Welcome to The Upgrade

Welcome to my weekly newsletter, which focuses on the intersection of AI, media, and storytelling. A special welcome to my new readers from Apple, Microsoft, The Washington Post, and many other top organizations — you’re in good company!

In today’s issue:

• 🗞️ The Week’s Top AI Stories

• 💡Insights: OpenAI’s Rollercoaster Week

•  🦃 What I’m Grateful for This Year

• 🎙️The Big Interview: Overtone AI co-founders on News + Ad-tech

The Week’s Top AI Stories

Open AI

• OpenAI chaos: A timeline of firings, interim CEOs, re-hirings and other twists — Axios

• Inside the Chaos at OpenAI — The Atlantic

• Before Altman’s Ouster, OpenAI’s Board Was Divided and Feuding — The New York Times

• OpenAI ‘was working on advanced model so powerful it alarmed staff’ — The Guardian

Regulation & Policy

• The OpenAI Mess Is About One Big Thing — The Atlantic

• E.U.’s AI Regulation Could Be Softened After Pushback From Biggest Members — TIME

• Businesses, tech groups warn EU against over-regulating AI foundation models — Reuters

Ethics & Safety

• OpenAI researchers warned board of AI breakthrough ahead of CEO ouster, sources say — Reuters

• Sam Altman’s Second Coming Sparks New Fears of the AI Apocalypse — WIRED

• Meet the 25-year-old AI model earning up to $11,000 a month — Business Insider

• Virtual Influencers Taking Off With Generative AI — Voice of America (Video)

Legal & Copyright

• Parts of Sarah Silverman-Led AI Copyright Case Against Meta Dismissed, But Not Core Argument — Rolling Stone

• OpenAI, Microsoft hit with new author copyright lawsuit over AI training — Reuters

In the Workplace

• 71% Of Employers May Be Left Behind In The Generative AI Race — Forbes

• I tried Microsoft's AI-powered assistant Copilot. It could be a game-changer for admin. — Business Insider

💡Insights: OpenAI’s Rollercoaster Week

The recent events surrounding Sam Altman's ouster and subsequent reinstatement as CEO of OpenAI have been dramatic and revealing of the tensions and challenges within not just OpenAI but the fledgling artificial intelligence industry.

It is now clear that the turmoil began with a letter from several OpenAI staff researchers to the board of directors, warning about a powerful AI discovery, dubbed "Q*,” that they feared could pose a threat to humanity. This letter, along with other grievances, including concerns over the premature commercialization of AI advancements, contributed to the board's decision to dismiss Altman. Despite these warnings, the specifics of the safety concerns mentioned in the employee letter have not been disclosed as of the time of writing. The only details disclosed so far are that the new OpenAI model demonstrated its potential by solving basic math problems, which was a significant development given that generative AI typically excels in tasks like writing and language translation but has long struggled with mathematics.

Altman's firing led to a major upheaval within OpenAI. Over 700 employees threatened to quit in solidarity with Altman and potentially join Microsoft, a major backer and partner of OpenAI. This internal conflict and investor pressure initiated intense negotiations for Altman's return. During this period, Microsoft CEO Satya Nadella played a crucial role, offering Altman and his allies positions at Microsoft and expressing openness to Altman's return to OpenAI under improved governance.

Altman's return was finally agreed upon, and a new board of directors was formed. This new board, led by former Salesforce co-CEO Bret Taylor, includes former U.S. Treasury Secretary Larry Summers and Quora CEO Adam D'Angelo. The reshuffle also indicated a potential shift in OpenAI's approach, suggesting a stronger partnership with Microsoft and a more aggressive commercial strategy​​​.

This episode highlights the fragile nature of the AI industry, especially regarding the governance and ethical considerations of AI development. It also underscores the significant role that large tech companies like Microsoft continue to play in shaping AI's direction. The 88 billion dollar question remains: what exactly spooked the OpenAI employees, ultimately kicking off this entire saga?

What I’m Grateful for This Year🦃

Celebrating Thanksgiving this week, I find myself reflecting on the many blessings and opportunities that have come my way this year, especially the exciting journey of launching The Upgrade.

