AI Today and Everyday — March 19, 2024

What’s new in the world of AI.

Apple and Google in talks on iPhone AI integration

Image source: DALLE-3

The Rundown: Apple is reportedly in talks with Google to integrate its Gemini AI model into the iPhone — potentially rolling out advanced AI capabilities to billions of users as soon as iOS 18 later this year.

The details:

• Gemini would enable new AI-powered functions for creating images, writing text, and more, complementing Apple’s on-device AI enhancements.
• Apple also held previous discussions with OpenAI about using its models, with OpenAI now also reportedly in talks with Samsung.
• Tim Cook has promised major AI announcements at WWDC in June, though a potential Google deal may not be finalized by then.

Why it matters: Despite pushing forward with their own AI tech, Apple integrating Gemini would be a massive power play — though OpenAI siding with Samsung would be an equally compelling counter. However it shakes out, powerful AI-integrated phones are arriving VERY soon.

NVIDIA launches world’s most powerful AI chip

Our Report: NVIDIA has unveiled the B200 “Blackwell” chip which can complete tasks 30x faster and reduce cost and energy consumption by 25x, compared to its predecessor (the H100).

Key Points:

• Manufactured by TSMC, its features include improved connectivity with other chips and a new method for processing AI data, which will speed up the process.
• Blackwell will be available as part of NVIDIAs next-generation “super-chip” lineup and is expected to be the engine powering data centers for Amazon, Microsoft, and Google.
• With Microsoft (after declaring they were reducing reliance on NVIDIA) announcing a collaboration, which will see it leverage the new chip to power its cloud platform, Azure.

Why you should care: Although NVIDIA owns 80% of the AI chip market, they’re having trouble expanding its technology beyond major cloud computing companies which it plans to address by making its AI systems more accessible.

YouTubers must now label AI content

• YouTube has announced a new policy that now requires content creators to add an ‘altered or synthetic content’ label to their videos, if AI tools were used.
• An exception to this rule is if AI was used to write scripts, generate ideas, add beauty filters, or create unrealistic content (eg. imaginary figures and specialist effects).
• Although they haven’t introduced specific enforcement measures, they’ve warned YouTubers that non-compliance could result in account suspension and demonetization.

Sam Altman speaks on the future of AI

Image source: Geeky Gadgets

The Rundown: OpenAI CEO Sam Altman just offered new glimpses into GPT-5, Sora, his vision for AGI, and more in a new interview on the Lex Fridman Podcast.

The details:

• Altman confirmed OpenAI will release ‘an amazing new model’ this year and ‘many different things’ in the coming months.
• When asked about GPT-4, Altman said it “sucks“ and that the leap in capabilities for GPT-5 will be similar to GPT-3’s jump to GPT-4.
• Altman highlighted Sora’s ability to understand and generate worlds, seeing it as a step towards AI assisting the physical world via robotics.
• He also predicted that compute will be the currency of the future and possibly the most precious commodity in the world.

Why it matters: It’s rare to get raw, long-form thoughts from one of the world’s most important CEOs. Altman speaks like someone who knows game-changing releases are coming soon — and the last time he went on with Lex, we got GPT-4.

Check out the video and transcript for the full conversation.

A ChatGPT for Music Is Here (13 minute read)

Suno’s AI model creates music from text prompts. The model creates all the music itself while using OpenAI’s ChatGPT to generate lyrics and titles. Suno’s newest model is still a few weeks from public release. A link to an example of a track generated by the AI model is available in the article.

OpenAI, GPT-5, Sora, Board Saga, Elon Musk, Ilya, Power & AGI (2 hour read)

Sam Altman recently appeared on a podcast where he discussed OpenAI, GPT-5, Sora, the board saga, Elon Musk, and more. This page contains a transcript of the podcast. Each line includes timestamps with links to take readers to specific parts of the video. A link to the full two-hour video of the interview is available.

Why Figma CEO Dylan Field is optimistic about AI and the future of design (47 minute read)

More empowering software should increase the demand for human creativity, let more people in, and let those people work better and faster.

Stability AI released Stable Video 3D, a model based on Stable Video Diffusion that creates high-quality and consistent 3D generations from a single image.

OpenAI is reportedly nearing a deal to hire former Airbnb policy chief and Clinton White House member Chris Lehane as a senior executive.

Microsoft announced an AI event scheduled for May 20 ahead of Build 2024, highlighting Windows and Surface AI integrations.

Roblox launched Avatar Auto Setup and Texture Generator, two new AI features designed to significantly reduce the time it takes for creators to develop avatars and 3D models on the platform.

The U.S. Dept. of Homeland Security announced three new pilot projects expanding its use of AI, aiming to improve its investigative processes, immigration officer training, and hazard mitigation planning.