imustachyou

Discussion on HN https://news.ycombinator.com/item?id=44576352

I’m missing something, won’t the input to the llm necessarily be plaintext? And the output too? Then, as long as the llm has logs, the real input by users will be available somewhere in their servers

Very fun! Would love to see a tier list to start checking off myself, and maybe a caption for the pictures

Reminiscent of On Kawara's "Date Paintings", where he painted each day's date in the local date format, every day for 48 years.

https://www.phaidon.com/agenda/art/articles/2014/july/14/on-...

What was the moment, if you don’t mind sharing?

He also left $200M for the Summer Science Program. https://www.forbes.com/sites/marybethgasman/2023/10/12/200-m...

After the Apple Vision announcement today, I was reminded of this short story by Rich Larson. Not so dystopian now

S4 and its class of state-space models are an impressive mathematical and signal-processing innovation, and I thought it was awesome how they destroyed previous baselines for long-range tasks.

Have there been any state-space models adapted for arbitrary text generation?

Language models like ChatGPT are trained to predict new words based on the previous ones and are excellent for generation, a harder task than translation or classification. I'm doubtful about the adaptability of text models that deal with fixed-sized input/outputs and don't have an architecture that is as natural for generating indefinitely long sequences.

Based on the 1953 short story by Arthur C. Clarke

https://urbigenous.net/library/nine_billion_names_of_god.htm...