As I embark on this new path, I’m deeply thankful for the unwavering support and inspiration from mentors, colleagues, and friends whose guidance has helped illuminate the way forward. I’m also grateful to each of you who has joined or plans to join me in this venture – whether by signing up for a course, collaborating via my consulting services, or simply subscribing to this newsletter. Your trust and eagerness to learn together mean the world to me! 🙏

This Thanksgiving, while we express our gratitude for our families and the joy they bring into our lives, I also want to extend my heartfelt thanks to YOU! Let's continue to learn, innovate, and grow together at the intersection of AI and storytelling.

Wishing you all a Happy Thanksgiving filled with love, joy, and endless possibilities! 🦃✨

The Big Interview: Overtone AI Co-founders

This week, I spoke with Overtone AI co-founders Reagan Nunnally and Christopher Brennan about their AI-driven approach to serving publishers and advertisers by turning qualitative text content into actionable metrics.

Peter: Tell me about the core problem you solve at Overtone AI and who benefits from your product.

Christopher: Our mission is to reestablish the connection between publishers and advertisers that has been disintermediated. A study from the ANA, an advertiser's association, found that 21% of ad impressions were from made-for-advertising sites. These sites are simply chucking up something online to capture ad revenue by getting clicks. That's worthless for brands and is part of increasing online content pollution.

Using AI models, we help brands monitor the web and provide insights on the content where their ads are being placed. We look at domains and evaluate website content. By looking at the content itself, we can tell you at a glance that these domains aren't doing the work of journalism or are filled with undesirable content.

We also provide actionable analytics for publishers. We help them see what type of content will do best to serve their audiences. Our content recommender system is optimized for distribution and helps guide publishers’ strategy by answering key questions: Is this the sort of piece that works best in a newsletter? Or is it one that works on a homepage, on social media, or for a specific user group like subscribers or non-subscribers?

Peter: Can you give me an example of how Overtone AI helps serve both publishers and brands?

Christopher: In the advertising business, news gets a bad rap. You don't know what's going to be said in the news. Some brands don't want to attach their messages to what is sometimes unpredictable content about adverse events. When brands can better understand where their messages are going to be in the mindset of the potential customer, that's going to really increase ad performance and benefit publishers, too.

Here’s an example. If you can understand that, let's say a tornado is going to hit, you don't want to be there serving an ad in that news article. That may not be the right time to interrupt with a brand message. But when the community is getting back together, you know, there are cause marketing teams at the agencies that can understand that, you know, an insurance company might want to be there and match dollar for dollar at the end of that article when your consideration and attention is there. We can differentiate those types of articles and provide actionable insights to better serve both brands and publishers.

Peter: Can you explain how Overtone AI metricizes text content via LLMs (large language models)?

Reagan: We use AI and natural language processing models to assess stories based solely on the content itself, not secondary metrics like clicks or shares. Our data is an editorial metric, which takes journalistic know-how and reproduces it at speed and scale. We create an Overtone Score, which is designed to reflect editorial differences between articles and is available immediately upon the existence of the text.

Christopher: Let me provide another example to help explain how it works. We can look at whether a news story is a quick hit about a car crash outside of town or an in-depth investigation about brake pads causing car crashes in the community. And that's our depth score. And that goes from one to five. We then analyze each paragraph to assess whether it's putting forth a fact, presenting an opinion, bringing in someone else's opinion in the form of a quote, and so on.

After analyzing all that at the paragraph level, those scores are aggregated and amalgamated into a content type. We can say, ah, this is more sort of a standard news piece, and that's going to do well on social media where things are a little bit faster, or it may do better in newsletter format. And that data is available to publishers.

Peter: With the genesis of ChatGPT and other LLMs to generate content quickly and at scale, how important is it to provide data on these qualitative analytics?

Reagan: If there’s an ability to generate content exponentially, you're going to need to learn to filter it so you can get to a place where you can manage it. We need to ensure that wherever you have natural language generation, you have processing and analysis tools in place.

Christopher: It’s clear that within publishing and media, we need new common metrics to talk about content. We open doors for creatives to talk to creatives that have never happened before. That’s exciting for us!

💻 New AI course dates out soon! 🗓️

AI Boost for Professional Communicators & Marketers is full and now underway! Be on the lookout for next year’s course dates, which will be released in the coming weeks. Learn how to use the latest and greatest AI tools in your professional workflows!

Don’t be shy—hit reply if you have thoughts or feedback. I’d love to connect with you!

Until next week,

Peter

Psst… Did someone forward this to you? Subscribe here!