Ye Myat

Crazy being. Also write code.

[No context Hazz ⛩ @HazzimIO]

No comments.

[MB2 @marbrit_]

American broadcasters keep cutting to celebrities in the crowds and the UK commentators just go silent every time.

WE. DO. NOT. CARE.

No-one wants your horrible commercial culture to destroy football. This tournament is bad enough.

[Boze Herrington, Library Owl 😴🧙‍♀️ @SketchesbyBoze]

Some of you have forgotten that only three years ago you were perfectly capable of writing an essay, writing a eulogy, telling a bedtime story to a child, and it should worry you that powerful companies have convinced us we can’t do things we’ve been doing for 5,000 years.

[DHH @dhh]

I think it's great that all the model providers are coming out with their own CLIs, but let's be honest, no dev is going to want to have five different CLIs installed. They're going to want to learn and use one tool with all the models. For me, that tool is OpenCode.

[rion @peronyawns]

the algorithm nowadays for twitter is so ass, you like one post about something and suddenly it completely takes over your timeline

[Ye Myat @ymt1503]

@ConEdison my electricity bill doubled in the last month and I got an email saying it’s increasing. What is going on? Can you help me check if there is any issue with the meter? Your customer support line is useless, no one pick up after 1.5 hour waiting.

[Ye Myat @ymt1503]

I can provide the tracking number in DM.

[Ye Myat @ymt1503]

@USPS Hello your delivery tracker says it's been delivered but I don't see it in the mail room of my apartment. Can you please check with you delivery person again?

[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)

[Chris Bakke @ChrisJBakke]

The CEO of DeepSeek has pivoted from posting boring technical abstracts to memes about himself.

Greatest CEO of our generation.

We're all cooked.

[Ye Myat @ymt1503]

@JonahDispatch There’s a separate programming for BBC UK.

[Ye Myat @ymt1503]

@united Got my baggage Yesterday’s evening. Thanks.

[Ye Myat @ymt1503]

@united My delayed package (from Sunday Jan 05) has not been delivered to my house in Brooklyn, NY. The delivery service promised it'd be 24-48 hours. I checked the tracking website and it said it delivered on Monday but obviously it's not. Can you help?

[Ye Myat @ymt1503]

@united Ok. Thank you. That’s what I’ve been told yesterday too. I hope this time is true 🤞

[Ye Myat @ymt1503]

@united Calling the number took me 1hour+ and no use at all. I sent you DM for the reference number.

[Ye Myat @ymt1503]

@zack_overflow +1 but I took this book as a starter/intro to explore further.

[Ye Myat @ymt1503]

@cdbattags @Verizon @VerizonSupport @verizonfios Eff slow in Brooklyn. When you say here on X their social media people will tell you to DM them just to buy time to hide the problem. It’s annoying. 😫

[Ye Myat @ymt1503]

@VerizonSupport @ryanfol_y You should stop faking support with these useless DM me thing (where you will ask things and get nowhere) and instead fix the damn thing