• Everyday AI
  • Posts
  • OpenAI Sora's data training: Should we be worried? 😬

OpenAI Sora's data training: Should we be worried? 😬

Apple's MM1 model, Sam Altman exclusive interview, xAI live on GitHub, and more!

Author

Everyday AI
March 18, 2024

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Should we be concerned about OpenAI Sora’s data? Is Apple finally creating an AI model? Here’s this week’s AI news that matters. Give it a listen.

🕵️‍♂️ Fresh Finds: Sam Altman talks Elon Musk in exclusive interview, AI that helps you launch your startup in days, what to expect at GTC 2024, and YouTube adding AI-generated content label. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Apple’s MM1 multimodal model, OpenAI Sora data training concerns and xAI now live on GitHub. For that and more, read on for Byte Sized News.

🚀 AI In 5: Can you ACTUALLY make a video with only text using this GPT? Find out here

🧠 AI News That Matters: Wondering why you should care about AI models plagiarizing and fridges powered by AI? We’re breaking down the latest AI news and trends. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about AI and neurodiversity, Morgan Stanley adding a Head of AI, and a Microsoft Copilot notebook review. Check it here!

AI News That Matters - March 18th, 2024 📰

Should we be worried about what OpenAI said about Sora?

What does Google's new video game-playing model mean about AI agents?

Is Apple finally creating an AI model?

There’s so much going on this week in the world of AI.

Luckily we’re breaking down the AI news that matters this week and why it’s important to you!

Join the conversation and ask Jordan any questions on AI here.

Also on the pod today:

• Figure 01, by OpenAI and Figure 🤖
• NVIDIA GTC Conference 🟢
• Apple's New Large Language Model, MM1 🍎

It’ll be worth your 33 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Upcoming Everyday AI Livestreams

Tuesday, March 19th at 7:30 am CST ⬇️

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – UNUM is an AI-powered social & bio link tool, Upscale.media Plugins gives you AI image upscaling plugins, and NextStarterAI helps you launch your startup in days.

Trending in AI – NVIDIA’s GTC is kicking off today, here’s what you can expect.

(FYI we’ll be there too! Sign up here to attend virtually or in person and a chance to win an NVIDIA GeForce RTX GPU!)

Big Tech - Sam Altman sat down with Lex Fridman to talk Elon Musk, Sora, and more!

AI Tech – Qualcomm is creating an AI chip to bring GenAI to more affordable phones.

Social Media - YouTube is adding an option to label AI-generated content in your videos.

AI in Politics – The US House of Representatives’ AI Task Force had its first meeting. Here’s what they talked about.

1. Apple Researchers Unveil Multimodal Models 🍎

Apple researchers have introduced the MM1 family of multimodal models, boasting an impressive 30 billion parameters. These innovative models have set a new benchmark in performance across a wide range of metrics, signaling a significant leap forward in the field of multimodal understanding.

2. OpenAI’s CTO Avoids Sora Data Training Questions 🧐

Mira Murati, OpenAI's CTO, dances around questions about Sora's data sources during a sit-down with The Wall Street Journal. From murky social media hints to licensed data nods, the $80 billion company's upcoming video-generating AI model remains shrouded in mystery.

3. Apple Talks with Google & OpenAI for Chatbot Integration 🤯

Apple is in discussions with Google and OpenAI to incorporate advanced chatbot technology into the upcoming iOS 18! This potential collaboration signifies Apple's efforts to step up its AI game and compete with industry leaders. If this deal goes through, it could revolutionize the way users interact with their iPhones, setting a new standard for artificial intelligence integration in mobile devices.

4. Elon Musk's xAI Drops Massive AI Bombshell on GitHub 🚀

Elon Musk just shook the AI world by open-sourcing Grok-1, a 314 billion parameter Mixture-of-Experts model, allowing devs to tinker with its base model weights and network architecture. This move comes after Musk's acquisition of Twitter and his push for AI transparency, taking a direct hit at rivals like OpenAI.

5. Homeland Security Explores New Generative AI Technology 🇺🇸

Homeland Security is diving headfirst into the AI world with three new generative AI pilot projects, aiming to enhance investigative reports summarization, immigration officers' training, and hazard mitigation planning. This move highlights the agency's commitment to leveraging cutting-edge technology for improved efficiency and effectiveness in various critical areas.

This GPT allows you to create videos with JUST text?

Can you really make a complete video for free with a GPT?

Veed GPT allows you to turn text into video inside ChatGPT.

Is it worth your time? We try it for you!

Or see this related video:

1 – Google DeepMind announces SIMA 🎮

An AI that plays video games? 

Yup. 

Google DeepMind has announced a new AI model, called SIMA, that can play video games based on natural language instructions.

The model has been trained with the help of 8 video game studios and can perform over 600 basic skills, making it more efficient than models trained on one specific game.

What it means: 
Here’s a hint — this isn’t about video games. Not at all. 

As we talked about on today’s show, we see SIMA as Google Deepmind’s first foray into agents, or AI-powered tools that can perform tasks autonomously without human intervention.

