TOTW 7 - The R-AI-CE to Supremacy

Hold on to your hats, folks! It's been an incredibly eventful week in the world of artificial intelligence, with major players announcing groundbreaking new technologies. Here's a recap of the latest developments in large language models (LLMs) and diffusion models, and why they matter for the future of AI.

Google's PaLM-E: An Embodied, Multi-modal Large Language Model.

On March 9, 谷歌 unveiled PaLM-E , an embodied version of their PaLM LLM. PaLM-E is a game-changer, as it introduces multi-modal capabilities, allowing it to work with both language and vision while translating these inputs into robotic actions. Next, the PaLM API was released, enabling developers to create conversational chatbots like ChatGPT, to summarize text, to write code. At the same time, MakerSuite is introduced, a quick prototyping tool, helping developers to quickly get started. As if that weren't enough, Google is bringing AI integrated features to their Google Workspace, making the most collaborative office suite even stronger.

Meta's Open-Sourced LLaMa: Innovation Through Accessibility

In a bold move, AI at Meta announced LLaMa and open-sourced the LLM, allowing it to run on a single GPU, making it accessible even for home users. What is important to notice, is that as LLMs grow in size, they develop new abilities, like mathematical reasoning and protein folding . The open-sourcing of LLaMa is expected to ignite innovation within the AI community, similar to the impact of open sourcing Stable Diffusion, resulting in gems like ControlNet and ComfyUI . Or creating a thriving community training specialized models, found on Hugging Face and CivitAI . I wonder which innovative solutions we will see in the coming months.

OpenAI's GPT-4: Their Most Advanced System Yet

微软 Germany's CTO, Andreas Braun, confirmed the arrival of GPT-4 , another multi-modal LLM last week. Released on March 14, GPT-4 is touted as OpenAI's "most advanced system, producing safer and more useful responses." The release comes just four months after ChatGPT, highlighting the rapid pace of development in this field. In response to Google's announcement, Microsoft also introduced Microsoft 365 Copilot , also integrating AI features into their widely-used office tools.

The Impressive Midjourney 5: An Overshadowed Pearl

Last but not least, the release of Midjourney 5 should not be overlooked. Midjourney is a diffusion algorithm, allowing people to create images. Though it may have been overshadowed by the previous announcements, its impressive capabilities make it a favorite for daily users like myself. The mind-blowing results achieved with Midjourney 5 demonstrate the speed and potential of these rapidly-evolving technologies.

Conclusion

This whirlwind week of AI advancements is just the beginning. With the pace at which LLMs are being developed and released, we can expect even more exciting news in the near future. As AI continues to evolve and integrate into our daily lives, it's crucial to stay informed and engaged with these groundbreaking technologies.