Hacker News: icpmacdo

New comment by icpmacdo in "Claude 4"

icpmacdo — Fri, 23 May 2025 02:32:26 +0000

not that much because its getting better at all benchmarks

New comment by icpmacdo in "Claude 4"

icpmacdo — Fri, 23 May 2025 02:28:17 +0000

"It feels like these new models are no longer making order of magnitude jumps, but are instead into the long tail of incremental improvements. It seems like we might be close to maxing out what the current iteration of LLMs can accomplish and we're into the diminishing returns phase."

SWE bench from ~30-40% to ~70-80% this year

New comment by icpmacdo in "One-Minute Video Generation with Test-Time Training"

icpmacdo — Tue, 08 Apr 2025 02:32:36 +0000

incredible results

New comment by icpmacdo in "O1 isn't a chat model (and that's the point)"

icpmacdo — Sat, 18 Jan 2025 21:25:15 +0000

Modern AI both shortens the useful lifespan of software and increases the importance of development speed. Waiting around doesn’t seem optimal right now.

New comment by icpmacdo in "GPT-5 is behind schedule"

icpmacdo — Tue, 24 Dec 2024 19:47:25 +0000

He only criticizes ai capabilities, without creating anything himself. Credentials are effectively meaningless. With every new release, he clamors for attention to prove how right he was—and always will be. That’s precisely why he lacks credibility.

New comment by icpmacdo in "GPT-5 is behind schedule"

icpmacdo — Mon, 23 Dec 2024 02:03:47 +0000

Gary Marcus is continuously lambasted and not taken seriously

New comment by icpmacdo in "GPT-5 is behind schedule"

icpmacdo — Mon, 23 Dec 2024 02:01:40 +0000

This is literally just the scaling laws, "Scaling laws predict the loss of a target machine learning model by extrapolating from easier-to-train models with fewer parameters or smaller training sets. This provides an efficient way for practitioners and researchers alike to compare pretraining decisions involving optimizers, datasets, and model architectures"

https://arxiv.org/html/2410.11840v1#:~:text=Scaling%20laws%2....

New comment by icpmacdo in "GPT-5 is behind schedule"

icpmacdo — Mon, 23 Dec 2024 01:57:11 +0000

can you share learning resources on this topic

New comment by icpmacdo in "OpenAI O3 breakthrough high score on ARC-AGI-PUB"

icpmacdo — Sat, 21 Dec 2024 22:29:42 +0000

ARC co-founder Mike Knoop

"Raising visibility on this note we added to address ARC "tuned" confusion:

> OpenAI shared they trained the o3 we tested on 75% of the Public Training set.

This is the explicit purpose of the training set. It is designed to expose a system to the core knowledge priors needed to beat the much harder eval set.

The idea is each training task shows you an isolated single prior. And the eval set requires you to recombine and abstract from those priors on the fly. Broadly, the eval tasks require utilizing 3-5 priors.

The eval sets are extremely resistant to just "memorizing" the training set. This is why o3 is impressive." https://x.com/mikeknoop/status/1870583471892226343

New comment by icpmacdo in "SpaceX Super Heavy splashes down in the gulf, canceling chopsticks landing"

icpmacdo — Wed, 20 Nov 2024 02:07:23 +0000

Why don't they already carry payloads? Is there anything worth taking up with the current expected value of it exploding ect?

New comment by icpmacdo in "Ask HN: What would you preserve if the internet were to go down tomorrow?"

icpmacdo — Sun, 03 Nov 2024 17:50:11 +0000

Can you link some of the best youtube channels for those?

Stanford CS229(Machine Learning) Building Large Language Models

icpmacdo — Wed, 23 Oct 2024 20:05:51 +0000

Article URL: https://www.youtube.com/watch?v=9vM4p9NN0Ts

Comments URL: https://news.ycombinator.com/item?id=41928824

Points: 2

# Comments: 0

New comment by icpmacdo in "iPhone 16 is much easier to repair"

icpmacdo — Mon, 23 Sep 2024 07:40:21 +0000

Always cool to see innovative solutions like this

New comment by icpmacdo in "OpenAI threatens to revoke o1 access for asking it about its chain of thought"

icpmacdo — Fri, 13 Sep 2024 22:12:03 +0000

Scaling The Turk to OpenAI scale would be as impressive as agi

"The Turk was not a real machine, but a mechanical illusion. There was a person inside the machine working the controls. With a skilled chess player hidden inside the box, the Turk won most of the games. It played and won games against many people including Napoleon Bonaparte and Benjamin Franklin"

https://simple.wikipedia.org/wiki/The_Turk#:~:text=The%20Tur....

New comment by icpmacdo in "OpenAI threatens to revoke o1 access for asking it about its chain of thought"

icpmacdo — Fri, 13 Sep 2024 22:07:31 +0000

This is what an incredible level of product market fit look's like, people act like they are forced to pay for these services. Go use a local LLAMA!

New comment by icpmacdo in "Learning to Reason with LLMs"

icpmacdo — Thu, 12 Sep 2024 19:12:52 +0000

because it is still the most interesting field of study

New comment by icpmacdo in "Patagonia gave its staff 3 days to decide to relocate or quit"

icpmacdo — Sun, 30 Jun 2024 01:10:58 +0000

He's alive

New comment by icpmacdo in "Claude 3.5 Sonnet"

icpmacdo — Thu, 27 Jun 2024 16:28:37 +0000

its a native swift app

New comment by icpmacdo in "Claude 3.5 Sonnet"

icpmacdo — Thu, 20 Jun 2024 16:58:26 +0000

No there is not

New comment by icpmacdo in "Safe Superintelligence Inc. – Ilya Sutskever"

icpmacdo — Wed, 19 Jun 2024 17:07:57 +0000

More info with an interview with Ilya here https://www.bloomberg.com/news/articles/2024-06-19/openai-co...