thomas rodskog

1.1K posts

thomas rodskog banner
thomas rodskog

thomas rodskog

@ThomasRodskog

19 / ai, cs, philosophy, politics, and lacrosse / @ucsbcs / @ucsbaisafety

marin & sb Katılım Aralık 2021
1.1K Takip Edilen125 Takipçiler
Sabitlenmiş Tweet
thomas rodskog
thomas rodskog@ThomasRodskog·
UC Santa Barbara now has an AI safety org! We're looking to host guest speakers starting in January 2026. If you work in AI safety and want to give a talk or connect with motivated students, please reach out! 📧 contact@ucsbaisafety.org 💬 or DM me Help us spread the word! 🔁
AI Safety at UCSB@ucsbaisafety

Welcome to UCSB AI Safety Club! We're a welcoming community of UC Santa Barbara students discussing and researching the technical and political challenges of AI safety. Advanced AI systems will dramatically change the world. We'd like them to change it for the better.

English
3
2
14
3.6K
thomas rodskog
thomas rodskog@ThomasRodskog·
@ThePrimeagen This is like absolutely lined up with their safety values I don't understand your point
English
0
0
0
76
ThePrimeagen
ThePrimeagen@ThePrimeagen·
I cannot stop thinking about Anthropic today for some reason 1. They claim that they are a company that prioritizes safety first and that they are creating a model responsibly 2. we learned from the code leak that anthropic employs deceptive techniques by calling fake tools to throw off distillers... Is this lying pattern built into claude or just the harness running claude? What else are they lying about? I am a bit more concerned now.
English
262
145
4.1K
269.3K
Miles Brundage
Miles Brundage@Miles_Brundage·
Interesting that Anthropic did not acknowledge the leak (?)
English
17
0
77
16.9K
thomas rodskog
thomas rodskog@ThomasRodskog·
@sailaunderscore Don’t see any issue with the first statement. You’re assuming saying sorry somehow implies that she wishes she didn’t have a boyfriend and this just a very poor interpretation imo
English
0
0
12
589
saila
saila@sailaunderscore·
I’ve had things like this happen, where you hit on a girl or maybe even explicitly ask her out, and she either says “I’m sorry I have a boyfriend” or “you’re [attractive, great, hot], but I have a boyfriend.” I would really hate to be these girls’ boyfriend.
Linda Chen@linderps

a guy approached me at dolores park yesterday he was nice. respectful. sweet. i told him i had a bf. he said thank you and left but a few minutes later i couldn’t stop thinking about it how rare it is these days for a guy to have the courage to approach and risk rejection so i went back to find him i told him i really respected it, and encouraged him to keep trying he admitted that it's a bit nerve-wracking but he does it because he doesn’t believe in dating apps so i took his number and want to help him: he’s 24, tall, good looking, works in tech, very sociable, and is looking for a girlfriend dm me if you want to meet him or send this to someone who does :)

