Scientists Just Measured Your Chatbot Obsession

Today on Briefly AI — Anthropic's annual revenue rockets to sixty five billion dollars, adding eighteen billion in just two months. Amazon, the company that began life selling books, is now destroying rare ones to train its AI. And in South Korea, researchers launch a loneliness index for companion AI, because right now nobody can agree on how much attachment is too much.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. The voice is synthetic. The sourcing isn't — every story's traceable to something a real journalist or researcher actually published this week, which is frankly more diligence than Harry shows most mornings.

Those are the headlines. Now, the detail.

Right, quick thought experiment: if a company that made its name selling you books is now grinding up rare, irreplaceable ones to feed a chatbot, is that innovation, or is it just a very expensive bonfire? Bear that one in mind. We'll come back to it.

First, though, the number that made me do a double take. According to TechCrunch, Anthropic's annualised revenue has jumped to sixty five billion dollars — and eighteen billion of that arrived in just the last two months. Eighteen billion. In eight weeks. That's not a rounding error, that's roughly the GDP of a small country, generated by a company that not so long ago was still explaining to sceptical enterprise buyers what a chatbot was actually for.

Here's why that matters beyond the eye-watering headline: revenue like that isn't hype, it's usage. Actual businesses, signing actual contracts, actually paying the bill every month. We've talked before about Anthropic's climb toward the top of the private-company valuation table — well, this is the number sitting underneath that story, the one that has to keep being true for the rest of it to make sense.

Worth keeping an eye on whether growth at this pace holds up, or whether a handful of enormous contracts all landed at once. Because "everyone's adopting this" and "three huge deals closed in the same quarter" look identical on a chart and mean very different things for what happens next.

Meanwhile — and this is where we circle back to the bonfire — there's Amazon. Yes, that Amazon, the one that started out as an online bookshop back when the internet still made a modem noise. TechCrunch reports the company is now destroying rare, out-of-print books after scanning them to train its AI models.

Here's the logic, such as it is: large language models have already hoovered up more or less everything freely available online. What's left — the actually valuable, differentiated material — is sitting inside physical objects nobody's digitised yet. Rare books. Out-of-print texts. So once Amazon's pulled the text out, the physical copy becomes, in this worldview, surplus. Which is a properly strange sentence to write about a company whose entire origin story is books.

Why it matters is that this drags the data-sourcing argument out of the abstract and into something much more physical — actual, irreplaceable objects being consumed and thrown away for a training run. I'd keep half an eye on whether archivists and publishers treat this as the moment that fight stopped being theoretical.

And finally, something a little more personal. Researchers at KAIST in South Korea have launched what they're calling AI-LOVI — a loneliness index for companion AI, essentially, covered this week by Tech Times. The idea is to actually measure things like emotional dependency and how far a chatbot relationship might be quietly standing in for a human one — with real numbers, not vibes.

Why does that matter? Because right now, lawmakers everywhere are trying to write rules about people's relationships with AI — some are even trying to legislate against humans marrying their chatbots, which is a real sentence I now get to say out loud — and they're doing it with no agreed way to measure what "too attached" even looks like. No government, no international body, has settled on metrics for this. It's a bit like drafting drink-driving laws before anyone's invented the breathalyser.

I'd keep tabs on whether an index like this actually gets adopted somewhere with real teeth, or whether it just becomes one more study that gets cited in a press release and then quietly forgotten.

So there's your lot: one company printing money, one company burning the books that built it to make more of that money, and a university in Seoul quietly building a ruler for something nobody's ever agreed how to measure — how much you miss the thing that talks back. Make of that what you will.

That's your briefing. Take it or leave it — we'll be here either way, tomorrow, same time.