AI Weekly Digest #12: AI adoption, integrating your favorite tools

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

• The Main Headlines: Big Companies accelerate AI integration across the board (Adobe, MS, Google, OpenAI, Meta) & Open Source initiatives can now compete head to head with chatGPT!

• Beyond the Hype, the story: Diagnostics and drug discovery, your doctor might be an AI sooner than you think.

• Bonus: Blog to Video, the ultimate tool for modern content creators!

Beyond the Hype, The Story

The future of healthcare is being shaped as we speak! Here are 5 amazing news that are reshaping this sector.

• Human Level Diagnostics: The competition is raging between OpenAI & Google. GPT4 and Med-PaLM2 both score very high as a medical diagnostician (rivaling the abilities of the best doctors today). What does this mean about the future of healthcare? With current shortage of doctors, can we envision a world where you’ll go through an AI first?

• Automated Drug Discovery: now possible thanks to AlphaFold, read about this medical revolution enabled by AI, “The Most Important Achievement In AI — Ever” according to Forbes.

• Digital Bridge enabling a man to walk again! watch video here.

• A chip in your brain soon? FDA approved human trial for Elon Musk’s Startup Neuralink. They can now implant AI chips in your brain!

• Brainwaves to video: AI is learning fast, and it can now approximate what you are imagining in your brain (e.g., people walking, animals you are imagining, and so on). This new experiment can generate videos based on your brain’s fMRI input. Read about it and watch demos here.

Main Headlines

Photoshop released its new generative AI feature, and it’s a game changer for photographers, illustrators and artists in general. You have to see it in action! Although the technology itself isn’t new (we’ve been seeing inpainting and outpainting features since Dalle2), this demonstrates the power of integration of such features within mainstream tools.

Microsoft Build: Several announcements (summarized here), but the main one is probably the integration of AI at the OS level for Windows 11, “where you can ask it to summarize, rewrite, and explain text in any of the apps you’re using, as well as adjust your computer’s settings“!

Google Ads supercharged with AI: Google is introducing a new natural-language conversational interface within Google Ads to simplify campaign creation. “Simply add a preferred landing page from your website and Google AI will summarize the page. Then, it will generate relevant and effective keywords, headlines, descriptions, images and other assets for your campaign”. Performance Max, powered by AI, has shown to increase conversions for businesses on average by over 18% at a similar cost per action​​.

OpenAI:

• Sam Altman in Europe meeting political leaders - OpenAI vs European Regulation.

• ChatGPT updates: it now requires you to validate your age before using the service in an attempt to become compliant with European new regulations. ChatGPT browing feature relies on Bing AI now. You can also share a conversation with others (e.g., trip plans, data analysis, and so on).

• $100k Grants: OpenAI announced a program to fund experiments in setting up a democratic process for deciding what rules AI systems should follow. They are offering 10 grants of $100,000 each to teams that can develop proof of concepts for this process.

Meta: This is probably the news of the week. Speech to text, and text to speech covering 1100 languages & dialects. This model covers dialects for which it was impossible to build a strong dataset ! But with Meta's huge conversational base, it became possible! Looking forward to playing with this one !

Open Source & Research: 

• Meet Falcon, the new Open Source LLM leader: a new 40B parameter model beating all other LLMs! The competition is still raging. What you need to know?

  • Fully Open Source, even for commercial use

  • It matches closed models such as Bard, ChatGPT & Claude according to authors.

• QLoRA: a new approach reduces drastically the memory usage needed to train large language models. What does it mean for us? We should be able train state of the art ~30B-65B parameter models, in less than 12h, with a single GPU! HUGE! This means we could have our own Falcon models running at home!!

Bonus!

Blog to Video, the ultimate tool for modern content creators! Here are Three tools to try!

• Lumen5: https://lumen5.com/blog-to-video

• FlikiAI: https://fliki.ai/use-cases/blog-to-video

• Synthesia, news like videos: https://www.synthesia.io/tools/blog-post-to-video

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!