llm_trw

7 ms·

31.
by llm_trw 2y ago
The point is that you can force the respective of a giant or an ant if you so choose. Using two cameras I had a time lapse of a fog coming in where you could see the 3d structure becuase they were at two different streets looking out. Like
32.
by llm_trw 2y ago
And people obscure this fact by talking about spending controlled by XYZ instead of total federal spending. Who cares that it's an act of congress that funds a given project instead of a regular budget? Your money is still being spent.
33.
by llm_trw 2y ago
If I was sending it as a business email I'd edit it before sending it off. But the first draft saved me between 10 to 30 minutes of trying to get out of the headspace of "this is a fucking disaster and I need to look for a new job
34.
by llm_trw 2y ago
>The second incorrect method to save a matrix of embeddings to disk is to save it as a Python pickle object [...] But it comes with two major caveats: pickled files are a massive security risk as they can execute arbitrary code, and the
35.
by llm_trw 2y ago
It is heaven. With a well working rag system you can find the reason why any decision was made so long as it was documented at some point somewhere. The old share point drives with a billion unstructured word documents starting from the 199
36.
by llm_trw 2y ago
No. I'd already spend 30 Min looking at how to solve it myself. The search on perplexity was a hail marry before I started calling handymen.
37.
by llm_trw 2y ago
Linus Trovalds: bottom 20% coder. https://www.youtube.com/watch?v=VHHT6W-N0ak&pp=ygUMTGludXMgQ...
38.
by llm_trw 2y ago
I fixed the hinge in my oven by giving perplexity.com the make and problem. I saved an hour on the phone calling people to organise a visit some time in the next week. Maybe you should stop using the Ai slop tools that don't work?
39.
by llm_trw 2y ago
Counterpoint: the PC. The minicomputer revolution was meant to replace all the big iron in the basement with computers shared by teams. This lasted a decade until we got computers meant for one person and we've used that format of comp
40.
by llm_trw 2y ago
Man if we only had some type of markdown meant for machines to understand, that was specifically designed to handle arbitrary information nesting and tagging our lives would be so much better now. We could have called it something like exte
41.
by llm_trw 2y ago
The best argument against the EU is a 5 minute conversation with anyone who supports it.
42.
by llm_trw 2y ago
That you (and apparently the German bureaucracy) think people who can start a business and people who can tick administrative boxes are the same type of person explains German economic performance in the last 5 years.
43.
by llm_trw 2y ago
How long did it take you from start to finish? In Australia, Canada and the US I can open a company in about 30 minutes from picking the name to opening a bank account I can get money into. In Germany the last time I tried it was about 7 we
44.
by llm_trw 2y ago
There is a huge gap in the market, I'd say between capital investment of $50,000 and $5,000,000, where a ton of ideas which can make 20%+ ROI per year once mature live. But because they will never hyper scale they can't be finance
45.
by llm_trw 2y ago
Try opening a business in Germany to find out why.
46.
by llm_trw 2y ago
Because is the bitter lesson were true no one would be wasting their time with convolutions or attention blocks. You'd just replace them with the general tensor that allows every hyper relation possible between all points instead.
47.
by llm_trw 2y ago
Teenage me who wanted information to be free didn't imagine this would be the result. It's like everything I wanted to happen in the 00s did, but from the monkeys paw. The only way I can think of fixing this is by giving rights to
48.
by llm_trw 2y ago
It was the only place where you could purchase the ebook.
49.
by llm_trw 2y ago
I contacted a well known author about the shit latex rendering of his ebooks on Amazon. He said sorry and send me the a pdf copy he build with my name on all the pages. I really like the fact that I have a book dedicated to me by the author
50.
by llm_trw 2y ago
The torrent sites is how meta and friends get their books. They are safe.
51.
by llm_trw 2y ago
I know they do good stuff, it's just the fire hose of bullshit is a lot wider than the tap of innovation.
52.
by llm_trw 2y ago
I don't know. Google has some great papers but a lot more hot air. The best thing to do when you see a google paper is ignore it until someone you trust more writes a second paper about it.
53.
by llm_trw 2y ago
Well it was a nice run. I expect all LLMs to be illegal within a year if this is the sort of high caliber defense the top minds of meta can come up with.
54.
by llm_trw 2y ago
It's a tough one, I've never seen a book that actually covers the _bare_ minimum of the maths you need for ML. The little learner comes close but I'd only really suggest that to people who already know the maths because the
55.
by llm_trw 2y ago
I mean yes, but without knowing the maths then knowing how to optimize the maths is a bit useless? At the very least you should know enough linear algebra that you understand scalar, vector and matrix operations against each of the others.
56.
by llm_trw 2y ago
The only reason why suburbia is possible is because it is _heavily_ subsidized. If you had to pay the real bill for road maintenance alone suburbs would no longer be viable. So the suburbs take from the commons and don't give back in y
57.
by llm_trw 2y ago
>Interesting, how are you remixing the order of questions? If we're talking about an academic benchmark like MMLU, the questions are independent of one another. Unless you're generating multiple answers in one go? Short version
58.
by llm_trw 2y ago
>This brings us to our current limitations. Right now, DeepEval’s primary evaluation method is LLM-as-a-judge. We use techniques such as GEval and question-answer generation to improve reliability, but these methods can still be inconsis
59.
by llm_trw 2y ago
Which is my point. You need a ton of specialized knowledge to use compute effectively. If we had infinite memory and infinite compute we'd just throw every problem of length n to a tensor of size R^(n^n). The issue is that we don'
60.
by llm_trw 2y ago
I am in correspondence with a number of worth while authors, it's just that there isn't any place where they congregate in the (semi) open and without the weirdos who do stuff with the models you're missing out on a lot. My f

More ›