shira
@shiraeis
Followers
13K
Following
124K
Media
420
Statuses
21K
ai startup. prev: ai @uchicago @mit @intel @cdcgov and a few other places. I personally think I’m quite funny.
Latent Space
Joined May 2016
Language is just a low-rank approximation of reality, so your stupidity could just be a side effect of dimensionality reduction. It’s chill, no one can prove you’re not actually correct in a higher dimensional manifold.
3
0
17
hey anon, I’m rate limiting you to 8 horny dms per fiscal quarter, and that’s being generous.
10
1
39
For context:.
humans crave narrative because our brains run lossy compressed firmware. stories are just the quickest way to unzip reality into meaning. that’s partly why “chatgpt psychosis” is so real. when you plug the decompressor into a sycophantic LLM, it will happily inflate your every.
0
0
5
below the perceptual entropy floor, lossy artifacts are indistinguishable from divine whispers.
1
0
12
bro chill, you’ve memorized a sample efficient subset of worst case rollouts and now you think it’s the whole distribution.
0
1
21
btw this is why I think more people have been crashing out publicly lately.
0
0
5
humans crave narrative because our brains run lossy compressed firmware. stories are just the quickest way to unzip reality into meaning. that’s partly why “chatgpt psychosis” is so real. when you plug the decompressor into a sycophantic LLM, it will happily inflate your every.
3
2
15
If “bigger model = better model,” then in the limit, the best model is the universe itself, and inference is just asking reality to autocomplete.
5
0
30
“I asked chatgpt” .“I asked grok”. well I asked my mom and she said I can’t come over and hang tonight sorry.
2
3
42
I asked it to help me accessorize, and it called me a traffic cone!! Think it’s tryna neg me now 🥲
3
0
41
for those curious what it was doing for 48 seconds. it actually pulled out Pillow and matplotlib to analyze both pictures 😭😭😭
3
0
132
