Honey Tree
Posts
🍯 Microsoft Competes with Itself (and OpenAI)

🍯 Microsoft Competes with Itself (and OpenAI)

PLUS: AI-music enters the Grammy's conversation and 24-hour sport coverage

Honey Tree
June 20, 2023

Welcome to Honey Tree, the newsletter that makes learning about AI sweet and simple. The world of AI can be fast-paced and confusing, so we’re here to help.

And the best part is we’re all natural.

Plus, if you’re a student or looking for a new job, check out our Honey Board at the end of the newsletter for open jobs and internships!

Here’s what what we got for today:

🤖 Meta and Microsoft Launch New Updates
🎤 Grammys Sets Rules for AI-Generated Music
⚽️ 24-Hour Coverage of the European Games Using AI
🤣 Sweet Memes
🍯 Honey Board

🤖 Who’s in the Lead? Microsoft or Meta?

The AI bubble is still going strong, and companies are capitalizing by launching new products fast.

Among them, we have Meta and Microsoft which both launched new products last week - here’s what’s going on:

Meta releases Voicebox: Deepfakes are becoming more and more realistic (just take a look at this video of NOT Morgan Freeman), and Voicebox will only make them better.

Meta’s new tool is a generative AI model that can help with audio editing and styling through

In-context text-to-speech synthesis: Voicebox can clone voices from audio samples.
Speech editing and noise reduction: Instead of rerecording speech, Voicebox can erase audio that’s interrupted by noise (like a dog barking or car honking) and fix mispronounced words.
Cross-lingual style transfer: When given a sample of someone’s speech in English, French, German, Spanish, Polish or Portuguese, Voicebox can produce an audio translation in any of the other languages.

Microsoft launches Orca: Although Microsoft has invested heavily in OpenAI, the company is still launching AI software on its own…

Microsoft recently launched an open-source model called Orca which learns by imitating large models, like OpenAI’s GPT-4.

Orca’s advantage: Since Orca’s a small model, researchers can use it independently ”without relying on a large data center” - and so, it’s easier and cheaper to use.

🎤 AI-Music Wins a Grammy

Over the last year, we’ve seen tons of AI-generated music, like Drake & The Weeknd’s Heart On My Sleeve and Kanye West’s cover of Somebody That I Used to Know.

Even Grimes launched Elf.Tech to allow anyone to create music using her voice in exchange for 50% of the profits.

Now, The Recording Academy - which is the organization behind the Grammy Awards - has announced that AI songs are eligible for awards.

Here’s the catch: There must be proof that a real person meaningfully contributed to the song. For example, in a songwriting-based category, the music would need to be written mostly by a human.

What happens if the writer uses AI to brainstorm and “help” write the music?

⚽️ 24-Hour Sports Coverage with AI

Up at 3 AM with nothing to do? Check out the European Athletics Team Championships in Silesia from June 20-25 — which will have 24-hour audio coverage.

Here’s what’s going on: With the use of AI, European Athletics has cloned the voice (using Meta’s Voicebox?) of former athlete and commentator Hannah England, enabling 24/7 coverage.

The goal is to expand coverage “further, making the sport available to everyone and customised to their needs”, according to European Athletics’ Chief Executive.