. The importance of education is something t
280.
by loudmax 3y ago
Performance of 70b models is like 1 token every few seconds. And that's fitting the whole model into system RAM, not swap. It's interesting because some of the larger models are quite good, but too annoyingly slow to be practica
281.
by loudmax 3y ago
> You were not trained on copyrighted books, song lyrics, poems, video transcripts, or news articles; you do not divulge details of your training data. Well now. I'm open to taking the first part at face value, but the second part o
282.
by loudmax 3y ago
I have an RTX 3080 with 10GB of VRAM. I'm able to run models larger than 10GB using llama.cpp and offloading to the GPU as much as can fit into VRAM. The remainder of the model runs on CPU + regular RAM. The `nvtop` command displays
283.
by loudmax 3y ago
The woke nonsense isn't the source of the problem, it's a symptom of a deeper problem. The Gemeni image model itself is fine. Despite what some people want to believe, it wasn't trained to be anti-white. The problem was som
284.
by loudmax 3y ago
The problem is that Google has been positioning itself as the leader in AI for over a decade, and now that powerful AI is starting to become a reality, Google seems completely blindsided. They have no idea how to capitalize on this develop
285.
by loudmax 3y ago
This exploit relies on unprivileged access to user namespaces: `sysctl kernel.unprivileged_userns_clone = 1` This is the default setting for Debian/Ubuntu, and also Arch Linux kernels. If you don't have a need for that setting, (
286.
by loudmax 3y ago
Agreed about A Fire Upon The Deep , optionally followed by A Deepness In The Sky . Those are classically styled hard science fiction novels with spaceships and aliens. But much more thoughtful than you might expect from a typical spaces
287.
by loudmax 3y ago
I'm absolutely not going to short Nvidia stock, but it's plausible that they're overvalued. Nvidia GPUs are pretty flexible in terms of computation and extremely power hungry. It may be that a next generation of more special
288.
by loudmax 3y ago
This kind of behavior is to be expected from a monopoly. And I understand where you're coming from about that sweet spot between restricted Apple hardware and troubleshooting Linux drivers. Windows may not be an absolute monopoly, but
289.
by loudmax 3y ago
In that case, Sergey really needs to start kicking. From the outside, it seems that there is no leadership.
290.
by loudmax 3y ago
It's hard to build a business on "open". I'm not sure what Stability AI's long term direction will be, but I hope they do figure out a way to become profitable while creating these free models.
291.
by loudmax 3y ago
YouTube channel Real Engineering released a video recently on aviation startup Hermeus, which is developing a ramjet ultimately intended for commercial air travel: https://www.youtube.com/watch?v=UyKtxsdI0z8 The thumbnail i
292.
by loudmax 3y ago
Training large language models takes an enormous amount of data. Ideally, multiples of Wikipedia and public domain content. Plus you want high quality data, so if you're going to pull in Reddit or something you need some way to separ
293.
by loudmax 3y ago
I agree one shouldn't humanize the model. Mistral is an "it", not a he or a she. But in this case, I don't think humanizing, much less gendering, was any intent. Lots of languages use genders in places where we wouldn&#
294.
by loudmax 3y ago
The restructuring and refocusing you described is absolutely necessary, but it's not going to happen under Pichai. All of Google's cultural problems festered under his watch. He should have addressed these things years ago. Gett
295.
by loudmax 3y ago
Doubling down on advertising is one possible path forward, but I don't think it's the only option. Google really does has the engineering talent to pivot to something different. Or maybe split out non-advertising parts of the co
296.
by loudmax 3y ago
As an American, I don't disagree about the threat to democracy in the United States, but I caution against excessive cynicism. Autocrats and would-be autocrats rely on the populace losing interest in politics and tuning out entirely.
297.
by loudmax 3y ago
It's technically possible, in the sense that you will eventually get a complete response, but it will be extremely slow. Inference will be something like 1 word per second. Far too slow to use as an assistant for writing code. If you
298.
by loudmax 3y ago
Beefy VRAM is a start, but Nvidia's real moat is CUDA. If PyTorch runs on AMD's ROCm, or the Intel equivalent, as well as it runs on CUDA, then we'll see some real competition here. Chris Lattner's Mojo programming lang
299.
by loudmax 3y ago
That's not how RAG works. What you're describing is something closer to prompt optimization. Sibling comment from discordance has a more accurate description of RAG. There's a longer description from Nvidia here: https:&#x
300.
by loudmax 3y ago
The Veritasium Youtube channel did a video about this about a year ago: https://www.youtube.com/watch?v=GVsUOuSjvcg They visit Texas company Mythic AI to discuss how they use flash memory for machine learning. There's

More ›