• AI Academy
  • Posts
  • 🤖 We Had a Quarrel with a New Sassy AI

🤖 We Had a Quarrel with a New Sassy AI

Plus, the NYT is using ChatGPT and OpenAI’s Security Breaches

July 10, 2024

Hello AI Enthusiast,

Before we jump into the latest AI developments, we've got a quick announcement to make. We're introducing a new edition to our newsletter family! 🎊

This additional newsletter will focus on AI tutorials, use cases, and experiments we've been working on. We'll share our back stories and how you can apply them in your own projects. Keep an eye on your inbox for this new content coming your way early next week.

⚠️ Also, our regular AI news roundup will now be landing in your inbox on Thursdays instead of Wednesdays.

Now, about the latest AI developments, a French company has created Moshi, a new AI that understands emotions. Unexpectedly, The New York Times is using OpenAI's technology while suing them. Meanwhile, OpenAI had to fix security issues that put users' chat privacy at risk.

Let's take a closer look at these AI stories. 👇​

The Big Picture 🔊

Kyutai Launches Moshi for Real-Time Interaction

Kyutai (we’ll avoid commenting on the name), an AI open-science lab in Paris, launched Moshi, an open-source AI model capable of understanding emotions and handling two audio streams simultaneously. Moshi is trained on both text and audio, allowing it to interact in real-time. A smaller version of Moshi can run on everyday devices, making it easy for more people to use.

Gianluca Mauro
Gianluca MauroCEO and AI Rockstar

We tested Moshi while commenting on this news. This AI is quick, so quick it kept interrupting me. My chat turned into a quarrel, and I caught it all on video. See for yourself below.

The small team behind Kyutai focused on real-time audio. Moshi's key features are local operation even on small devices, open-source availability, and simultaneous input and output processing. These characteristics likely won't be part of ChatGPT-4o's upcoming new version.

@gianluca.mauro

I had a fight with a sassy AI. It was weird. #Ai #learnontiktok #artificialintelligence #business #machinelearning #product #ux #entrepren... See more

New York Times Experimenting with OpenAI

A leak revealed that the New York Times developed a tool using OpenAI's AI for generating headlines and applying their style guide. Despite suing OpenAI for copyright infringement, the NYT experimented with AI, instructing it to "not improvise" and adhere strictly to their guidelines.

Andrea Mattiello
Andrea MattielloCM and Outdoor Fanatic

AI tech is controversial, but its potential is undeniable. The NYT, despite suing OpenAI, is using its tech to boost employee productivity. They even created an internal AI tool for styleguide during a company experiment.

NYT trying to limit the AI by instructing it in the prompt to "not use rules learned somewhere" won't prevent it from accessing the data NYT is so concerned about. Our Bootcamp addresses exactly this need, helping professionals and companies boost productivity with AI while understanding its real capabilities and limitations.

OpenAI's Security Challenges

Early last year, OpenAI experienced a security breach where a hacker accessed internal discussions among researchers, sparking fears about foreign threats like China. Additionally, recently the ChatGPT Mac app was found to be storing conversations in plain text, making them easily accessible to malicious actors; OpenAI has since fixed this issue by encrypting the data.

Gioele Mottarlini
Gioele MottarliniCOO and Image Addict

People often jump to blame AI for security issues, but AI usually has nothing to do with it. This could've happened to any tech company. This said, self-regulation by companies isn't really helping. Zuckerberg once expressed the idea that regulators should provide clear directives for companies to follow.

It’s worth mentioning that OpenAI offered ChatGPT downloads directly from their website, bypassing some of Apple's stricter rules, including EU regulations.

Bits and Bobs 🗞️

Tribal News 🫂

Last week we had our first-ever demo day for our GenAI Bootcamp, and our jaws are still somewhere on the floor. We decided to shake things up this time, giving our students a chance to pitch their projects and get live feedback. Why? Because the previous projects were so impressive, we thought they deserved their own spotlight.

Five brave souls took the stage, each presenting a unique project. Gianluca was so thrilled that he couldn't resist recording a video for LinkedIn.

Looks like we've set the bar high for future cohorts. No pressure, current students!

Monday marked the beginning of our 7th GenAI Project Bootcamp, which sold out. We met participants from various industries and countries, filling our class to capacity.

We've now opened the waiting list for our 8th cohort. The next program starts September 16th and runs for two months. It's hands-on, so be prepared to put in some real work.

By joining our waiting list, you'll get early access to subscriptions and a special discount. Ready to roll up your sleeves?

LOLgorithms 😂

We are excited about the potential of AI video generators, but there is still work to be done.

That's a wrap on our newsletter! Here’s a quick recap before you go:

Catch you next week! 👋