If the ultimate goal is for SIMA to be able to play video games by itself, you have to realize the bigger picture that SIMA is a proving ground for future autonomous agents to be used outside of video games. 

2 – Cringy OpenAI response on training data 😬

Welp. That was kinda awkward.

OpenAI has been under some increased criticism lately after its CTO Mira Murati kinda avoided questions about Sora’s training data in a recent interview.

If you’re a bit confused — here’s the basics of Sora, like we talked about on today’s show. 

Sora is a not-yet-released GenAI text-to-video model from OpenAI.

Right now, only internal employees and a select few visual artists have access. And the results — crazy. 

In an interview with the Wall Street Journal, Mira got a bit hesitant on direct questions about how OpenAI trained its Sora model. 

Watch for yourself: 

What it means: 
We talked about this in-depth, but Sora is SUPER impressive. (And not just for its actual AI video outputs.) 

But one thing the collective world has asked was — how?

How did OpenAI seemingly put out higher quality AI video outputs than other companies like Pika Labs and Runway that were more than a year ahead? 

We expect this to open a Pandora’s box into how OpenAI actually trained Sora.

And, that could lead to more lawsuits. (Like the whole New York Times vs. OpenAI lawsuit thing.) 

Go check out this episode for more on Sora. (What no one’s actually talking about.) 

3 – Apple (kinda) announces its LLM 🍎

Well… that was anticlimactic. 

Apple sort of released details on its new LLM, or more specifically, its Multimodal Large Language Models. 

Apple researchers released a paper detailing their new (and much anticipated!) model, dubbed MM1. 

Here’s the gist – 
Apple’s been working feverishly behind the scenes on its internal GenAI and LLM offerings, reportedly spending millions of dollars A DAY on its own AI tech. 

And instead of the traditional Apple release, bundled up in secrecy and a big in-person reveal, we instead got an impressive research paper. 

The not-yet-public model comes in three sizes (3B, 7B, and 30B parameters) and outperforms most competitors in tasks such as image recognition and natural language reasoning, but it still lags behind Google and OpenAI's models.

What it means: 
Will Siri finally become super smart? We think so.

Whatever Apple unveils at its WWDC conference in June, the expectations are that it will be AI-heavy. 

Will we see MM1 as an edge device offering a super smart Siri and iPhone with a MLLM built-in?

This paper could be the first sign that that’s where Apple is headed. 

4 – Figure demos ChatGPT-powered humanoid bot 🤖

Well, that’s equal parts scary and cool AF.

The OpenAI-backed robotics company Figure released a reportedly real-time demo of its humanoid bot, Figure 01, completing real-time tasks. 

(Better than me, TBH). 

And it’s CRAAAAAZY impressive. 

Our livestream audience agreed. 

"Figure 01" is equipped with advanced AI and visual recognition capabilities. It showcases its ability to perform tasks and engage in conversation with a human, potentially revolutionizing multiple industries.

The Figure 01 robot is equipped with a new Visual Language Model (VLM) that allows it to smoothly interact with humans and perform tasks simultaneously.

What it means:
We talk a lot about 2024 being the year of both GenAI implementation and also AI agents.

This Figure/ChatGPT mashup is a culmination of both of those.

Like we talked about on today’s show, we can’t emphasize how impactful this partnership could be to our everyday lives if development continues to go in a positive direction. 

Check it out for yourself: 

5 – NVIDIA GTC is here 🤩

And so are we. 

After a five-year hiatus, the in-person NVIDIA GTC conference is back in action.

Today, make sure to follow us on LinkedIn and Twitter, as we’ll be reporting LIVE and in real-time from NVIDIA CEO Jensen Huang’s keynote address. 

The event will bring together AI leaders and top companies across various industries to showcase breakthroughs and advancements in AI.

What it means:
NVIDIA has powered not only the Generative AI movement, but also the U.S. economy at large. (We talked about this almost a year ago.) 

Whatever happens this week at GTC is a signal of what’s next in business. 

Don’t worry – you can attend virtual for free.

Find out that info here, and even enter to win a GPU from our partners at NVIDIA. 

From Our Partners

We're going to be broadcasting live all week at the NVIDIA GTC Conference in San Jose!

To celebrate, we're partnering with NVIDIA to give away a GeForce RTX 3080 Ti chip!

(You know, the GPU chip you need to run their new and free Chat With RTX chatbot.)

AND to make it even better, we’re also giving you a chance to win credits for NVIDIA’s Deep Learning Institute.

First, register for the FREE virtual NVIDIA GTC Conference or buy tickets to the in-person event and then fill out this form here!

You will then be entered to win. Simple as that.

Winners will be announced later this month.

⌚

Numbers to watch

7%

Alphabet shares went up 7% after news of Apple potentially using Gemini for iPhones broke.

Now This …

What are your thoughts on this week’s news?

Vote to see live results

What are you most intrigued by?

Login or Subscribe to participate in polls.

Reply

or to participate.