←back to thread

Show HN: Gemini Pro 3 imagines the HN front page 10 years from now

(dosaygo-studio.github.io)
3338 points keepamovin | 1 comments | 09 Dec 25 15:00 UTC | HN request time: 0.443s | source
Show context
hn_throwaway_99 ◴[10 Dec 25 02:00 UTC] No.46213179[source]▶
This is awesome, but minor quibble with the title - "hallucinates" is the wrong verb here. You specifically asked it to make up a 10-year-in-the-future HN frontpage, and that's exactly what it did. "Hallucinates" means when it randomly makes stuff up but purports it to be the truth. If some one asks me to write a story for a creative writing class, and I did, you wouldn't say I "hallucinated" the story.
replies(4): >>46213931 #>>46215177 #>>46215634 #>>46219940 #
zwnow ◴[10 Dec 25 07:44 UTC] No.46215177[source]▶
If someone asked you, you would know about the context. LLMs are predictors, no matter the context length, they never "know" what they are doing. They simply predict tokens.
replies(1): >>46215398 #
block_dagger ◴[10 Dec 25 08:25 UTC] No.46215398[source]▶
This common response is pretty uninteresting and misleading. They simply predict tokens? Oh. What does the brain do, exactly?
replies(3): >>46215589 #>>46215987 #>>46216410 #
wafflemaker ◴[10 Dec 25 10:01 UTC] No.46215987[source]▶
It does exactly the same, predicts tokens, but it's totally different and superior to LLMs /s

OTOH, brain tokens seem to be concept based and not always linguistic (many people think solely in images/concepts).

replies(2): >>46216725 #>>46219733 #
1. ricardobeat ◴[10 Dec 25 11:54 UTC] No.46216725[source]▶
LLMs are “concept based” too, if you can call statistical patterns that. In a multi-modal model the embeddings for text, image and audio exist in the same high-dimensional space.

We don’t seem to have any clue if this is how our brain works, yet.