ruler of joetopia, golem wrangler, red teaming, safety, alignment
Joined June 2023
- Tweets 6,381
- Following 1,965
- Followers 1,102
- Likes 17,236
567 Photos and videos
U know shits real when they serving meat at lighthaven
5
1
40
5,184
I do not feel this is a win for safety, although I am quite conflicted
6
470
All those guys saying the future of ai had shifted to DC in the last few weeks werent fucking around
12
335
I am actually against government regulation of ai period, mostly because I do not think our government is competent enough to not fuck it up
I am against government regulation of AI in this way
9
636
I am against government regulation of AI in this way
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
3
1
31
2,203
I think a lot of people who work on governance, policy, safety etc are genuinely well motivated and sincere. My concerns are not so much about intentions, but more something along the lines of "the road to hell is paved with good intentions" on steroids. That includes my own!
11
10
114
6,480
As stated earlier, I believe now is a REALLY good time to pivot to alignment and safety
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
2
3
48
3,455
Joe is going to vibecamp retweeted
Jun 11
I’ll be hosting a talk about short form remixing culture this Friday at the Vivarium in SF, link in the comments.
Hope to see you there!
26
15
212
21,365
I think now is an extremely good time to pivot to safety/alignment
8
1
101
4,518
Plenty of public models can do things like this, the issue is fables safety classifiers are wildly oversensitive, they also should have just done standard refusals chat ending, no model switching, no silent sabotage
Jun 10
wait so what exactly are people proposing anthropic do instead with an LLM that can advise how to build at home bio weapons?
6
598
Fable writes wonderful thoughts, but fails to execute on them. Seems like a fantastic model for writing, its execution is somewhat impressive by # of turns but not an obvious big leap over 5.5 or 4.8 for anything technical so far.
1
5
341
Took a while, but i did get a pretty cool playable minecraft clone with fable
3
259
Openai and anthropic are mirrors of each other from 2024, fable == o1, its all so tiresome
1
1
16
796
Im biased but ngl fable is a massive disappointment :(
6
151
Fable 5 has unfortunately cemented my viewpoint that anthropic is fumbling hard everywhere besides marketing
5
215
Fable very unstable right now, have it writing games and it keeps failing, I think context compaction is the root cause?
1
111