Nitish

354 posts

Nitish

Nitish

@StrongDuality

Palo Alto, CA Katılım Kasım 2013
1K Takip Edilen3.2K Takipçiler
Nitish retweetledi
Eric Wallace
Eric Wallace@Eric_Wallace_·
Today we release gpt-oss-120b and gpt-oss-20b—two open-weight LLMs that deliver strong performance and agentic tool use. Before release, we ran a first of its kind safety analysis where we fine-tuned the models to intentionally maximize their bio and cyber capabilities 🧵
Eric Wallace tweet mediaEric Wallace tweet media
English
106
327
2.6K
642.9K
Nitish
Nitish@StrongDuality·
something something alien with extraordinary ability something
English
0
0
8
1.2K
Nitish retweetledi
OpenAI
OpenAI@OpenAI·
We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond. These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. openai.com/index/introduc…
English
936
3.9K
17.3K
8.3M
Nitish retweetledi
Bob McGrew
Bob McGrew@bobmcgrewai·
I hope you enjoyed the livestream! It's been a privilege to work with the incredibly talented team that put GPT-4o together.
English
7
1
70
3.3K
Nitish retweetledi
Greg Brockman
Greg Brockman@gdb·
Announcing Sora — our model which creates minute-long videos from a text prompt: openai.com/sora
English
1K
3.3K
19.3K
8.2M
Aidan Clark
Aidan Clark@_aidan_clark_·
This sounds stupid but it took till a surprisingly far point in my career to start resuming experiments when they hit a problem -- I'd just restart from scratch. This is a terrible instinct, get rid of it!
English
3
0
14
2.1K
Nitish
Nitish@StrongDuality·
Managing such a large effort (from design, training, eval, deployment, …) is, unsurprisingly, quite difficult. The model looks very strong, looking forward to playing with it. Congratulations to the Gemini team!
Demis Hassabis@demishassabis

The Gemini era is here. Thrilled to launch Gemini 1.0, our most capable & general AI model. Built to be natively multimodal, it can understand many types of info. Efficient & flexible, it comes in 3 sizes each best-in-class & optimized for different uses blog.google/technology/ai/…

English
0
0
5
1.6K
Wojciech Kryściński
Wojciech Kryściński@iam_wkr·
Personal update: After nearly 5 years with @SFResearch I decided to embark on a new professional adventure. Excited to share that I have joined the amazing team @cohere to work on LLMs. Very grateful for the exceptional colleagues and collaborators I have had in the past years!
English
8
1
119
13.8K
Nitish
Nitish@StrongDuality·
@_aidan_clark_ are you staring at the loss curves sufficiently? i have on good authority that that matters.
English
1
0
2
236
Aidan Clark
Aidan Clark@_aidan_clark_·
Nothing like the feeling of waiting to find out if you’ve fixed a bug or not….
English
4
0
10
1.4K
Nitish
Nitish@StrongDuality·
@savvyRL It's an OWCA conspiracy.
GIF
English
0
0
1
357
Rosanne Liu
Rosanne Liu@savvyRL·
10th citation of the platypus paper! Yet still can't get it accepted to any conference... Have fun with open vocab classification! arxiv.org/abs/2209.03320
Rosanne Liu tweet media
English
6
4
78
18.6K
Nitish retweetledi
Barret Zoph
Barret Zoph@barret_zoph·
Our team at OpenAI is hiring! We're looking for engineers/researchers who do rigorous and thoughtful work understanding and evaluating LLMs like ChatGPT. If you're interested, please apply online and DM me with work that you've done!
English
38
97
704
624.9K
Sam Bowman
Sam Bowman@sleepinyourhat·
To give credit, the key method here is largely the same as in prior work from @StrongDuality here. We just show that it works extremely well in settings like this where conventional wisdom had been to use RLHF (or a few other less-effective alternatives). arxiv.org/abs/1909.05858
English
1
0
9
2.2K
Sam Bowman
Sam Bowman@sleepinyourhat·
Proud to see Tomek et al.'s _pretraining with human feedback_ highlighted as a core technique on the first page of Google's PALM-2 (Bard) write-up! ai.google/static/documen…
Sam Bowman tweet media
Tomek Korbak@tomekkorbak

You can (and should) do RL from human feedback during pretraining itself! In our new paper, we show how training w/ human preferences early on greatly reduces undesirable LM behaviors, including under adversarial attack, w/o hurting downstream performance. arxiv.org/abs/2302.08582

English
5
7
89
32.2K
Nitish retweetledi
OpenAI
OpenAI@OpenAI·
Announcing GPT-4, a large multimodal model, with our best-ever results on capabilities and alignment: openai.com/product/gpt-4
English
2K
16.2K
60.4K
12.5M
Red
Red@adventurared·
I got a handful of interesting questions about working @OpenAI in my DMs and some standard ones. Open to more questions!
English
4
0
18
5.1K
Nitish retweetledi
OpenAI
OpenAI@OpenAI·
ChatGPT and Whisper are now available through our API (plus developer policy updates). We ❤️ developers: openai.com/blog/introduci…
English
641
2.4K
10K
2.4M

Keşfet

@j_mcgraph @_aidan_clark_ @gdb @iam_wkr @SFResearch @cohere @cgivancouver @savvyRL