🇪🇺 EU's AI Crusade: Progress and Protection

Hello AI Enthusiast,

As AI continues to weave into our daily lives and work, it's time to reflect on how this impacts us all. The way humans and AI work together is changing fast, and it's something we can't ignore.

Look at the recent headlines: The EU's new Artificial Intelligence Act, Meta's Purple Llama project, and MIT's AI governance papers are reshaping regulatory practices. These steps are crucial for making sure AI is used safely and wisely.

And it's not just about rules. Companies like Microsoft are joining forces with labor groups to make sure people are ready for an AI-powered future. These initiatives show how AI can help us in real-world jobs.

But let's not forget the other side of the story. The recent incident at Sports Illustrated is a reminder that AI is a powerful tool, but it needs to be used responsibly.

That's where we come in. Our Master in Prompt Engineering course, starting January 8th, is tailored to blend your creativity with the practical power of AI. We're here to equip you with the skills to work on AI projects and stay ahead in this rapidly changing world. But you'll need to act fast – we've only got three spots left!

Make a resolution to learn and innovate with AI this new year.

Now, let's move on to our News Bytes section and explore more about these AI developments and beyond.

News Bytes 🗞️

• French AI startup, Mistral AI, raised a whopping $415 million in its Series A Funding¹ round, giving it a solid $2 billion value. Celebrating this, they've launched a clever new model called Mixtral 8x7B and opened a developer platform for folks to access and play with their AI models. Is OpenAI seeing a new rival in town?

• EU lawmakers have made a deal on the EU's new Artificial Intelligence Act. It's aimed at protecting people’s rights while still allowing for tech advances, by putting some limits on AI use, though governments get a pass on some of these rules. This big step could change how AI is made and used, affecting tech companies, governments, and everyone's rights.

• Google launched Gemini, its most advanced Multimodal AI² model, last week. However, after claiming it outperforms ChatGPT-4 on numerous benchmarks, Google revealed that its impressive demo was scripted. The AI model didn't perform the tasks live as implied, but rather worked through a series of carefully tuned text prompts with still images. Also, we tried to perform the same tasks with GPT4, and the results were astonishing…

• Microsoft is partnering with a major labor group, to demystify AI's future and quash fears of job replacement. The alliance includes AI learning workshops, niche AI career focus sessions, and labor summits to collect feedback. Amid growing concerns about tech rendering jobs obsolete, this alliance aims to share detailed AI knowledge, integrate worker inputs in AI development, and influence policies supporting worker technology skills.

• The UK Judicial Office's approval of AI tools like ChatGPT for legal rulings brings out a contradicting reality where some sectors restrict AI use due to fears of accuracy and data privacy. Notwithstanding these potential risks, the approval emphasizes how these tools can enhance our efficiency, but it also urges us not to let total reliance blur critical checks.

• And while some are leveraging the power of AI to enhance people some others are misusing it. After Sports Illustrated, under The Arena Group, got caught publishing AI-generated articles under fake names, their CEO, Ross Levinsohn, was let go. This incident spotlights the potential misuse of AI, underscoring that it should be used to empower human creativity and productivity rather than attempt to wholly substitute human involvement.

• Meta launched Purple Llama, an open-source project with tools to make AI development safer. It includes CyberSecurity measures for checking risks and 'Llama Guard' for cleaning up unsafe outputs. Using an open-source evaluation system will simplify following safety rules.

• Meta is allowing users to try out its AI features on the Meta Ray-Ban smart glasses, which include a multimodal AI assistant capable of describing objects and providing suggestions. While the test period is limited to a small number of people in the US, this move reveals Meta's efforts to integrate AI technology into its wearable devices.

• A committee of MIT leaders and scholars has released a series of policy papers on AI governance with the aim of helping policymakers create better oversight of AI in society. The papers suggest using existing regulatory agencies and legal liability frameworks to regulate AI and emphasize the importance of defining the purpose and intent of AI applications in advance.

• Social Media platforms are focusing on AI-generated images. Meta has launched a standalone image generator with visible watermarks indicating they were created with Meta AI. Through its testing, Meta is aiming to make AI a core part of its services.

• Meta is not the only one. Snapchat is releasing new AI-powered features for Snapchat+ subscribers, including the ability to create and send AI-generated images based on text prompts.

• StableLM has launched the StableLM Zephyr 3B, an AI assistant model that offers accurate and fast text generation on regular hardware. This smaller model allows for the development of capable AI tools on more average hardware setups, opening up possibilities for personal tools on devices like phones and laptops.

This week’s glossary 📖

1. Series A Funding: This is a level of investment aimed at early-stage startups that are beginning to generate revenue but are not yet profitable. It follows initial seed funding rounds.

2. Multimodal AI: A type of AI system that can process and understand multiple types of input data, such as text, images, and sounds, to perform tasks or make decisions.

Educational Pill 💊

Multimodal AI

Multimodal AI systems, exemplified by Google's recent Gemini launch, mark a significant leap in artificial intelligence, integrating multiple types of data like text, images, and audio for richer, more context-aware interactions. Unlike unimodal AI, these systems offer enhanced understanding and accuracy, crucial for complex tasks. These advancements give us a glimpse of how we’ll change the way we interact with technology, from search engines to smart devices.

From our community 🤝

A touching share from one of our students as the third edition of the Master in Prompt Engineering course nears its end. ❤️

That is the end of our newsletter.

Remember, if your company is looking to implement AI technologies, we also offer customized corporate training.