Surprise! You've Been Hired by AI (PWW #14)

Congratulations on your new job! No, you didn’t apply for it, and yes, it's strange to find out about it in my newsletter. But it’s true – you’ve been working another job, unbeknownst to you. What’s your new job, you ask? To create as much content as possible! Text, audio, video, images – it doesn’t matter. As long as you keep posting, your new employer, AI, is happy.

This article was originally published in my newsletter Product Whispers Weekly.

The unseen workforce

Basically, everyone who uses the Internet has been hired by big tech on a zero-hour, zero-wage contract. This is part of the AI arms race that's speeding up at an alarming pace. I call it an arms race because there are many competitors trying to create the top AI model. But this arms race is also about something else – data.

The Large Language Models being built by OpenAI, Google, Anthropic and many others rely on sucking in as much data as possible. They scrape the Internet for all text they can get their hands on. OpenAI was caught scraping YouTube videos without permission. And last week it became clear that Slack is using your work DM's to train their AI models.

And that's just the tip of the iceberg. Deals are being made left and right between AI companies and content platforms like Reddit, Stack Overflow, the New York Times, and many more.

But at the end of the day this all means one thing: Big Tech needs you to create more content.

Content is the new gold

We've all known that when you are using a service that is free, you are the product. What's changing now is that the actual content you create is becoming valuable.

Companies like Google and Facebook have always used your content to create a profile of you, which they could sell ads against. The actual content of your posts didn't matter: they only cared about what it said about you.

Now that's changing. Content is worth money. Major publishers and platforms are signing deals with AI companies left and right. News Corp signed a deal with OpenAI to provide their content. Reddit signed a similar deal with Google. And there are plenty more examples.

It's one thing for news outlets and publishers to strike deals like this, but the Reddit deal is different. The Reddit deal is about selling user-generated content on the platform. The posts and comments you and I create are being sold for millions of dollars, yet we receive nothing.

A new age of the Internet

We are at the start of a new paradigm for the Internet. Two massive problems are starting to appear: the erosion of privacy and the flood of AI generated content.

Privacy has been eroding at an alarming rate since Web 2.0 and the rise of big-tech. But we are now at a stage where you should assume everything you do within the big-tech ecosystem is used to train AI. Instagram posts, Gmail messages, YouTube videos – they are highly likely to be fed into an AI training data set. You have zero control over this, and it's unclear what the implications are in the long-term.

At the same time we are all being flooded by AI generated content. Most of it still fairly easy to spot, but that's getting harder every week. How do you feel about a future where everything you read is generated by AI? Seems like a bleak future, doesn't it?

In other AI news

• OpenAI launched their GPT-4o model. It is now multimodal, meaning ChatGPT can now work with text, audio and images. The new voice chat mode is incredibly impressive! Definitely worth checking out the announcement video.
• In the Google IO keynote the term 'AI' was used 121 times. Too bad they failed to announce anything impressive. The keynote felt excruciatingly long (110 minutes) and lacked any overarching story line. It felt like every department in Google was given some speaking time to talk about what they are trying to do with AI.
• Anthropic researchers are figuring out how an LLM actually works. I know this sounds weird. A maker of large language models is trying to figure out how their models work on the inside. But this is actually a problem: no one knows exactly how an LLM gets to a specific result... Sounds scary right?

Podcast tip: AI For Humans

The AI For Humans podcast is great for everyone who wants to stay somewhat up to date on what's happening the world of AI. Kevin Pereira & Gavin Purcell update you on the latest news. But what I find most fun is their own experiments. Being tinkerers, they try all kinds of fun things with AI. For example their AI Arnold Schwarzenegger was hilarious ??