amelius

15 ms·

451.
by amelius 2mo ago
> The most private AI is the one running on my own hardware, not in some giant data center. I want that too, but you gotta ask yourself the question how efficient that is compared to running it in a datacenter shared with everybody else.
452.
by amelius 2mo ago
I also drink a lot of coffee and I really hate it when I'm visiting a client and they offer me only 1 or 2 cups of coffee during the day!
453.
by amelius 2mo ago
We need to get commerce out of social media. Anything "social" should not have to do with commerce. There are no billboards in a church, so why should there be ads in places where people socialize? Having ads on sports fields is b
454.
by amelius 2mo ago
Why is floating point faster?
455.
by amelius 2mo ago
You can already do this with VNC or Xpra (the latter is a screen/tmux for graphical applications).
456.
by amelius 2mo ago
Working in AI was once reserved for intellectually gifted people. But with the recent advances in compute performance, working in AI today is basically not much different from pounding on a black box as long and as hard as necessary to make
457.
by amelius 2mo ago
Give me the freakin bit error rate please.
458.
by amelius 2mo ago
> inputs, process, or outputs that require meatbag intelligence Yeah but only because a meatbag is going to use the product :)
459.
by amelius 2mo ago
Apple.com looks ok. It doesn't have the cookie banner.
460.
by amelius 2mo ago
Yes, I was assuming they'd use external tools. Using only the raw LLM doesn't sound like a good strategy. Considering that agents are not a new concept, why isn't this a solved problem by now?
461.
by amelius 2mo ago
Then they shouldn't store the data as embeddings. Instead: use an LLM to build a large (old-school) database of products with all their specifications. The LLM can also build the schema for that database as it finds more data. Then use
462.
by amelius 2mo ago
An LLM can read websites, right? And turn them into structured data.
463.
by amelius 2mo ago
Can anyone explain why LLMs are so bad at finding products (their webpages) with given specifications? You'd think they would have solved it by now.
464.
by amelius 2mo ago
I was hoping for a blog about things I don't understand yet.
465.
by amelius 2mo ago
Instead of watermarking you could add a cryptographic signature.
466.
by amelius 2mo ago
How does that watermark feature work if you can just point the camera at an AI-generated image?
467.
by amelius 2mo ago
The feature I want most is NOT sending a notification to the app whenever I take a goddamn screenshot.
468.
by amelius 2mo ago
Not a fan of N=1 experiments.
469.
by amelius 2mo ago
Why not if they are working in your interest and produce open source code.
470.
by amelius 2mo ago
EU should heavily subsidize Mozilla, so that this is not necessary.
471.
by amelius 2mo ago
How many mankinds did he create? I cannot imagine he created just one and then picked another hobby.
472.
by amelius 2mo ago
At this point I want to see some human-drawn pelicans on bicycles. I suspect the LLMs aren't doing all that bad.
473.
by amelius 2mo ago
I'm an IT professional and noticed that memory went scarce from looking at the shelves.
474.
by amelius 2mo ago
Think of it as looking at your screen through AR goggles, where the goggles modify the image before it reaches your eyes. Everything happens at the image level so there's no way Meta can obfuscate the DOM to make ads pass through.
475.
by amelius 2mo ago
I didn't understand your use case, so it could also be the way you write your prompt, I suppose ...
476.
by amelius 2mo ago
You can still do that. The only thing I'm proposing is to do the filtering on graphical data (what the user actually sees) rather than on the DOM (which can be obfuscated; and which may change every week).
477.
by amelius 2mo ago
Maybe a better approach is to convert the page to an image, run OCR over it, use an LLM to remove the ads, present the result to the user.
478.
by amelius 2mo ago
You forgot the lines necessary for downloading the dependencies, and the dependencies of dependencies, etc. (and compiling and installing those dependencies)
479.
by amelius 2mo ago
How did it work out for the pastafarians?
480.
by amelius 2mo ago
Nvidia's great superpower is flexibility. You can easily run models of very different types on the same card; and their hardware is great for R&D. However, at some point AI may be good enough for most people and then it makes sense

More ›