Good morning. Yesterday, a group of authors filed a copyright infringement lawsuit against Anthropic. Just a few hours later, OpenAI announced a content licensing deal with Conde Naste (the publisher of Wired, The New Yorker and more).

That marks 11 publishers — AP, Axel Springer, The Atlantic, Dotdash Meredith, FT, LeMonde, NewsCorp, Prisa Media, Time and Vox Media — who’ve signed deals.

Other organizations remain committed to battling it out with OpenAI in court.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

• 🏞️ AI for Good: Predicting water pollution before it happens

• 👨‍💻 New OpenAI feature allows users to customize GPT-4o

• 🛜 Amazon pulls AI-generated books after Deep View inquiry

• 🏛️ In Wyoming, an AI is running for mayor

AI for Good: Predicting water pollution before it happens

Last year, British IT company CGI launched a pilot program in southwest England meant to predict water pollution before it happens in an effort to prevent it. 

The details: The project tied together data from a variety of sources — satellite imagery data alongside data gathered by sensors planted in fields and rivers — with an artificial intelligence model. 

• The model was trained to predict when the local river would be most vulnerable to “pollution events.” Based on those predictions, local actors — farmers, for example — could hold off on using chemicals or fertilizers that would seep into the river. 

• The AI tool was intended to act as a big step toward cleaning up the river system in the North Devon Biosphere Reserve. 

CGI said the model performed with greater than a 90% accuracy rate in test preceding the pilot program. 

Fix joint pain with this powerful protein

Staying healthy and feeling good are not easy. And they get a lot harder as we age. 

Collagen, for example, supports joint health, bone strength, heart health, skin health and muscle mass. But, starting in your 30s, the body just starts to produce less and less of it (sometimes by around 1% each year). 

But the good news is that it’s restorable. 

You can help restore your collagen levels with NativePath — a natural supplement that comes without any fillers or “junk.” (Plus, it’s flavorless, so it won’t ruin your breakfast smoothie). 

Click here to uncover seven reasons why you should consider taking collagen today.

New OpenAI feature allows users to customize GPT-4o

In the midst of steadily mounting pressure around enterprise AI adoption (part of which involves slimming patience for returns on AI investments), OpenAI is rolling out a new feature for ChatGPT enterprise users: customization of its most powerful model, GPT-4o. 

What happened: The new feature will allow enterprise customers to train custom GPT-4o bots on company data. 

• This kind of customization (also known as fine-tuning) was not previously available for GPT-4o, though it was available for GPT-4o mini. 

• OpenAI said that, at first, companies will only be able to fine-tune custom models with text-based data. 

“We’ve been extremely focused on lowering the bar, the friction, the amount of work it takes to get started,” Olivier Godement, the head of product for OpenAI’s API, told Bloomberg. 

The context on GPT-4o: OpenAI recently published a system card for GPT-4o, saying that the model only poses a “medium” risk and so is safe enough to release. 

OpenAI added that fine-tuned models remain “entirely under your control, with full ownership of your business data, including all inputs and outputs. This ensures your data is never shared or used to train other models.”

Save Time & Scale Video Creation with AI

Say goodbye to video production bottlenecks. PlayPlay's AI Video Assistant removes the hassle of video editing — helping teams create professional-looking videos in seconds.

Simply describe your video needs in a sentence and watch the Assistant create your video using the perfect template, text, images, and music.

• Cut down time and money, while delivering consistent video comms.

• Boost engagement and conversions with captivating videos.

• Streamline content creation workflows across your entire organization.

Trusted by 3,000+ brands like Dell, CVS Health and L'Oréal.

Create engaging videos now — start your 14-day free trial.

• Military logistics startup Defcon AI closed a $44 million funding round.

• AI customization startup Eppo raised $28 million in Series B funding.

• Authors sue Anthropic for copyright infringement over AI training (Reuters).

• Waymo says it has doubled its weekly paid robotaxi trips to 100,000 since May (CNBC).

• Microsoft is building a data center in a tiny Indian village. Locals allege it’s dumping industrial waste (Rest of World).

• Conde Nast signs deal with OpenAI (Wired).

• Disney gives up on trying to use Disney+ excuse to settle a wrongful death lawsuit (The Verge).

If you want to get in front of an audience of 200,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

Amazon pulls AI-generated books after Deep View inquiry

On Monday, we wrote about a family who was poisoned after using what later turned out to be an AI-generated mushroom foraging guidebook. He didn’t name the online retailer where he purchased the book.

In researching the story, I made my way to Amazon — which, as we mentioned Monday, was inundated with AI-generated mushroom-hunting books a year ago — to see if AI-generated books on the topic were still available for purchase.

They were. 

