Looks like my previous account got permanently deleted, because I was deactivated for more than 30 days.
11 years of my online history is lost.
So here I am, starting all over again.
7) Companies like OpenAI and Anthropic have weaponized their policy folks to align their narratives with the company product dominance goals. If someone employs you, you cannot be a free-thinker about many topics, especially if your paycheck and promotions depend on it.
It’s a certain Anglo-centric, particularly America-centric, myopic understanding of the people of the world that thinks that things happen only here. If nuclear technology was quickly replicated elsewhere (USSR, China, India), LLMs are nothing compared to them in complexity.
استارتآپ ما امسال داره رشد خیلی زیادی تجربه میکنه. از ۲۰ هزار کاربر رسیدیم به ۲۲۰ هزار و سودده شدیم. ماه پیش هم ۲.۶ میلیون دلار سرمایه seed جذب کردیم و حالا داریم کلی استخدام میکنیم.
اگر در ونکوور هستید یا میتونید تا ۶-۱۲ ماه دیگه نقل مکان کنید (ویزا و اجازه کار با ما) ... ۱/
Asked a question about pandoc from Llama 2 13B (chat) and it gave me a number of commands to try. All the arguments to pandoc were made-up and it was really disappointing!
@yassersouri یکی از اولین پروژههایی که تو آزمایشگاه انجام دادم شبیه همین بود (احتمالا از همین مقاله الهام گرفته بودم)، اینکه یه مدل random forest (!) روی ویژگیهای هر پیکسل ترین کنم برای edge detection! یادش بخیر :))
Was just testing the Google PaLM through Poe and asked it a research question and it gave me a short summary and links to three papers.
ALL THREE were made up and the links were for unrelated arxiv papers!
@yassersouri Yeah, I tested it with Bing too and it gave me really good results. I had done a Google Search too, but the links Bing gave me were really useful (much more than Google Search results).
We're launching Keras Core, a new library that brings the Keras API to JAX and PyTorch in addition to TensorFlow.
It enables you to write cross-framework deep learning components and to benefit from the best that each framework has to offer.
Read more: https://t.co/xmmxBfSZgh
Exciting updates to #stablediffusion with Core ML!
- 6-bit weight compression that yields just under 1 GB
- Up to 30% improved Neural Engine performance
- New benchmarks on iPhone, iPad and Macs
- Multilingual system text encoder support
- ControlNet
https://t.co/wVMDG5FZjc 🧵
We are pleased to announce the return of CLIC this fall, as part of DCC 2024. Schedule and details to follow in a few weeks. Please join our forum here https://t.co/TRMa2bPVu1 to stay up to date.
One of the minor perks of presenting at WWDC the last four years is that Apple has accidentally given me the world's highest-budget transition timeline.
Reeling from the reputational damage SBF caused to EA, this became somewhat of an existential risk to the EA movement itself: nukes are too obvious, mosquito nets are too small, putting AI x-risk on the map was the path to show the world the enormous value EA offers society.
yeah the model makes occasional mistakes, is demonstrated to be full of biases, and cannot be trusted to supply reliable information. but arent you thankful for all the hard work that we do to make sure it doesnt actively try and kill you? you woke up alive today. thanks to us.