🍿 Microsoft Will Hold Another AI Event

Good morning, individuals.

Welcome to our 144 new subscribers since Friday. For those who are unaware, this is the best AI newsletter in existence.

In the email today:

• How Congress will consider AI copyrights ⚖️

• Microsoft to hold “future of work with AI” event 🍿

• Headlines 📰

• Tool of the Day 🔨

• Tweet of the Day 🦅

• Links 👀

On to it.

Congressional Research Service Report Examines AI and Copyrights ⚖️ 

It’s tough to find impartial actors when it comes to the AI copyright debate.

Content creators who generate “new” works via AI want copyright protection for their “creations.”

Artists, writers and others who have already organically created works want the Federal government to protect their creations from AI pirating— training AI datasets on their work and then generating new outputs based on said training.

The US Copyright Office, courts, and Congress will ultimately have to deal with this issue.

That’s why last week the Congressional Research Service — a “nonpartisan shared staff to congressional committees and Members of Congress” — prepared a report for Congress to summarize the issue.

Unsurprisingly, this report is dense and, at times, a little dull. But there are important considerations to be culled, which we have done for you here:

• The use of generative AI programs raises questions about who holds the copyright to works created using these programs.

• The Copyright Act protects "original works of authorship," but the US Copyright Office recognizes copyright only in works "created by a human being."

• As we described recently, Kris Kashtanova registered a copyright for her graphic novel, which was illustrated entirely with images generated by Midjourney. The Copyright Office determined that the images did not have a human author and were not copyrightable.

• It is uncertain whether copyright law will recognize protection for generative AI outputs, but it is worth noting that the Copyright Office is not the final authority on US copyright law. Applicants can ultimately challenge Copyright Office decisions in U.S. District Court.

All of this naturally leads to the inevitable question: If the Copyright Office eventually grants copyright protection to AI-generated works, who owns the copyright?

The AI “creator” will naturally claim credit, as they “created” something new by feeding the AI prompts.

So too will the artists whose images helped train the AI.

Then you’ll hear from the companies that create the AI (Midjourney, Stable Diffusion, OpenAI, etc.). Do they want a piece of the pie, or is their service fee enough for them?

The remainder of the report deals with the to-date unanswerable question of whether AI itself is actively violating existing copyrights by its trained practice of taking existing works, copying, and repackaging them.

• AI systems are trained to create artistic works by analyzing large amounts of data, including existing works like text and images from the internet.

• The training process involves making digital copies of existing works, which poses a risk of copyright infringement.

• The reproduction of entire works or substantial portions thereof is almost inevitable in this training process, according to the U.S. Patent and Trademark Office.

• OpenAI, for one, acknowledges that its programs are trained on copyrighted works obtained from publicly available datasets.

• Ultimately, though, making copies of copyrighted works without permission may infringe upon the copyright owners' exclusive right to reproduce their works.

There is no doubt in our minds that this question will make its way out of the administrative world of the Copyright and Patent Offices and into the legal realm, beginning in the Federal District Courts.

Headache-inducing stuff.

Microsoft to Hold Another AI Event 🍿

From The Verge:

Getyo popcorn ready.

Headlines 📰 

Doctors in Hungary are using AI to improve breast cancer detection: None too soon, as there is at least one case already where two different human radiologists failed to identify a possible cancer marker that AI caught. 🩻 

AI is also a new tool for detecting Alzheimer’s Disease from routine brain imaging tests: Doctors at Massachusetts General Hospital are cross-comparing brain MRI’s from with and without Alzheimer’s to sharpen diagnostic evaluation. 🥼 

China’s Science and Technology Minister concedes that his nation finds keeping up with ChatGPT “very difficult to achieve”: At the same time, it remains unclear how the new technology will account for China’s ongoing internal censorship and technological restrictions. 🤷‍♂️ 

62% of surveyed insurance carriers have already cut staff after implementing AI: Underwriters and analysts cannot keep up with AI’s speed and are already on the chopping block. 🧑‍💼 

Quora’s Poe AI is now available for Desktop: Poe has been lauded by AI Twitter for its usefulness and speed. It allows access to multiple chatbots, including Anthropic’s Claude and the fastest instance of ChatGPT. 🪶

Tool of the Day 🔨

Scribble Diffusion.

Fascinating! We try not to feature ultra raw, gimmicky-type tools here, but we’re making an exception today for Scribble Diffusion.

The tool allows you to scribble AND describe the image you would like to see, and then it creates the output of it. The results are mixed, but you can see the beginnings of a really compelling use-case for image-generation apps with this.

Simple. Effective. Fun.

Get your eighth grade scribbles out of the way early:

Tweet of the Day 🦅

Character.ai is decidedly cool. It lets you hold an imaginary conversation with famous people and characters. Like our Tool of the Day, the results can be very mixed, ranging from the mind-blowing to the very generic.

Whatever your experience, the business model is unclear and the app seems to be built atop a large language model API. In other words, how on Earth can this justify a $1 billion valuation?

Meanwhile, Stability AI (Stable Diffusion) is raising money at a $4 billion valuation, after a $1 billion valuation in October. Frothy, but as one of the image-generation market leaders, this at least makes some sense. Friendly reminder that Mark Zuckerberg once acquired Instagram for $1 billion.

Links 👀

• LSU gymnast and NIL mogul Olivia Dunne put the university in a difficult spot when she promoted an AI coursework tool on TikTok 🤸‍♀️ 

• Given the repetitive nature of human emotion, it’s not surprising that AI is getting better at reading feelings by evaluating data 👀 

• Only 10% of employees are using AI at work, a number that their bosses deem far too low 🖥️ 

• Bad actors are using AI voice technology to scam people by mimicking the voices of loved ones purportedly in distress 😢