The era of personalized content with AI is here; music labels take on AI startups; wonky consultants are the new stars of generative AI; DeepMind is building knowledge of how genes work

Chatbots struggle to cite sources; 10% of research may be co-authored by AI; Aramco is betting big on AI; the AI we could've had; OpenAI wants AI to help humans train AI

Jun 28, 2024

∙ Paid

This week, Synthesia introduced a suite of products and features designed to help companies create and distribute AI videos at scale. Most of the press coverage focused on the creation part of the product bundle—and it’s easy to understand why: the new AI avatars are faster to make and perform better than the previous generation tech while products like the AI Video Assistant and AI Screen Recorder will turn anyone into a skilled video creator.

In this newsletter, I want to focus instead on the role that generative AI will play in the distribution and consumption of media, and why we’re entering a new era of the internet where hyper-personalized content could reshape how we access information and entertainment.

The video below offers a summary of this article, generated with my AI avatar in Synthesia, though I encourage you to read the entire article:

Today, the main interface for accessing content online is the personalized social feed. It started in 2006 when Facebook replaced the chronological home page with a new feature called the News Feed. Powered by a ranking system, the News Feed used predictive AI-driven algorithms to curate existing content made up largely of posts shared by your friends, Facebook Pages you liked or Groups you joined (alongside a healthy sprinkle of personalized ads).

How News Feed ranking system works — How AI powers Facebook’s News Feed ranking algorithm

TikTok perfected the concept of a personalized feed by building a recommendation system that delivers content not just from your social graph but also based on your interests. While effective, TikTok’s approach still has one limitation: it can only rearrange pre-existing content. So whereas my For You feed might be different to yours, our feeds are picking from the same pool of standout videos.

That’s because traditional video is, at the end of the day, a one-to-many broadcast medium: you make it with a camera, you edit it with an app, and you upload it to TikTok or YouTube in the hopes of reaching millions of people.

Introduction to the TikTok recommendation system | TikTok

Generative AI is set to introduce a new paradigm of personalization by allowing you to create multiple versions of the same core content, each tailored to individual preferences, demographics, or cultural contexts. For instance, a company could generate one version of a product advertisement for a viewer in the United States and another for someone watching in France. The core message would remain the same, but elements such as language, cultural references, and even visual aesthetics or actors could be automatically adjusted to resonate with each specific audience.

This level of customization extends beyond just geographical differences. Generative AI could potentially create hundreds or even thousands of variations of a single piece of content, each fine-tuned to an individual's interests, viewing history, or even current mood.

The potential for increasing user engagement is enormous. Imagine an educational video that adjusts its tone and depth based on the viewer's expertise level, or that adapts its examples to align with the student's personal interests.

But in order to do that, we first need to rebuild the entire video delivery infrastructure to support these hyper-personalized and interactive experiences—and I’m excited to see Synthesia is doing exactly that. The most obvious place to start with generative AI and hyper-personalization is language: if I want to learn something new or watch an entertaining video, it’s easier to do it in my preferred language.

Click on the button below and you can watch an example video delivered by a new web player built by Synthesia. The video player allows you to easily switch between different versions of the same video through the language button in the bottom right corner. With just one click, you can toggle between the three versions I’ve generated and therefore watch variations of the same video in English, German, Romanian and Japanese. I’ve selected just four languages but Synthesia supports over 130.

Click to watch the video

Look closely and you’ll notice other subtle tweaks I’ve made beyond translating them. Generative AI allows you to fully localize videos right down to the varieties of a given language—in my example, I’ve created American and British English versions. Or you can even toggle between several AI avatars that are targeted for a specific audience. For example, in the Romanian version I’ve replaced the stock AI avatars with my own.

I’d be remiss if I didn’t recognize that this new level of personalization will bring new challenges or amplify existing ones. When misused, hyper-personalization could drive further fragmentation of shared experiences and the amplification of echo chambers. There are also concerns about data privacy and the ethical implications of such granular content targeting. The good news is that policymakers are already thinking about these issues. Under the EU AI Act for example, there are provisions related to transparency for AI generated content, including the disclosures that the content has been artificially generated or manipulated in a clearly visible manner. Meanwhile, Synthesia has implemented robust content moderation policies and systems to ensure that harmful or disagreeable content isn’t being generated with its platform.

For creators of content, the ability to reach audiences with precisely tailored messages could dramatically increase effectiveness. However, it also means rethinking content strategies to accommodate a much more diverse and granular approach to creation.

For consumers of content, the age of one-size-fits-all media is coming to an end, replaced by a world where every piece of content could be as unique as the person consuming it.

And now, here are the week’s news:

❤️Computer loves

Our top news picks for the week - your essential reading from the world of AI

MIT Technology Review: Synthesia’s hyperrealistic deepfakes will soon have full bodies
WSJ: Music Labels Take On AI Startups With New Lawsuits
Fast Company: Perplexity CEO Aravind Srinivas responds to plagiarism and infringement accusations
The Economist: Non-white American parents are embracing AI faster than white ones
FT: The AI we could have had
The Atlantic: Generative AI Can’t Cite Its Sources
The Economist: At least 10% of research may already be co-authored by AI
WSJ: AI Work Assistants Need a Lot of Handholding
Wired: OpenAI Wants AI to Help Humans Train AI
Bloomberg: AI Wreaks Havoc on Global Power Systems
Fortune: The knowledge economy is being eroded by AI. Welcome to the ‘relationship economy’, LinkedIn’s chief economist says
New York Times: The A.I. Boom Has an Unlikely Early Winner: Wonky Consultants
Wired: With AI Tools, Scientists Can Crack the Code of Life
Fortune: Energy giant Saudi Aramco is betting on AI to thrive after the ‘peak oil’ era
WSJ: No One Wants to Sound Clueless About AI. Especially Your Boss.

⚙️Computer does

AI in the wild: how artificial intelligence is used across industry, from the internet, social media, and retail to transportation, healthcare, banking, and more

🧑‍🎓Computer learns

Interesting trends and developments from various AI fields, companies and people

Keep reading with a 7-day free trial

Subscribe to Computerspeak by Alexandru Voica to keep reading this post and get 7 days of free access to the full post archives.