leobg

12 ms·

211.
by leobg 1y ago
https://memrussian.com/ ?
212.
by leobg 1y ago
I’d say reading is fine tuning. It changes your weights. That’s the whole point of doing it.
213.
by leobg 2y ago
This is a verbatim quote from Jason Calacanis.
214.
by leobg 2y ago
Nobody is copying their model parameters or inference code. What people “suck out” of their API are the general ideas. And they do it specifically so they can reassemble them in their own way. It’s like reading all the Jack Reacher novels a
215.
by leobg 2y ago
OpenAI is taking the position similar to that if you sell a cook book, people are not allowed to teach the recipes to their kids, or make better versions of them. That is absurd. Copyright law is designed to strike a balance between two iss
216.
by leobg 2y ago
I’d be interested in hearing more about how to “abuse” compression algorithms to detect outliers in datasets or to filter out repetitions. For example, LLMs or Whisper sometimes get stuck in a loop, repeating a word, a group of words, or wh
217.
by leobg 2y ago
Good question. When BF Skinner used to train his pigeons, he’d initially reinforce any tiny movement that at least went in the right direction. For the exact reasons you mentioned. For example, instead of waiting for the pigeon to peck the
218.
by leobg 2y ago
The whole point of RLHF is to make up for the fact that there is no loss function for a good answer in terms of token ids or their order. A good answer can come in many different forms and shapes. That’s why all those models fine tuned on (
219.
by leobg 2y ago
After skimming the landing page, I still have no idea what this does. Generate animated videos based on my notes and sketches maybe? If so, how would that help me learn? And learn what? Is this for physics classes? Language learning? Learni
220.
by leobg 2y ago
Most Pardons: Clinton & Reagan (4/month). Fewest Pardons: Bush senior / Biden (1.5/month) Since Reagan.
221.
by leobg 2y ago
This should be a terminal utility. xx ffmpeg video1.mp4 normalize audio without reencoding video to video2.mp4 And have sensible defaults. Like auto generating the output file name if it’s missing, and defaulting to first showing th
222.
by leobg 2y ago
If by LLM you mean embeddings I agree. Though you can often get away with using much smaller models for that. I was talking about people who actually make a call to a completion endpoint and then have the LLM repeat the input text token for
223.
by leobg 2y ago
They can actually survive if you let them. Here’s a video of a guy filming himself being stung by a bee, then watching as the bee spins around, works its stinger loose, and flies away: https://youtu.be/G-C77ujnLZo
224.
by leobg 2y ago
There are AI bros that will call an LLM to do what you could do with a regex. I’ve seen people do the chunking for RAG using an LLM…
225.
by leobg 2y ago
Made this for my kids for Christmas: - take an ebook in any language - AI translates it to German - AI speaks it using the voice of their fav narrator - a UI showing the text as it is being read Now they can read Asimov, Kulansky, Bryson, r
226.
by leobg 2y ago
iOS Voice Dream Reader. First app I install on a new iPhone since 2010 I believe. I will even cut and scan physical books just so I can read them in the app. The story of the guy who made it is also interesting!
227.
by leobg 2y ago
Someone who remembers good old Skinner. Cool!
228.
by leobg 2y ago
It seems to me that you have no idea what the word fraud means. It’s a legal term. It has nothing to do with whether you like or agree with something.
229.
by leobg 2y ago
Good point with the HTML in the training data. I always found the “structured output feature” thing odd. LLMs will follow any structure present in the prompt. So with a few few-shot examples, you have always been able to make them do anythi
230.
by leobg 2y ago
That’s meta. Literally. Edit: Sorry. This was based on the false assumption that this was research by Meta, Inc..
231.
by leobg 2y ago
I also find the conclusions odd. Could it be that for the rich people, Wikipedia is just entertainment, whereas people in poorer countries actually seek to understand a topic in order to solve a problem?
232.
by leobg 2y ago
Context: We do pay ~0.20 Euros of taxes per kWh. You have to add that to the actual power price. (And we do pay tax on income before we use that income to pay for power to our homes.)
233.
by leobg 2y ago
He has been conditio sine qua non. For ChatGPT, too, arguably (impetus for founding of OpenAI as a counter balance to Google, hiring of Ilya Sutskever, funding, all-in position on Andrej’s belief in the superiority unsupervised learning). A
234.
by leobg 2y ago
What makes you think he isn’t just messing with people here? He has said before that he’s an alien or that he’ll guard Taylor Swift’s cats with his life. Why should this one be different?
235.
by leobg 2y ago
What country are you in? I’m in DE and have filed a criminal complaint about a company that runs fake personal ads targeting the elderly: When an elderly person calls, they schedule an appointment at the person’s home. Then, over several ho
236.
by leobg 2y ago
The “stripes” in the keogram look very regular. A few days of clear skies followed by a few days of clouds. Is that what’s happening? Why the regularity?
237.
by leobg 2y ago
https://solarcan.co.uk/
238.
by leobg 2y ago
Isn’t that what all the dating apps have been doing for decades?
239.
by leobg 2y ago
I suggest you read “The Kindly Ones” by Littell. Or, for a smaller investment of time, “Just Revenge" by Dershowitz. Putin is not even on the same axis as Hitler in terms of evil. I can only take your words to mean that you have no ide
240.
by leobg 2y ago
Love this: > VW Group proved their incompetence with regards to vehicle data and everything people accused Tesla of seems to happen at VW Group.

More ›