AI Weekly Digest #7: Open Source GPT4 competitor?

Welcome to this new edition of our AI newsletter, where we bring you the latest updates on artificial intelligence!

Here’s what you must know: Open Source models are gaining moment, unlocking several scenarios previously thought to be too expensive or time consuming. Fast iterations on these models are bringing the community to better, smaller models, challenging more and more OpenAI’s ChatGPT and GPT4 models.

• The Main Headlines: AutoGPT, GPU shortage, new text-to-video models, AI wins a photography context, and Spotify takes down a very popular AI generated song! As usual we have a lot to cover!

• Beyond the Hype, the story: MiniGPT4, the open source GPT4 competitor made possible in just 4 weeks of open source iterations.

• Bonus: Current state of the art AI, and its impact on society in a great 20min video.

Beyond the Hype, The Story

Welcome MiniGPT4: Open Source Competing with GPT4? How is this even possible?

The AI community has recently witnessed the rapid development of miniGPT4, an open source AI model with capabilities similar to GPT-4.

In just a few weeks, advancements like Blip2, LLaMA, Alpaca, and Vicuna have paved the way for miniGPT4's impressive features, such as generating detailed image descriptions, creating websites from hand-written drafts, and providing solutions to problems shown in images.

Built on a pretrained ViT and Q-Former vision encoder with a linear projection layer and the advanced Vicuna large language model, miniGPT4 demonstrates the power of open source models and efficient computing resources.

The journey from Alpaca to miniGPT4 in just four weeks highlights the potential for even more astonishing breakthroughs in the future.

Read the full story on MiniGPT4 here, or check out the summary in this tweet!

Main Headlines

• GPUs: Microsoft is developing its own AI chips to reduce dependency on Nvidia amid GPU shortages and high prices. This in-house solution aims to meet Microsoft's growing demand for AI applications and cut costs significantly.

• Text-2-video: Nvidia builds a new text-to-video generative AI on top of stable diffusion models. The results are getting closer to what we would expect from a consistent video generated via AI.

• AutoGPT crossed 100k stars on Github as we speak making it one of the most popular code repositories out there. This showcases again the trending hype revolving around Autonomous AI. Yet we do believe it’s still in its early stages and not ready yet for production. But we can bet that it will improve very (very!) quickly over the coming weeks.

• AI wins a photography context: yet again AI proves that it can fool the best of us as it can enter a photography contest and win it!

  

•  AI generated Drake x TheWeeknd goes viral, and Spotify decides to shut it down! AI generation is getting better on all aspects, text, images, videos and of course voice cloning & songs!

Bonus!

Autonomous AI Agents, realistic AI music of Drake, AI generated photography wining a prestigious award?

If you think it’s all going too fast, you’re not alone!

Watch this great video by ColdFusion TV. 20 min summarizing today’s state of the art in AI, and analyzing its impact on all of us.

That’s it for today! If you made it this far, I’d appreciate a quick feedback 😋! I know there’s room for improvement! So don’t hesitate to share with me the things you liked and those that you didn’t.

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