LucasLanglois

Funnily enough it's not their first attempt at launching a prediction market. They tried exactly the same in 2021 -> https://qz.com/2069284/facebook-is-shutting-down-its-experim...

If you don't know Tom Scott, he has done a great video 4mn vide on the model where you can see it in action: https://www.youtube.com/watch?v=i70wkxmumAw

Love that Cloudflare put together a participative and community-driven advent calendar!

Ask 3-5 domain experts to share their experience on that.

If a book comes up multiple time start with it and then iterate on that knowledge.

You can also look for a university syllabus about the topic to get an overview.

Interestingly, this highlights that despite popular belief, the scientific process is not about being right or wrong but rather what's powerful.

A theory is promoted because it is powerful in explaining observations and providing tools for humanity until it is improved by the next one. You can build a gun with gravitation but you need relativity for a rocket.

Wikimedia net assets grew by $50M in 2021. Check the article to get more information.

The point of the piece is to highlight the discrepancy between the tone of Wikipedia's fundraising and mission and the reality of their system and financial situation.

We learned a lot by looking at communities successfully producing knowledge. The scientific community is one example, Wikipedia is another one. We have looked at them carefully and it’s become clear to us that they build knowledge in a very different way than media, politics and people do it in their personal life, on social media and even on the Internet in general.

It has led us to ask ourselves: what if thinking of the truth couldn't be done in the same way as it used to, and instead requires a new framework?

Pallas is a technology to think of facts, truth and knowledge in a completely new way.

That's one part of the answer.

The other is that you can use Inferrd from your notebook and have it running in 1mn.

Hey, Lucas here. I’m excited to share Inferrd along with my co-founder Theo. It’s a product we started working on after dealing with our fair share of deploying models on GPU in traditional clouds (GCP, AWS, Azure) for consulting contracts.

Hosting AI is challenging because it requires GPUs and most of the time, data scientists are left to design and build the infrastructure. Configuring Nvidia and Cuda is another issue that can consume days or weeks (add a few more weeks for TensorRT).

That’s why we built Inferrd, it’s a Heroku-style GPU hosting platform specifically designed for AI models. All the libraries are pre-configured, no cold start and out-of-the-box support for most major frameworks (spaCy, TensorFlow, PyTorch, Keras, Scikit, XGBoost). In addition, we leverage GPU sharing to provide very competitive pricing. Lastly, we provide key authentication, smooth scaling and monitoring tools.

AI hosting is a very interesting new space, and we’ll be around to answer any questions!