The details: The search “mushroom guide” returned two books — Mushroom Cultivation for Beginners and the Mushroom Bible — on the first results page that seemed likely AI-generated. Neither page included disclosures of any kind.

Both authors had published a decent quantity of books on a variety of topics in a short period of time, beginning in 2023 or 2024. 

• Neither author’s Amazon page included a bio and neither author — David Odamah or Emily C. Morrison — appeared to exist in any capacity online. 

• Both names, in fact, seemed remarkably similar to existing names in the mushroom space: Adamah Farms is famous for its mushrooms, and Kendall Morrison was profiled by the New York Times in 2010 for his at-home mushroom cultivation. 

I then ran the available samples of each book through a series of half a dozen AI detectors (which are not necessarily consistently accurate). Each detector said the likelihood that Morrison’s Mushroom Bible was AI-generated was 100%; Odamah’s received a likelihood of 80% to 90%.

After The Deep View contacted Amazon for comment, Amazon removed the listings and said it is continuing to investigate other books on this topic. 

“Amazon is constantly evaluating emerging technologies and is committed to providing the best possible shopping, reading and publishing experience for authors and customers,” Amazon spokesperson Tim Gillman told me. “We do not allow content that violates our content guidelines, whether AI-generated or not, and we have proactive and reactive measures to evaluate the content in our store.”

In Wyoming, an AI is running for mayor

In Wyoming’s primary yesterday, voters in Cheyenne, Wyoming faced the usual slate of differing mayoral candidates. Unusually, one of these candidates was a generative AI chatbot called VIC.  

I didn’t have this on my 2024 bingo card, but here we are. 

The details: Earlier this year, Cheyenne librarian and AI enthusiast Victor Miller had the thought that perhaps an AI chatbot wouldn’t make the kind of mistakes that the people who run local governments make. 

So he used ChatGPT-4 to create a custom chatbot, named VIC (Virtual Integrated Citizen), which he put on the ballot. 

• Miller, who has referred to himself as a “meat avatar,” for the bot, said in a recent post on Twitter that, if elected, the chatbot “will be doing one hundred percent of the decision-making completely on its own.”

• Chuck Gray, Wyoming’s secretary of state, said in June that, after investigating state law, an AI bot “can’t run for office.” OpenAI also reportedly shut down Miller’s ChatGPT account, saying that its policies don’t allow ChatGPT to be used for campaigning. But Miller just made a new account and another bot. 

The clerk of Laramie County — which encompasses Cheyenne — said in July that it would be allowed so long as it is Miller’s name, not VIC’s, that appears on the official ballot. 

Miller has said that the chatbot, which he claims has an IQ of 155, will be more thorough and less biased in running Cheyenne when compared to human options. He told Cowboy State Daily that, though he doesn’t expect to win, he expects the likelihood of elected AI officials to increase as people get more comfortable with the tech. 

• As Princeton computer science professor (and author of AI Snake Oil) Arvin Narayanan said, IQ isn’t a valid measure for AI, and intelligence in that sense isn’t really relevant to politicians. 

• “Governing is hard not because of a lack of smarts but because adjudicating between incommensurable values is hard and because allocating scarce resources among stakeholders is hard,” he said. “Politics is frustrating because it is the forum that we have chosen as a society for dealing with our deepest differences. To try to automate that is to miss the point entirely.”

This is a wonderful example of the variety of problems around allowing AI to just run things. One, you have all the issues with reliability, hallucination and bias, which, together, mean that decision-making and AI should be kept very far apart. 

• But let’s say all those problems were solved. What happens when the power goes out? What if the wifi isn’t stable? What if it gets hacked? What if OpenAI decides to shut down the account, temporarily or permanently? 

• What if OpenAI’s servers go down, or the company folds or is bought out? What if OpenAI gains access to all of VIC’s private data regarding the city and people of Cheyenne and sells that data to third parties? 

And what’s the environmental cost of powering an AI mayor? 

Even in the best-case scenario, the system isn’t robust enough to handle vital societal infrastructure. And these truths about energy and power will never go away, AGI or not.

A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

Here’s your view on whether an AI ban would be preferable to no regulation at all:

Around a quarter of you said “yes.” A quarter of you also said that neither regulation nor a ban seems likely. 15% said “no” and the rest said something else.

One of you said that we must start by regulating slowly, focusing on areas of obvious harm (nudify apps), and tightening regulation over time.

Ban it:

• “In an absolutely ideal word, if it can’t be regulated, I would love to see it all go away. Actually, I would love to see it all go away anyway, even if our legislators can figure out effective regulations.”

Something else:

• “Transparency. Ideally, sources used to train/learn. The idea that that means no novel or new ideas, IMHO, is incorrect. AI is looking at existing data that is too complex for us to analyze and often comes up with novel observations and conclusions.”