Gm. Fed head honcho Jerome Powell is scheduled to make his annual speech at Jackson Hole today. Will he finally signal the end to interest rate hikes that people have been predicting since last year’s Jackson Hole address? Or will he continue to attack inflation like it owes him money? Tune in at 10 AM to find out.

— Stephen Flanders

🏠 AI

Meta’s New AI Models

When it comes to AI contenders, nobody pumps out more open-source models than Meta. It’s kind of their whole schtick.

And Team Zuck did it again this week with two fresh models.

SeamlessM4T

The first model is known as SeamlessM4T, and it’s been imbued with the ability to translate and transcribe almost 100 different languages via text and speech.

It sounds a bit gimmicky, but it’s a huge breakthrough for AI-powered translation services.

With SeamlessM4T, it is possible to put a headphone in and converse with somebody speaking a totally different language. The model will simply translate for you in real time.

But the model isn’t perfect. It still overuses the masculine form, is a bit toxic about translations involving sexual orientation and religion, and isn’t quite consumer-use-ready for long-form translations.

However, with some improvements, SeamlessM4T might one day act as people’s personal Rosetta Stone.

Code Llama

The second model is none other than Code Llama, which is basically Meta’s version of GitHub’s Copilot.

In English, Code Llama has the ability to write code in Python, C++, Java, PHP, Typescript, C#, and Bash.

Now, when I say writing code, I’m really saying that it will help coders write code. In other words, it’s great at debugging and finishing code, not writing an entire codebase.

It’s a copilot, not the pilot.

And tools like this are extremely popular among devs, with 70% of coders now using or thinking about using AI tools, which makes sense considering AI tools are helping devs code 55% faster.

That’s a lot of potential users for Code Llama.

The Bottom Line: Meta is on an AI tear right now. Real-time translation and devs shipping code faster? Sign me up.

🏠 SCIENCE

A Mind-Blowing Achievement

Welcome to the coolest story of the week.

Ann was a normal 30-year-old woman with two kids and a husband. But then she suffered a severe brainstem stroke that left her so paralyzed she couldn’t even breathe at first.

Even after five years of physical therapy, Ann couldn’t move her facial muscles enough to speak.

So researchers tried something new out on her.

First, they implanted a paper-thin rectangle of 235 electrodes on the sections of her brain critical for speech. These electrodes would then intercept the brain signals that would have normally fired up had Ann been able to speak. Then, a cable was plugged into Ann’s head that transports the signals to a bank of computers.

This computer then uses AI to translate Ann’s brain signals into words and project them onto a digital avatar that literally talks in Ann’s voice.

In essence, this brain chip, AI, and digital avatar combo is giving Ann the ability to speak again.

Told you this was the coolest story of the week.

The Big Picture: Brain chips have already given paralyzed people the ability to walk again, and now it’s giving people the gift of speech. There’s a reason companies like Neuralink are able to raise so much money.