Forward Future News Episode 3

tl;dr:

• AI Fears Grow - The Open Letter

• Local LLMs Everywhere!

• Musk Tried to Take Over OpenAI

• Dalai Updates

The Open Letter

Over 1,100 AI researchers and thought leaders signed an open letter asking the AI industry to “pause” AI development beyond GPT-4 for 6 months. Check out the letter here. The letter, signed by Elon Musk, Woz, and many of the most prominent names in AI lays out a clear argument for why we need to slow down and think about the implications of training models more advanced than GPT-4 and what that could mean for society.

There are so many angles to view this letter from. The obvious one is that of course OpenAIs competition wants it to stop development for 6 months, they are so far ahead! This would give its competitors time to catch up. The thing is…that’s true, but it’s also true that taking a minute (6 months) to pause and come up with a framework to guarantee AI alignment would be a good thing.

Take a look at the full video I made on the open letter:

Local LLMs are Everywhere!

I am incredibly happy to see a flourishing open-source community, building LLMs that anyone can install on their local computer. This is the best way to distribute the power of AI so it’s not in the hands of a few powerful closed-source companies like OpenAI. This also allows people to customize their AIs how they want.

This week, GPT4ALL was released, which takes an Alpaca-like approach to fine-tune the LLaMA model using GPT-3.5 Turbo. It’s super easy to install on your computer and only takes 4 GB of space. It works surprisingly well, especially compared to Alpaca, which was pretty good but not great.

You may be asking yourself: if it was trained on GPT-3.5, is it legal? Well, the creators of GPT4ALL believe they read through the terms of service and found a loophole that allows them to train the model using GPT data.

Interested in installing it yourself? Here’s how:

Musk Tried to Take Over OpenAI

A story that came out this week revealed that in 2018, Elon Musk tried to take over OpenAI from Sam Altman. The story is fascinating and explains the tension between these two tech leaders.

Before OpenAI released ChatGPT to the world and became the fastest-growing product of all time, they were building for years without much to show for it. At one point, Elon believed they had fallen fatally behind Google in the AI race and attempted to usurp control from Altman.

Here’s the full story of how it went down:

Dalai Keeps Getting Better

Dalai, the UI interface for local LLMs, continues to get better! After Meta, the company behind LLaMA, realized that their leaked model was powering Alpaca, they started sending cease and desist letters to different developers, including Dalai.

However, the creator, Cocktail Peanut, is a staunch believer in AI being open-sourced and decided to make Dalai and Alpaca distributed through Torrent architecture. This means there’s no way to “take down” the distribution of Alpaca and Dalai. What an awesome story.

Additionally, Dalai keeps getting better by the day. A bunch of new features was released (such as prompt templates) and the UI looks better. If you haven’t had a chance to install Dalai LLaMA locally, here’s how you do it:

Thanks for reading and I’ll see you next week!

PS. I’m ALWAYS open to feedback, so feel free to drop a comment and let me know how I can improve this newsletter.