kevin @kevinmt24 - Twitter Profile

about 1 month ago

every time you catch yourself correcting your Claude agent the same way twice, turn it into an instruction in CLAUDE.md. compounding returns. I know this is a no-brainer for all, but it gives does give huge returns!

0

61

kevin @kevinmt24

6 months ago

@chuhakabacha Let's gooo!

0

34

kevin @kevinmt24

6 months ago

Switched from Spotify to Apple Music a month ago. Loving the lossless audio, Automix, and recommendations. But the UX still bugs me — that's where Spotify shines. And don't get me started on the liquid glass mess.

kevinmt24's tweet photo. Switched from Spotify to Apple Music a month ago. Loving the lossless audio, Automix, and recommendations. But the UX still bugs me — that's where Spotify shines. And don't get me started on the liquid glass mess. https://t.co/6iaG4iwjyN

0

2

0

164

kevinmt24 retweeted

Jim Fan

@DrJimFan

almost 2 years ago

OpenAI Strawberry (o1) is out! We are finally seeing the paradigm of inference-time scaling popularized and deployed in production. As Sutton said in the Bitter Lesson, there're only 2 techniques that scale indefinitely with compute: learning & search. It's time to shift focus to the latter. 1. You don't need a huge model to perform reasoning. Lots of parameters are dedicated to memorizing facts, in order to perform well in benchmarks like trivia QA. It is possible to factor out reasoning from knowledge, i.e. a small "reasoning core" that knows how to call tools like browser and code verifier. Pre-training compute may be decreased. 2. A huge amount of compute is shifted to serving inference instead of pre/post-training. LLMs are text-based simulators. By rolling out many possible strategies and scenarios in the simulator, the model will eventually converge to good solutions. The process is a well-studied problem like AlphaGo's monte carlo tree search (MCTS). 3. OpenAI must have figured out the inference scaling law a long time ago, which academia is just recently discovering. Two papers came out on Arxiv a week apart last month: - Large Language Monkeys: Scaling Inference Compute with Repeated Sampling. Brown et al. finds that DeepSeek-Coder increases from 15.9% with one sample to 56% with 250 samples on SWE-Bench, beating Sonnet-3.5. - Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters. Snell et al. finds that PaLM 2-S beats a 14x larger model on MATH with test-time search. 4. Productionizing o1 is much harder than nailing the academic benchmarks. For reasoning problems in the wild, how to decide when to stop searching? What's the reward function? Success criterion? When to call tools like code interpreter in the loop? How to factor in the compute cost of those CPU processes? Their research post didn't share much. 5. Strawberry easily becomes a data flywheel. If the answer is correct, the entire search trace becomes a mini dataset of training examples, which contain both positive and negative rewards. This in turn improves the reasoning core for future versions of GPT, similar to how AlphaGo’s value network — used to evaluate quality of each board position — improves as MCTS generates more and more refined training data.

DrJimFan's tweet photo. OpenAI Strawberry (o1) is out! We are finally seeing the paradigm of inference-time scaling popularized and deployed in production. As Sutton said in the Bitter Lesson, there're only 2 techniques that scale indefinitely with compute: learning & search. It's time to shift focus to the latter.

1. You don't need a huge model to perform reasoning. Lots of parameters are dedicated to memorizing facts, in order to perform well in benchmarks like trivia QA. It is possible to factor out reasoning from knowledge, i.e. a small "reasoning core" that knows how to call tools like browser and code verifier. Pre-training compute may be decreased.

2. A huge amount of compute is shifted to serving inference instead of pre/post-training. LLMs are text-based simulators. By rolling out many possible strategies and scenarios in the simulator, the model will eventually converge to good solutions. The process is a well-studied problem like AlphaGo's monte carlo tree search (MCTS).

3. OpenAI must have figured out the inference scaling law a long time ago, which academia is just recently discovering. Two papers came out on Arxiv a week apart last month:

- Large Language Monkeys: Scaling Inference Compute with Repeated Sampling. Brown et al. finds that DeepSeek-Coder increases from 15.9% with one sample to 56% with 250 samples on SWE-Bench, beating Sonnet-3.5.
- Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters. Snell et al. finds that PaLM 2-S beats a 14x larger model on MATH with test-time search.

4. Productionizing o1 is much harder than nailing the academic benchmarks. For reasoning problems in the wild, how to decide when to stop searching? What's the reward function? Success criterion? When to call tools like code interpreter in the loop? How to factor in the compute cost of those CPU processes? Their research post didn't share much.

5. Strawberry easily becomes a data flywheel. If the answer is correct, the entire search trace becomes a mini dataset of training examples, which contain both positive and negative rewards.

This in turn improves the reasoning core for future versions of GPT, similar to how AlphaGo’s value network — used to evaluate quality of each board position — improves as MCTS generates more and more refined training data.

133

6K

1K

3K

801K

Who to follow

Overwhelmingly mediocre thoughts underwhelmingly written for public scrutiny.

kevinmt24 retweeted

about 2 years ago

So Apple has introduced a new system called “Private Cloud Compute” that allows your phone to offload complex (typically AI) tasks to specialized secure devices in the cloud. I’m still trying to work out what I think about this. So here’s a thread. 1/

223

11K

2K

8K

4M

kevinmt24 retweeted

OpenAI Developers

@OpenAIDevs

over 2 years ago

GPT-4o is now available in the API. It’s as smart as GPT-4 Turbo, has improved vision capabilities, and is much more efficient—2x faster, 50% cheaper, 5x rate limits. It supports text and vision today, with audio and video coming soon. Details in thread 🧵 https://t.co/JT3VmueGEQ

80

2K

331

283

404K

kevinmt24 retweeted

Jeff Boudier 🤗

@jeffboudier

over 2 years ago

Breaking: we got 1,200 tokens per second for Llama 2 7B on H100! 🔥🔥🔥 Check out our new project optimum-NVIDIA with fp8 inference support and first benchmarks for LLM acceleration: https://t.co/qYFsD4kC3n 🧵 👇

22

730

106

327

143K

kevinmt24 retweeted

Rockstar Games

@RockstarGames

almost 3 years ago

61K

2M

337K

48K

248M

kevinmt24 retweeted

Elon Musk

@elonmusk

almost 3 years ago

Cybertruck deliveries start on Thursday

22K

151K

12K

1K

38M

kevinmt24 retweeted

OpenAI

@OpenAI

almost 3 years ago

ChatGPT with voice is now available to all free users. Download the app on your phone and tap the headphones icon to start a conversation. Sound on 🔊

1K

16K

3K

2K

10M

kevinmt24 retweeted

Rockstar Games

@RockstarGames

almost 3 years ago

We are very excited to let you know that in early December, we will release the first trailer for the next Grand Theft Auto. We look forward to many more years of sharing these experiences with all of you. Thank you, Sam Houser

49K

1M

299K

43K

202M

kevinmt24 retweeted

OpenAI

@OpenAI

about 3 years ago

ChatGPT for Android is now available for download in the US, India, Bangladesh, and Brazil! We plan to expand the rollout to additional countries over the next week. https://t.co/NfBDYZR5GI

585

6K

1K

211

2M

kevinmt24 retweeted

flipflop.xyz @flipflop_xyz

over 3 years ago

GM GM! Fed up with the funky NFT market action, fam? 🚨 Launch Alert - Today, we are sharing our flipping recipe that has worked for us in every market condition Learn it all in under 2 mins ↓ 🧵...