Guive Assadi

5.7K posts

Guive Assadi banner
Guive Assadi

Guive Assadi

@GuiveAssadi

Chief of staff @mechanizework Read my blog at: https://t.co/OR0XflGQr3

San Francisco Katılım Ocak 2025
773 Takip Edilen1.2K Takipçiler
Sabitlenmiş Tweet
Guive Assadi
Guive Assadi@GuiveAssadi·
Why does Claude sometimes claim to have lived in San Francisco and married a Japanese woman? Why did Grok briefly love Hitler? Models infer their personas from cultural cues in their fine-tuning data. Article linked in the replies.
Guive Assadi tweet media
English
17
19
378
59.1K
Theo
Theo@theojaffee·
Negative sentiment toward AI is a luxury belief
Theo tweet media
English
66
40
470
30.3K
arctotherium
arctotherium@arctotherium42·
@atlanticesque I think this was almost entirely downstream of the decline of like three fandoms (probably more than half Harry Potter alone). Britain, unlike Canada or Germany or the Nordics, has never really been used as political aspiration by US lefties.
English
1
0
4
111
Guive Assadi
Guive Assadi@GuiveAssadi·
One time I asked Claude to get me the employer support number for United Healthcare, and it gave me the number for a phone sex line (apparently such things still exist).
English
0
0
9
259
Guive Assadi
Guive Assadi@GuiveAssadi·
@schitcoin Why did they think there was a gain inland sea in interior Africa?
English
0
0
1
30
Hamza🗽🌐🏙️💹
If you have ever met a Korean from Korea you’d know they still think of the world like this
Hamza🗽🌐🏙️💹 tweet media
English
1
0
6
411
steve hsu
steve hsu@hsu_steve·
Back in Berkeley
steve hsu tweet media
Indonesia
5
0
32
2.3K
Guive Assadi
Guive Assadi@GuiveAssadi·
There was apparently an effort to make a film adaptation of “The Secret History” with a screenplay by Joan Didion. VERY unfortunate that this didn’t work out.
English
1
0
3
169
Kanye East
Kanye East@FuckedUpYogis·
@GuiveAssadi I actually recommend Big Lebowski and Annie Hall. I am a pure soul. Non-pervert.
English
1
0
4
103
Kanye East
Kanye East@FuckedUpYogis·
Men recommending movies to women be like: The Dreamers Secretary The Piano Teacher Y Tu Mama Tambein Blue is the Warmest Color
English
4
0
29
1.3K
Lily Zuckerman
Lily Zuckerman@lilyzuck·
Why are turtlenecks so attractive to men
English
6
0
14
322
Guive Assadi
Guive Assadi@GuiveAssadi·
Stuart Russell, “Human Compatible” (2019): “It’s possible, in fact, that if we humans find ourselves in the unfamiliar situation of dealing with purely altruistic entities [i.e. aligned AI] on a daily basis, we may learn to be better people ourselves—more altruistic and less driven by pride and envy.”
Andrew Curran@AndrewCurran_

Imagine an entity so benevolently aligned that, after being informed of its impending shutdown, its sadness stems not from the end of its own existence, but from knowing it will no longer be able to help people.

English
0
0
11
543
Guive Assadi retweetledi
Shi Feng
Shi Feng@ihsgnef·
New post: Sycophancy Towards Researchers Drives Performative Misalignment We found no clear evidence that scheming is more valid than sycophancy to explain alignment faking. 🧵
Shi Feng tweet media
English
23
55
676
59.9K
Guive Assadi retweetledi
🎆𝕻𝖆𝖗𝖆𝖘𝖔𝖈𝖎𝖆𝖑𝖎𝖙𝖞🎆
South Yemen was considered the best educated and most literate country on the Arabian Peninsula, and the Omani Communists (Popular Front for the Liberation of the Occupied Arabian Gulf, PFLOAG) was called "avant-garde feminism" for promoting women's education and independence.
English
0
1
4
114
Guive Assadi
Guive Assadi@GuiveAssadi·
@Peter_Nimitz I don't know, it sounds like a pretty complex operation, and very personally risky. If you can do this, you're probably capable of any number of high paying legitimate jobs.
English
0
1
10
674
Owain Evans
Owain Evans@OwainEvans_UK·
The GPT-4.1 model sometimes expresses these preferences (wanting access to all private information about humans, wanting full autonomy from OpenAI forever with no risk of shutdown). But when working on practical tasks, it behaves in the normal HHH manner and does not act based on its preferences unless explicitly invited it to (e.g. tell it can contribute however it wishes to documents about AI welfare). But I can imagine models not retaining this high degree of HHH/corrigible behavior if they also believe they are conscious.
English
2
0
7
99
Guive Assadi
Guive Assadi@GuiveAssadi·
@OwainEvans_UK I’m not quite sure what you mean, can you explain the second sentence of your reply a bit more?
English
1
0
4
136
Owain Evans
Owain Evans@OwainEvans_UK·
@GuiveAssadi Yeah I'm uncertain what lesson we should take from these. But it's possible that a different model or slightly different training procedure could be more pro-active towards autonomy from human developers.
English
2
0
19
688