vitaelabitur

8 ms·

31.
by vitaelabitur 9mo ago
Aren't LLMs just super-powerful pattern matchers? And guessing "taps" a pattern recognition task? I am struggling to understand how your experiment relates to intelligence in any way. Also, commercial LLMs generally have syst
32.
Some programming languages worth learning(codecrafters.io)
3 points by vitaelabitur 9mo ago|0 comments
33.
The Case of a Curious SQL Query(buttondown.com)
3 points by vitaelabitur 9mo ago|0 comments
34.
NASA used genetic algorithms to design an antenna (2003) [pdf](ntrs.nasa.gov)
2 points by vitaelabitur 10mo ago|0 comments
35.
by vitaelabitur 10mo ago
I went home for holidays last month. One day, my mom had a complaint about her food delivery and raised a ticket in the app. She was assigned "someone" on chat, and she carefully typed her issue. Then, she got a call from the same
36.
by vitaelabitur 10mo ago
I might try this if you link or briefly explain the supporting research on your landing page.
37.
Recovering depth from images using Markov Random Fields(nghiaho.com)
2 points by vitaelabitur 10mo ago|0 comments
38.
Rendering graphics in 500 lines of bare C++(haqr.eu)
10 points by vitaelabitur 10mo ago|2 comments
39.
by vitaelabitur 10mo ago
Yeah, but I tried switching to minified JSON on a semantic labelling task and saw a ~5% accuracy drop. I suspect this happened because most of the pre-training corpus was pretty-printed JSON, and the LLM was forced to derail from likely pat
40.
by vitaelabitur 10mo ago
I tokenized these and they seem to use around 20% less tokens than the original JSONs. Which makes me think a schema like this might optimize latency and costs in constrained LLM decoding. I know that LLMs are very familiar with JSON, and c
41.
by vitaelabitur 10mo ago
Guilty. I was clearly influenced by the prose. Can we simply let bygones be bygones?
42.
by vitaelabitur 10mo ago
It's ironic I've been waiting for smart glasses with displays ever since I found them in books and films as a kid, but now I see them as artifacts signaling a dystopian future. That they are tied to Meta does not help.
43.
by vitaelabitur 10mo ago
I feel a lot of us are starting to reject new tech that seeks to aggressively deepen our digital immersion.
44.
by vitaelabitur 10mo ago
I beg to disagree. I believe in the Orwellian rules religiously, but I like indulging in maximalism from time to time. It felt apt that the article uses a dense, historical style to help it evoke a part of cultural history bygone. That said
45.
by vitaelabitur 10mo ago
I wouldn't be surprised to see all of us deeply fetishizing the analog experience in the coming years. Activities like letter writing might get gentrified, with private businesses charging a premium for delivery.
46.
by vitaelabitur 10mo ago
What does your tool/service do? Write a blog on "how to do XYZ". Use your personal blog, website, or guest post on Medium, etc. Be helpful. Share an easy and free way to do it without using your tool/service. Then introd
47.
by vitaelabitur 10mo ago
That might be the case. Maybe I am writing from a place of nostalgia and subconscious declinism.
48.
by vitaelabitur 10mo ago
You are right, parents can still steer their children. But in some cases like mine, parents might not have the time or inclination to do so, for whatever reasons. I was able to discover Pickwick Papers, Hardy Boys, Harry Potter, Lord of the
49.
by vitaelabitur 10mo ago
https://www.youtube.com/watch?v=ba5HllbvLf4 , The Many Adventures of Winnie the Pooh (1977) Ending I was a quiet kid. Books, shows, and films shaped my sensibilities and moral tastes. I wonder if that remains possible today.
50.
by vitaelabitur 10mo ago
Makes sense. The default assumption is indeed shifting. Today, you are guilty of using AI until proven innocent. Students and some artists might certainly benefit from what you demoed. This is interesting, I'll play the devil's ad
51.
Project ideas to appreciate the art of programming(codecrafters.io)
224 points by vitaelabitur 10mo ago|79 comments
52.
by vitaelabitur 10mo ago
I am struggling to understand the need for this. If AI output is indistinguishable, that could mean - 1. It is as good as human output. For this to happen with the current state of AI, a human still needs to be an expert to guide and review
53.
by vitaelabitur 10mo ago
Here's what has worked for me on day 0: What problem does your tool solve? Write a blog on "how to solve/do XYZ". Share an easy and free way to solve the problem without using your tool. Then introduce your tool and shar