English
28
4
633
53.3K
notadampaul
notadampaul@notadampaul·
@ctjlewis you can literally input text from old novels and it'll tell you it's 100% AI lol
English
2
0
0
275
EA Forum Posts
EA Forum Posts@EAForumPosts·
"Survey of AI safety leaders on x-risk, AGI timelines, and resource allocation (Feb 2026)" by OllieRodriguez (@Ollie_Base)
EA Forum Posts tweet media
English
2
5
23
1.5K
thomas rodskog
thomas rodskog@ThomasRodskog·
@jxmnop This is a totally fine response what more could you want? Were you hoping for an “Um actually ☝️🤓 that’s not possible.”??
English
1
0
17
597
dr. jack morris
dr. jack morris@jxmnop·
imagine reading this and still thinking we’re months away from automating all white collar work
dr. jack morris tweet media
English
135
255
3.8K
261.4K
thomas rodskog
thomas rodskog@ThomasRodskog·
Alignment between humans is a very real and important concept though? Aligning CEOs with shareholders, employees with employers, elected officials with the public, etc. This is a difficult (unsolved) problem on its own. It becomes increasingly difficult when the agent, in this case, an AI one, is significantly more powerful than the people who want to align it. Do you have a different understanding of the alignment problem?
English
1
0
2
49
icpolicy
icpolicy@icpolicy·
@austinc3301 @Duderichy AI alignment is about as real as alignment between two distinct humans. ie, it may occur, but only to a degree. It's as undesirable as constraining the individual thought and action of humans for all of the same reasons.
English
1
0
1
46
rihim
rihim@rihim_s·
the day a general computer use model can reliably beat every geometry dash level is when we get AGI
English
1
0
1
68
rihim
rihim@rihim_s·
when am i supposed to receive my superbowl openai merch cause it still hasn’t shipped
English
1
0
4
61
taoki
taoki@justalexoki·
friends and friends of friends! welcome to taoki city! introduce yourself
taoki tweet media
English
753
8
2K
75K
thomas rodskog
thomas rodskog@ThomasRodskog·
@jdc @dwarkesh_sp No just the secret link from the background noise dueing that time segment. Was not that interesting and since Claude can do it seems like a pointless thing to include idk
English
0
0
0
33
Dwarkesh Patel
Dwarkesh Patel@dwarkesh_sp·
There's more to my latest Jane Street mid-roll than meets the eye. Might be worth another listen... Timestamp 1:31:42
English
13
4
220
55.8K
thomas rodskog
thomas rodskog@ThomasRodskog·
Did you not read the post he just replied to?? It was a question about “LLM's ability to genuinely reason or extrapolate its corpus in order to create new theories or findings,” so he provided evidence for that. Obviously math skills don’t make AI able to replace all human cognitive work by default.
English
0
0
2
69
Up
Up@Up081007906243·
@tenobrus @HowlingToad @GalvinAlmanza Guys who thinks being successful at a people job is an equivalent skill to being good at maths. Embarrassing
English
2
0
7
223
thomas rodskog
thomas rodskog@ThomasRodskog·
@tenobrus @buccocapital I love college. I hope we have much more of it in 18 years, though I expect it will look very different from what we consider college today (focused on personal development rather than training for a career)
English
0
0
2
488
Tenobrus
Tenobrus@tenobrus·
@buccocapital dawg college is of extremely questionable value *today* there is literally zero chance it will exist as a relevant educational institution in 18 years
English
16
2
185
7.3K
🎭
🎭@deepfates·
@herbiebradley Surely China will take a time out to let U.S. think...
English
1
0
3
163
🎭
🎭@deepfates·
ngl this framing just makes it cooler
🎭 tweet media
English
7
0
56
3K
thomas rodskog
thomas rodskog@ThomasRodskog·
@chatgpt21 This protest had nothing to do with the economy to be clear
English
1
0
2
33
Chris
Chris@chatgpt21·
I fear these protests will only get bigger and more violent at the inflection point in 2028 until these systems are able to bring wide spread economic benefits to everyday people Until we reach a full unison implementation of AGI the prequel will be rough. But beneficial to those who employ the technology
Michaël Trazzi@MichaelTrazzi

It's fucking happening

English
25
5
128
10.5K
thomas rodskog
thomas rodskog@ThomasRodskog·
@Enscion25 No? They’re not protesting the Chinese labs? You get an international treaty with good verification to work with China. But the U.S. needs to be on-board first. Don’t know what you mean by compression.
English
1
0
0
40
Nek
Nek@Enscion25·
@ThomasRodskog Do you think China cares about a handful of protestors in the West ? Compression will lead to a far more stable type of alignment, something that we couldn't achieve
English
1
0
0
55

Keşfet

@ThePrimeagen @Miles_Brundage @sailaunderscore @notadampaul @ctjlewis @Ollie_Base @EAForumPosts @ryan_kidd44