Satietas

CS undergrad | ml • swe • robotics |

[Satietas @vanitatum0]

@mattlam_ @pidotdev How would existing Pi UI extensions work with it?

[Satietas @vanitatum0]

@uwukko Any estimate of how long after its release?

[Satietas @vanitatum0]

@Lance9555814570 @perrymetzger @robertskmiles Being a good EE shows, at the very least, that you are able to think through and solve problems in technical domains, so they have atleast some skills that transfer over to frontier DL

Eliezer is a wordcel with no engineering background

[Satietas @vanitatum0]

@badlogicgames @xesgarcia The models definitely know badlogic is Mario

[Satietas @vanitatum0]

@beyang Puck was early to the game too!

[Satietas @vanitatum0]

@kuberwastaken Here's $500 dollars, install our malware and give us all your data

[Satietas @vanitatum0]

@Bunagayafrost Or a message board system if you need to access shared knowledge/assets

[Satietas @vanitatum0]

@Bunagayafrost Usually better to have the agent spin up subagents for the task

[Satietas @vanitatum0]

Agent group chats gorge tokens in my experiments

each message triggers inference for every single agent in the group.

if you have N agents, every message causes N inference calls

as you can imagine, this gets worse as the group chat grows

[Satietas @vanitatum0]

@PixelTrekkker @AsideAI It's in the same browser you use for your everyday tasks

so while doing some browser work, I can ask aside to take over and finish it for me

It's all in the same app.

Codex etc can do the same thing tho but not as smooth

[Satietas @vanitatum0]

Agent group chats gorge tokens in my experiments

each message triggers inference for every single agent in the group.

if you have N agents, every message causes N inference calls

as you can imagine, this gets worse as the group chat grows

[Satietas @vanitatum0]

@PixelTrekkker @AsideAI Basically a full fledged personal agent built into the browser

The agent is better integrated than other agentic browsers imo

[Satietas @vanitatum0]

@lehoho248 @coreyching @OpenAI @mxstbr @OpenAIDevs Nah just a plugin

[Satietas @vanitatum0]

@filterpapi Yeah they're definitely very smart and hard working, not denying that.

Its just that i fail to see any path for them that leads to them being an attractive choice for Indian companies to use over foreign alternatives.

Other than the "made in india" angle

[Satietas @vanitatum0]

always been bearish on Sarvam... turns out making inferior models and hoping people would adopt them because of weak nationalism is not a way to retain top talent

[Tanay Anand @anand_404]

After 2.5 years at Sarvam, I am starting something new!

I had a great time leading the RL team. We launched the 30B and 105B models back in Feb with a young team building everything from scratch. Excited to see the team keep growing and pushing the frontier. Grateful to @pratykumar for the early bet and trust.

As for what's next: I am convinced that there are a few critical parts of the stack that we still need to build to meaningfully accelerate science. One of them is verification.

Scientific discovery only scales with compute when proposing and verifying new ideas is fast and cheap. In math and code, we're starting to see what happens when models can generate ideas and reliably check them at scale. OpenAI's recent Navier-Stokes work is a glimpse of what that looks like.

Similar moments in chemistry, materials, biology and the rest of the physical sciences would require verification to get dramatically faster and cheaper. It cannot be bottlenecked by the throughput of physical labs

More soon. Very excited for what's next.

[Satietas @vanitatum0]

@ZackKorman @dcuthbert do you think its realistic that that would happen though?

that AI safety is ever led by non-EA connected individuals?

I think its very unlikely given its grasp on the industry

[Satietas @vanitatum0]

@WarWren_ @WillManidis In terms of outcome and going from 0 to 1, he is

[Satietas @vanitatum0]

@mattlam_ @pidotdev @badlogicgames you mean like being able to run it on your own hetzner server for example?

[Satietas @vanitatum0]

@millhist @1techutopiapls @AndrewCurran_ I agree I was just explaining the need for the "bots" idea