Reposted by Prakash | (Ate-A-Pi)Prakash | (Ate-A-Pi) @8teapi.bsky.social · 04/07/2025AI video trailer by Ai filmmaker KavanTheKid Yup. Trad film is done. 141
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 04/07/2025AI video trailer by Ai filmmaker KavanTheKid Yup. Trad film is done. 141
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 14/01/2025😆 leftist TikTok refugees migrating to Chinese red book app thinking it’s Maoist but it’s actually named for Stanford business school and Bain Capital is the best timeline. 030
Reposted by Prakash | (Ate-A-Pi)Dror Poleg @drorpoleg.com · 12/12/2024ChatGPT kept interrupting my conversation with @8teapi.bsky.social 🤣 021
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 28/11/2024I just found out anyone can download your posts on bluesky 050
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 22/11/2024Incredible things are happening in 🇨🇳 Haptic feedback tank built by autistic nerds just to play World Of Tanks. 020
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024AGI timeline > 2026/2027 on a straight line extrapolation > Add 1-2 years for inevitable stumbling blocks > Blockers in the field have receded but not disappeared 020
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> In Anthropic's dealings with corporation and governments including USG - A few people get it, and start the ball rolling in deployment - At that point a competitive tailwind kicks in - If China deploys, the US starts competing - “The spectre of competion plus a few visionaries” is all it takes 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> Pushes back on @tylercowen 's 50-100 years estimate > Most economists follow Robert Solow, “You see the computer revolution everywhere except the productivity statistics.”. 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024How fast will AGI create change? > Dario picks middle ground > Does not see AGI conquering the world in five days, since a) physical systems take time to build and b) AGI that we do build would be law abiding, and have to deal with human system complexities and regulations. 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024Who gets to decide the constitution for the strong AI? > Basic principles: - no CBRN risks - adhere to rule of law - basic principles of democracy > Outside of that believes users should be free to fine tune and use models as they please. 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> Promising areas for research with insufficient people working on them: - mechanistic interpretabilty - long horizon learning and long horizon task - evaluations particularly of dynamic systems - multi-agent coordination 120
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024On hiring > “talent density beats talent mass”, prefer “ a team of 100 people that are super smart, motivated and aligned with the mission“ 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> Anthropic's way of dealing with this uses an if/then framework: If "you can show the model is dangerous" then "you clamp down hard" > Mechanistic interpretability is where they hope to be able to verify and check model state in ways the model cannot access on its own 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024- Level 5 - Models become more capable than all humanity 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024- Level 4 - They start to increase the capacity of a state actor, and become the primary source of CBRN risks. If you wanted to do something dangerous, you would use the model. Autonomy and deception, including sandbagging tests and sleeper agents become an issue. 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024- Level 3 - Agents, which we hit next year. They start to be a useful assistant in increasing a non-state actors capacity for a CBRN attack. Can be mitigated with filters as the model is not yet autonomous, and efforts must be made to prevent theft by non-state actors. 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024- Level 2 - ChatGPT/Claude, about as capable as Google in providing CBRN info, and no autonomy 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> AI Safety Levels - Level 1 -> chess playing Deep Blue, narrow task focused AI 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024Risks and Safety > worries about a) cyber, bio, radiological, nuclear (CBRN) b) model autonomy 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024On complaints models have become dumber: > believes its hedonic adjustment > besides pre-release A/B testing the day prior, Anthropic does not change its models 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024> Mechanistic interpretability is a safety and transparency technique Anthropic leads at and is a recruiting draw > when they win a new hire, Dario tells them "The other places you didn’t go, tell them why you came here." > Has led to interpretability teams being built out elsewhere 100
Prakash | (Ate-A-Pi) @8teapi.bsky.social · 21/11/2024The Race To The Top Dario Amodei on Lex Fridman TLDR: AGI 2026-2029. Things could be great, but risks remain. Dario reveals some of Anthropic's Jedi mind tricks Highlights: On the Race to the Top: > "trying to push the other players to do the right thing by setting an example" 130
