leobg

47 ms·

901.
by leobg 4y ago
He made references to 42 long before 420 ever came up. Certainly long before the weed smoking Joe Rogan episode. The 420 joke is a reference to his tweets trying to take Tesla private. 42 wasn’t in the right range as a target price. But 420
902.
by leobg 4y ago
You're not the only one. It's the original blog author who made that mistake. Now his article is on archive.org and he can't edit anymore either. :)
903.
by leobg 4y ago
Facepalm. I hate typing on my phone. 1880s, of course. Thanks!
904.
by leobg 4y ago
In the 1970s, people moved to Ashrams in India to lose their ego. In the 2020s, people are anxious for AI to conserve it beyond death. Quite a generational pendulum swing… :)
905.
by leobg 4y ago
That’s precisely the funny thing with copyright: If you create a work where you can clearly tell what the source was for your inspiration because you stole from another source, it’s a violation of copyright. But if you create a work and you
906.
by leobg 4y ago
What’s your favorite dish/recipe?
907.
by leobg 4y ago
Exactly. And they get away with it precisely because we believe in this wrong notion of freedom. We’re still stuck at the 2nd stage of the three transformations that Nietzsche outlined back in the 1980s. Which is essentially to think we’re
908.
by leobg 4y ago
The idea copyright is to protect the author’s specific personality in the expression of an idea while, at the same time, allowing for the free circulation of the idea itself. An ML model has no personality. It is not a legal entity. Also, i
909.
by leobg 4y ago
“Freedom is not the ability to get what you want. It’s the ability to want what is good for you.”
910.
by leobg 4y ago
You you noticed the duplicate word?
911.
by leobg 4y ago
“It’s not the drones that kill, it’s the users.” I bet that’s what they tell themselves. (“And besides, the people getting killed are in the wrong anyway. And there’s the unseen good of the lives saved by these drones.” Continue this list f
912.
by leobg 4y ago
So, HN folks who bought one during the pandemic: Are you still using it?
913.
by leobg 4y ago
“Chat mode” just means that 1) you structure the prompt given to the model for completion as a chat, i.e. a list of lives in the form “NAME: UTTERANCE” and 2) specify as “end” sequence the “NAME:” part for the user (i.e. prevent the model f
914.
by leobg 4y ago
Maybe one day you’ll be able to tell ChatGPT what kind of model you need and it’ll automatically select the right architecture, gather the training data, and commission the training using the cheapest and/or fastest provider. :)
915.
by leobg 4y ago
The zero shot ability is convenient. But for tasks that you need to get done millions of times, I’d much rather spend $10 on GPU compute and maybe a day of training data generation to train a T5 which I then “own”. Also, running your own sp
916.
by leobg 4y ago
I've got 2 questions: 1) Can I run this, say, on Colab? I see no code. 2) Any interactive open source tool to visually explore the latent space of a model’s representations for any given prompt, or for faces, buildings, etc.? Like they
917.
by leobg 4y ago
Hitchhiker’s Guide to the Galaxy is one of his favorite books. The 420 is a reference to “42”being the answer to the meaning of life. Nothing to do with weed.
918.
by leobg 4y ago
sbert.net is all you need.
919.
by leobg 4y ago
What an interesting thought: What makes us intelligent beyond machines is our time spent silently and introspectively thinking and dreaming in the absence of outside prompts? If this is true, then we are certainly getting closer to an AI th
920.
by leobg 4y ago
Text to speech output, please. Google WaveNet voices. With a toggle to turn auto-reading of responses off or on. Add serpapi on top of that, and you might just become my phone’s most used app.
921.
by leobg 4y ago
If interested in consulting, perhaps you could shoot me an email. Address is in my profile. I may actually need exactly your expertise right now.
922.
by leobg 4y ago
I'd like to post a TD;DR, but there are just too many streams of information, and I can't process complex topics like that this fast...
923.
by leobg 4y ago
That’s what OpenAI’s labelers apparently thought, too. No human preference there, it seems :)
924.
by leobg 4y ago
I asked it yesterday to give me a regex that matches the numbers 1 to 365. And it literally gave me that response about hate speech.
925.
by leobg 4y ago
From the article: > Again, we do not know what really goes on in the mind of a starling
926.
by leobg 4y ago
Unique just means there is no exact duplicate. Since this already applies to almost anything as simple as a snowflake, such a claim is, on the contrary, a truism.
927.
by leobg 4y ago
I have, but haven’t bothered to publish it. POC to use by me and my daughter. Works like Siri. Fun to play with. But the hallucinated responses made us stop using it pretty quickly. Needs more work to be of any use (integrate search engines
928.
by leobg 4y ago
I recommend Google Text To Speech. WaveNet voices are pretty good.
929.
by leobg 4y ago
I know. I’m saying it may not have anything to do with an “image” at all. More with “composing oneself”. But I haven’t looked up the etymology.
930.
by leobg 4y ago
“Bild” is indeed “image”. Though one also says “bilden” to mean “generate”, “create”, or “consist of”. For instance, one might us it to say that two people “bilden” a team (i.e. make up a team).

More ›