Phi-3 does great on every benchmark, but it’s horrible in every other way.

It looks like they overfit the public benchmarks.

Goes to show how many people only care about playing stupid games to win silly prizes.

Building with AI gets easier every day.

Here is an open-source library that makes integrating AI into an application extremely easy:

github.com/CopilotKit/Cop…

Star the repository!

This library alone can make React the best front-end framework out there!

There are a bunch of

How long until Open Sora catches up with Sora?

(Has anyone outside of OpenAI seen Sora, anyway?)

By the end of 2024, open-source video generation will be mind-blowing.

I hate passwords. I think everyone does.

In graduate school, the final project for one of my classes was a solution for a password-less bank account. This is still not possible today, but we are getting close.

I use a password manager. It helps, but you still need a master

One cross-platform language to work with every cloud! This is the way it should be.

Winglang is an open-source language designed specifically for the cloud. It was created by the same guy who created the AWS CDK.

Winglang abstracts the differences between cloud providers by

How to use Gemini 1.5 Pro? (Free session!)

This is a live session that will show you how to use Gemini's new native audio, file API, JSON mode, and 1M context window features.

Friday, Apr 26th at 10 am PDT

FREE for everyone to attend!

RSVP here: events.singlestore.com/webinar-gemini…

Imagine a self-driving car that would take you home safe 99.9999% of the time, but it would kill you the other 0.0001% of the time.

Would you ride in that self-driving car or rather drive yourself?

Be careful with state-of-the-art models.

To build a model, you go through 3 stages.

First, your model sucks. You are better off making a random guess than using it.

After some care, you get it working and can show it has some predictive power. Using the model at this point is

The average MT-Bench score of Llama3 is 8.04. The new model's score is 8.25.

That's a large jump! And it only took a few days.

The name of the model is Llama-3-Smaug-8B, and it was created by the team Abacus.AI.

Here is the model in HuggingFace:
huggingface.co/abacusai/Llama…

Llama 3 hasn't been out for a week and we now have something better!

The 'MT-Bench' benchmark measures an LLM's ability to produce coherent, informative, and engaging conversations.

We now have a new Llama3-based model with a much better score.

Open source always wins!

↓

Kick-ass online machine learning courses:

• Coursera - Machine Learning Specialization
• MIT 6.S191 Introduction to Deep Learning
• DS-GA 1008 Deep Learning
• UC Berkeley Full Stack Deep Learning
• Cornell Tech CS 5787 Applied Machine Learning

It's now up to you.