AI Weekly Digest #1 Beyond ChatGPT & Text based models!

Welcome to the first edition of our AI newsletter, where we bring you the latest updates on artificial intelligence and its impact on our world. NoBS!

So much to cover! In this issue, we'll explore:

• The Main Headlines: David Guetta clones Eminem’s voice, Accurate AI generated images from brainwaves, Microsoft’s new multi-modal model and more!

• Hot Topics: The story of Silicon Valley Bank’s collapse and its impact on AI startups. Wild speculations about GPT4 release next week! Midjourney v5 and Dalle2 updates!

• Beyond the Hype, the story: How AI is transforming medical research. An in-depth analysis of a two year revolution enabling fast, cheap and efficient drug discovery!

Beyond the Hype, The Story

“The Most Important Achievement In AI — Ever” according to Forbes

The breakthrough of AlphaFold in November 2020 paved the way for a two-year revolution culminating in the first synthesized drug in January 2023, demonstrating the enormous potential of AI to expedite drug discovery and advance new treatments for various diseases. Read the story here:

You can also find a summary on Twitter

Main Headlines

(1) Voice Cloning: David Guetta & Emin-AI-em

Recent releases of voice cloning services, offered by companies such as ElevenLabs, have made the technology widely accessible. It’s available online, only requires a few seconds of your voice and $5/month!

Why it’s important: As usual when it comes to AI, it’s a doubled edged technology.

• On the positive side, it offers diverse applications, such as cloning your own voice for practical reasons, and recreating lost artist voices. David Guetta showcased how powerful this technologie is by leveraging one AI tool to generate a text music line, and a second AI tool to synthesize Eminem’s voice. It was supposed to be a joke and people hearing it at his show for the first time “went nuts”!

• However, on the negative side, it allows criminals and pranksters to exploit the technology for fraud and hate speech, which calls for new security practices in an era where voice identity is no longer reliable!

(2) AI generated Image based on brainwaves?!

A fascinating new study shows that AI can reconstruct an image based on a patient’s brainwaves! (Paper source).

In the experimental protocol, the researchers present an image to a patient, read the patient’s brainwaves through fMRI, and then try to reconstruct the image based on that brain signal (see image below).

The results show that stable diffusion is capable of interpreting such signals to reconstruct an image that represents similar concepts. This opens the way to medical applications that were previously inimaginable!

What if ChatGPT could also read images??

Microsoft revealed a new multi-modal ChatGPT like model, called Kosmos-1. In other words, this model can take multiple signals as input and isn’t limited to text only. Thus, it can

1. Answer textual questions and engage in chat conversations in a similar way to ChatGPT.

2. But it can also take into account images as well (without needing OCR). This means that it can both:

   • Describe an image,

   • While taking into consideration the text available in that image! (see examples in the images below).

Full Paper: you can read the full paper here.

Although the model is not currently available to the public, there are wild speculations that suggest Kosmos-1 may serve as the foundation for GPT4, which is set to be released soon (more on this topic below!)!

(4) US Copyright Office won’t protect AI generated images

Why is it important? the document announces two decisions:

• Images generated by AI, and not significantly altered by the author (via Photoshop for instance) are not protected by copyright in the US.

• Comic book’s text and image arrangement are protected by US copyright since they involved a human effort.

It is uncertain how this decision will impact artists in the US, and it remains to be seen whether similar decisions will be made in other parts of the world. The issue of copyright for AI-generated content is still unfolding! We’ll keep you updated ;)!

Hot Topics!

The end of this week has been shaken by several announcements!

(1) HOT: GPT4 to be released soon?

GPT4 is to be released soon and should be a multi-modal model, namely, capable of handling more than text. Speculation assume it will be based on Kosmos-1 (described above)! We should now more next Thursday March 16th!

(2) HOT: SVB, the bank holding 50% of US VC-backed Startups has collapsed…!

This story is unraveling as we speak!

Why is this important?

• Short-term: without support from the US government, many startups could miss payroll this month and announce layoffs

• Long-term: Startups play a key role in innovation and highly skilled job creation, and both will be severely impacted if the issue is not mitigated in the coming days.

AI innovation is fueled by VC-backed startups. We can imagine that many of them will see an end to their promising journey unless bold announcements are made next week. Stay tuned!

The full story in this tweet.

(3) HOT: Image generation models - updates! Midjourney-V5 and Dalle2 releasing new versions soon!

• Midjourney-v5 coming soon! has unveiled a new website for its paying members that enables them to provide ratings for images generated by their upcoming AI model, "V5"! These rating will be then be used by to help decide whether V5 should have a different style from V4 or not!

  

  Midjourney-v5 discord announcement.

• Dalle2 experimental v2: OpenAI is currently testing an experimental DALL·E 2 model with a small group of users for early feedback. The experimentation will last for a few days and feedback will be taken into account to prepare the coming release.

  

  source

Release dates are not available yet, but early feedback indicates that both model represent a substantial improvement with regard to their respective previous version! Stay tuned!

That’s it for today! If you made it this far, I’d appreciate a quick feedback 😋! This is the first attempt and there is 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!