Miguel Fierro

Happy birthday to one of my heroes: Claude Shannon.

Without him we wouldn't have information theory, artificial intelligence, computers, telecommunications or happiness.

The one and only 🐐

📣 Miguel Fierro shares insights on combining Large Language Models (LLMs) with Knowledge Graphs (KGs) 🧠💡

1️⃣ LLMs excel in understanding language but can hallucinate.
2️⃣ KGs provide structured, accurate responses but lack language understanding.

Together, they could overcome

Miguel Fierro Meta I have been $META shareholder for years. Most of that drop was just a market decline that everyone experienced. But there was a metaverse component

Miguel Fierro Robert Scoble 😎 Austrian Australian Response

Microsoft releases Phi-3, the new version of their very efficient model.

Here are some details:
🔹Sizes: mini (3.8B), small (7B), and medium (14B).
🔹Context lengths 4K to 8K (extendable with RoPE).
🔹Mini trained on 3.3T tokens, small and medium on 4.8T.
🔹Performs

The Recommenders team has been making a lot of progress lately.

Cool stuff coming soon!

Stay tuned!

Phi3-3.8B surpasses Llama3-8B

This is better than watching formula one.

An AI helping some fish sing is probably the best thing you'll see today.

When someone says that most AI problems can be solved with linear regression.

These last few days have been a rollercoaster in AI.

While people are focused on Large Language Models, Microsoft is full speed on Small Language Models (SLMs).

Here are some takeaways:

🔹Better data >> more data
Phi 3 has been trained on heavily filtered web data and

A company launches a robot with a flamethrower attached.

I think there are like 20+ movies explaining why this is not a great idea.

Shannon is better than Messi.

And you know it

Here is a very powerful model for recommendation systems that not many people know about: factorization machines (FMs).

Details:
🔹It is fast and has a low memory consumption.
🔹FMs can be used for binary classification, regression, and ranking.
🔹FMs model interactions of order

If you want to get into AI, if you are looking for a job, it's ok to have ups and downs.

Just keep pushing.

Vamos!!!🦾🦾🦾

Miguel Fierro This explains everything.
I was done with Google in 2020: docs.google.com/document/d/1U9…

For those who want to switch their career to AI, networking is a booster.

Here are some ideas for effective networking:

- Leverage your current network.
- Provide value first, then ask favors.
- Keep the relationship going via messages.
- Follow people who can teach you

An amazing YOLO-NAS starter notebook.

Thanks Miguel Fierro 👏

Today I'm presenting at XGS, an internal Data Science meetup at Microsoft Gaming.

I'll talk about the concept of the one-person video game, explain how the video game industry can use recommendation systems, and we will make an announcement!

I'll share some details about the