AI Weekly Digest #29: Gemini, EU AI Act and HeyGen Avatars

Hello, tech enthusiasts! This is Wassim Jouini and Welcome to my AI newsletter, where I bring you the latest advancements in Artificial Intelligence without the unnecessary hype.

You can find me on LinkedIn, Twitter and Medium! Let’s connect!

Now let's dive into this week's news and explore the practical applications of AI across various sectors.

Main Headlines

Here are the main trends to keep in mind if you are working in AI today.

#1 Google Gemini Models, The New State Of The Art

Google unveiled its new generation of models this week. Multimodal, more capable, and available in various sizes suited for different applications, the Gemini Family represents the new state of the art. Check out my blog post to delve into everything you need to know about this groundbreaking development!

#2 AI Video Startup HeyGen Launches Near-Instant Avatar Generator

Ever dreamed of cloning your digital self? With a realistic avatar?

Now, it’s possible in just a few minutes! All you need is a recording of yourself (2-5 minutes), and HeyGen will generate an avatar for you. You can then use this avatar to create endless videos by simply prompting it. The result is an avatar that mimics your movements and synchronizes its lips to the intended speech you've written!

And you can try it for free!

#3 Introducing SDXL Turbo: A Real-Time Text-to-Image Generation Model

Stability AI released a few months ago its new Stable Diffusion flagship image generation model: SDXL1.0!

This week, it announces a turbo version capable of generating the image in real time as you type your prompt!

#4 AMD unveils its New AI Chips: highest performing accelerator in the world 

The ongoing GPU shortage remains a critical bottleneck in AI development, with Nvidia's stocks witnessing a surge of +250% in a year (!), catapulting their market capitalization to $1 trillion.

In this competitive landscape, AMD has been actively making strategic moves to challenge Nvidia's dominance. They are focused on developing alternative solutions to mitigate the GPU shortage's impact on AI advancement, highlighting a dynamic and evolving arena in AI hardware.

"LLMs are escalating in size and complexity, necessitating vast amounts of memory and computational power. The availability of GPUs stands as the most crucial factor in the adoption of AI", emphasized AMD CEO Lisa Su as it introduced AMD's MI300X. It is now the world's highest-performing accelerator. She boldly compared it to Nvidia’s H100 chips, noting that while it matches the H100 in training Large Language Models (LLMs), it excels in inference tasks. Specifically, the MI300X outperforms the H100 by 1.4 times when operating with Meta’s Llama 2, a Large Language Model with 70 billion parameters.

This announcement comes days after OpenAI signed a letter of intent to spend $51 million on chip startup Rain AI, Wired reported, highlighting the need for a diversified source of AI chip providers!

#5 European Union strikes landmark deal on AI regulation

What you need to know for now?

1. Foundation Model Regulation: The legislation includes specific restrictions on AI foundation models but provides broad exemptions for open-source models. This could be significant for developers and users of open-source AI.

2. Risk-Based Approach: The EU's AI Act is designed around a risk assessment of software models. The greater the risk posed by an AI system to individuals' rights or health, the stricter the regulations it will face.

3. General-Purpose AI Systems: A pivotal part of the negotiations focused on how to regulate general-purpose AI systems such as ChatGPT. The agreement delineates transparency requirements for all such AI models and imposes more stringent rules for powerful models.

4. Biometric Surveillance Limits: The agreement includes regulations on remote biometric surveillance, such as facial recognition, with a ban on real-time facial recognition in public spaces, albeit with certain exceptions for law enforcement and national security.

5. Penalties for Non-Compliance: There will be a robust mechanism for monitoring and sanctions, with the potential for significant fines. The new EU AI office will have the authority to impose fines of up to seven percent of a company's annual turnover, or 35 million euros, whichever is greater.

As for the timeline, while the political agreement on the AI Act is a critical step, the law will still need formal approval from EU member states and the European Parliament.

We do not have the exact date when the regulations will come into effect, but it suggests a sense of urgency to finalize this regulation, implying that it may be implemented relatively soon after formal approval.

This is it for Today!

Until next time, this is Wassim Jouini, signing off. See you in the next edition!

Have a great Sunday and may AI always be on your side!