Saurabh Arya

Ze Monkey Mind

[Saurabh Arya @vandermyder]

Funniest thing is when Manjeraker calls Saim Ayub - Saiyyam (संयम) Ayub and Basit Ali calls Virat(विराट) - Virat (विरात). Plus both of them keep calling it a blockbuster while the stadium is empty! Media jokers

[Saurabh Arya @vandermyder]

@nishkumar1977 The amount of time you spend telling people to behave or to do some homework is crazy. The Excel sheet was a great first step. Here is a potential second step. Capture top requests. You are an admin, when you login with your gmail.

https://t.co/OTGnk1GM94

[Saurabh Arya @vandermyder]

@Citrini7 Is it coz J10 is a superior product or coz their customer has nowhr else to go & they will need more? CH is operating in a monopolistic envr (nobdy else is selling to their customer) & war means more demand. DA is in a competitive setup with competition from local/EU/US Cos

[Saurabh Arya @vandermyder]

@nishkumar1977 I get the severity part. The problem is with their conviction on urgency. They cook up a reason to claim now is the time. Doing it for yrs. I get it - they need to sell subscriptions & hence need headlines. Their track record is not very good. Who knows law of avgs may catch up

[Saurabh Arya @vandermyder]

@nishkumar1977 EWI - Coz of some random prop indicator + arbit media headlines they have been predicting a crash in the short term, for years. They run a very profitable content biz reliant on fear mongering.

[Andrej Karpathy @karpathy]

I don't have too too much to add on top of this earlier post on V3 and I think it applies to R1 too (which is the more recent, thinking equivalent).

I will say that Deep Learning has a legendary ravenous appetite for compute, like no other algorithm that has ever been developed in AI. You may not always be utilizing it fully but I would never bet against compute as the upper bound for achievable intelligence in the long run. Not just for an individual final training run, but also for the entire innovation / experimentation engine that silently underlies all the algorithmic innovations.

Data has historically been seen as a separate category from compute, but even data is downstream of compute to a large extent - you can spend compute to create data. Tons of it. You've heard this called synthetic data generation, but less obviously, there is a very deep connection (equivalence even) between "synthetic data generation" and "reinforcement learning". In the trial-and-error learning process in RL, the "trial" is model generating (synthetic) data, which it then learns from based on the "error" (/reward). Conversely, when you generate synthetic data and then rank or filter it in any way, your filter is straight up equivalent to a 0-1 advantage function - congrats you're doing crappy RL.

Last thought. Not sure if this is obvious. There are two major types of learning, in both children and in deep learning. There is 1) imitation learning (watch and repeat, i.e. pretraining, supervised finetuning), and 2) trial-and-error learning (reinforcement learning). My favorite simple example is AlphaGo - 1) is learning by imitating expert players, 2) is reinforcement learning to win the game. Almost every single shocking result of deep learning, and the source of all *magic* is always 2. 2 is significantly significantly more powerful. 2 is what surprises you. 2 is when the paddle learns to hit the ball behind the blocks in Breakout. 2 is when AlphaGo beats even Lee Sedol. And 2 is the "aha moment" when the DeepSeek (or o1 etc.) discovers that it works well to re-evaluate your assumptions, backtrack, try something else, etc. It's the solving strategies you see this model use in its chain of thought. It's how it goes back and forth thinking to itself. These thoughts are *emergent* (!!!) and this is actually seriously incredible, impressive and new (as in publicly available and documented etc.). The model could never learn this with 1 (by imitation), because the cognition of the model and the cognition of the human labeler is different. The human would never know to correctly annotate these kinds of solving strategies and what they should even look like. They have to be discovered during reinforcement learning as empirically and statistically useful towards a final outcome.

(Last last thought/reference this time for real is that RL is powerful but RLHF is not. RLHF is not RL. I have a separate rant on that in an earlier tweet
https://t.co/RMIpFPVpuM)

[Saurabh Arya @vandermyder]

@straits_times Scammers are increasingly creating Straits Times look-alike websites to mislead the public.

Link - https://t.co/KGp6T8ULx7