leobg

35 ms·

601.
by leobg 3y ago
I index 40M paragraphs of legal text, bm25 and vector similarity search, at < 200ms query time, on a single $80/month Hetzner server. Email in profile if you’d like to talk.
602.
by leobg 3y ago
Not in SF?
603.
by leobg 3y ago
iOS app that uses the camera to check if my toddler’s eyes are closing. And if so, turns on the flashlight and plays a nasty sound. We used it in the car when having her fall asleep before arriving would have caused problems (having to wake
604.
by leobg 3y ago
I would’ve been interested to learn how much it costs to train these models using their platform. Like, a 70b model - are we talking millions of dollars here?
605.
by leobg 3y ago
See sibling comment.
606.
by leobg 3y ago
Can’t share the model, since it was trained for a client. I don’t know if any public datasets exist. But Mistral will learn what you throw at it. So if you build a dataset of chat conversations that contains, say, answers in the form of {“a
607.
by leobg 3y ago
You can fine tune an LLM to output a particular token for category X. This way, you get 1) fast inference (just need to generate a single token) and 2) confidence metrics (just use the logprobs for the token that was chosen). Gets me to 93%
608.
by leobg 3y ago
Defendant: “My actions were pre-determined. I couldn’t do anything about it.” Judge: “The same applies to me convicting you.”
609.
by leobg 3y ago
I love Mistral. It’s crazy what can be done with this small model and 2 hours of fine tuning. Chatbot with function calling? Check. 90 +% accuracy multi label classifier, even when you only have 15 examples for each label? Check. Craaaazy p
610.
by leobg 3y ago
In what sense does Altman belong in the same sentence as Musk? I’m completely puzzled.
611.
by leobg 3y ago
> Mr. Musk explained that his plan was to colonize Mars to escape overpopulation and other dangers on Earth. I don’t think Musk has ever been worried about overpopulation. The opposite is true.
612.
by leobg 3y ago
He’s much better at giving it away when he thinks he is investing. Stuff like shorting Tesla. Hence, some traders have started to follow something called the Inverse Gates Index.
613.
by leobg 3y ago
Elon is the owner. Altman is a hired CEO.
614.
by leobg 3y ago
You know that Hass is German for “hate”? I always find it funny that I didn’t bother to change the name here.
615.
by leobg 3y ago
Not OP, but I asked myself that same question two years ago. Then I looked at the energy prices in Germany and knew I had no chance against cloud GPUs. Maybe you live in a country with lower energy prices, like Bermuda (or any other country
616.
by leobg 3y ago
In Germany, it is literally the authorities and conscientious journalist, who live in fear of being sued by the criminals, rather than the criminals being in fear of being sued themselves. It is agonizing to watch.
617.
by leobg 3y ago
I at first thought that by Arsenal of Democracy they meant an educated populace that actually felt some sense of responsibility towards both the generations that came before and those that will come after. Minds that have the capacity to po
618.
by leobg 3y ago
"No one person should be trusted here. I don’t have super voting shares. The board can fire me. I think that’s important." Sam Altman, June 2023 (Haha)
619.
by leobg 3y ago
Being called “a gem of a human being” by someone like Altman would be an insult to anyone. If it wasn’t just that one more flat out lie, spoken with zero conscience, that it is.
620.
by leobg 3y ago
Same. Never understood whether they’re buttons or just sound generators for blind people.
621.
by leobg 3y ago
The dead have estates.
622.
by leobg 3y ago
Holding a “State of X address” in order to position yourself as an authority used to be the popular thing for internet marketing crooks back in the 2010s. I guess it works.
623.
by leobg 3y ago
> Two years ago, Ekobot announced a collaboration with Swedish telecom company Telia that led to the integration of 5G mobile technology into the robot, which lets it communicate remotely with a central server to share collected learning
624.
by leobg 3y ago
Wow. What a refreshing directness: > He was a swine of a man and a jabbering dupe of a president. Nixon was so crooked that he needed servants to help him screw his pants on every morning. Even his funeral was illegal. He was queer in th
625.
by leobg 3y ago
Maybe this is what Altman was less than candid about. That the speed up was bought by throwing RAG into the mix. Finding an answer is easier than generating one from scratch. I don’t know if this is true. But I haven’t seen an LLM spit out
626.
by leobg 3y ago
Was that not precisely @Suhail’s business plan? I hear they have since pivoted to AI image gen.
627.
by leobg 3y ago
> over five percent of the output ChatGPT emits is a direct verbatim 50-token-in-a-row copy from its training dataset I don’t think this is typical behavior of LLMs. This is more typical behavior for retrieval augmented generation (RAG).
628.
by leobg 3y ago
I had a lot of fun, having ChatGPT generate a fake diary of, say, Adolf Hitler for certain dates in history, as well as time machining him into the present. I just like being able to pick up a piece of history and turn it around and look at
629.
by leobg 3y ago
Very interesting. I’m a lawyer and using sequential IDs in a fraud case right now, to determine the number of victims. Unfortunately, so far, I only have the IDs of two victims, and those are from just within about a month, whereas the frau
630.
by leobg 3y ago
Book recommendation (a novel that plays out your thought experiment): One Second After is a 2009 novel by American writer William R. Forstchen. The novel deals with an unexpected electromagnetic pulse attack on the United States as it affec

More ›