HackerTrans
TopNewTrendsCommentsPastAskShowJobs

icpmacdo

no profile record

Submissions

Stanford CS229(Machine Learning) Building Large Language Models

youtube.com
2 points·by icpmacdo·2 tahun yang lalu·0 comments

Safe Superintelligence Inc. – Ilya Sutskever

twitter.com
41 points·by icpmacdo·2 tahun yang lalu·28 comments

Jan Leike joins Anthropic on their superalignment team

twitter.com
99 points·by icpmacdo·2 tahun yang lalu·33 comments

Lucid CFO Steps Down

bnnbloomberg.ca
1 points·by icpmacdo·3 tahun yang lalu·0 comments

[untitled]

1 points·by icpmacdo·3 tahun yang lalu·0 comments

GitHub Raises $250M at $2B Valuation

wsj.com
569 points·by icpmacdo·11 tahun yang lalu·389 comments

comments

icpmacdo
·tahun lalu·discuss
not that much because its getting better at all benchmarks
icpmacdo
·tahun lalu·discuss
"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
icpmacdo
·tahun lalu·discuss
incredible results
icpmacdo
·tahun lalu·discuss
Modern AI both shortens the useful lifespan of software and increases the importance of development speed. Waiting around doesn’t seem optimal right now.
icpmacdo
·2 tahun yang lalu·discuss
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.
icpmacdo
·2 tahun yang lalu·discuss
Gary Marcus is continuously lambasted and not taken seriously
icpmacdo
·2 tahun yang lalu·discuss
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....
icpmacdo
·2 tahun yang lalu·discuss
can you share learning resources on this topic
icpmacdo
·2 tahun yang lalu·discuss
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
icpmacdo
·2 tahun yang lalu·discuss
Why don't they already carry payloads? Is there anything worth taking up with the current expected value of it exploding ect?
icpmacdo
·2 tahun yang lalu·discuss
Can you link some of the best youtube channels for those?
icpmacdo
·2 tahun yang lalu·discuss
Always cool to see innovative solutions like this
icpmacdo
·2 tahun yang lalu·discuss
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....
icpmacdo
·2 tahun yang lalu·discuss
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!
icpmacdo
·2 tahun yang lalu·discuss
because it is still the most interesting field of study
icpmacdo
·2 tahun yang lalu·discuss
He's alive
icpmacdo
·2 tahun yang lalu·discuss
its a native swift app
icpmacdo
·2 tahun yang lalu·discuss
No there is not
icpmacdo
·2 tahun yang lalu·discuss
More info with an interview with Ilya here https://www.bloomberg.com/news/articles/2024-06-19/openai-co...
icpmacdo
·2 tahun yang lalu·discuss
Thanks for the work you put into the Svelte platform, competition helps keep ecosystems more honest