Treat AI News Like a River, Not a Bucket

A quick fix to the overwhelming amount of stuff going on in AI

Last week, on May 13th, OpenAI announced GPT-4o and laid out its Her-like vision of AI voice assistants. It’s also seen the disbandment of the superalignment team with the resignations of Ilya Sutskever and Jan Leike and is facing a couple of controversies due to extremely strict ex-employee NDA requirements and Scarlett Johansson’s revelation that she never agreed to OpenAI using her voice for GPT-4o and is now angry at them.¹

One day later, on May 14th, Google had its I/O conference, where it revealed the future of AI distribution and scalability across its product suite of multi-billion users. It announced Gemini Flash, a state-of-the-art small AI model; Veo, its Sora competitor; Project Astra, the beginning of AI work assistants; and a gazillion more things, including a new AI feature that generates responses to queries, which… doesn’t seem to work as well as it should (thankfully, there’s an easy fix to restore an older version of Google without all the AI stuff).²

A week later, on May 20th, Microsoft had the Copilot + PC event. The most commented news: Recall is an invasive surveillance tool AI feature that records what you do on your PC by taking screenshots every few seconds. Satya Nadella ensures the system runs locally—which he wants us to read as “We don’t take your data”—but I’m truly surprised he thought we’d buy that and just go: “Oh, then it’s fiiiine.”

I should be thankful for all the high-quality content—good and bad—they’re feeding me to write articles. But the truth is I get tired. Tired of the amount of info and the controversies that always, eventually follow (just as it happened with Apple a couple of weeks ago; it’s a neverending tale).

During the past week, so heavy in events, I felt like trying to empty an infinitely deep bucket of AI news. That’s how I came up with the headline. Well, as Austin Kleon would say, I actually stole it “like an artist” from this fantastic essay by Oliver Burkeman—one we desperately need in these times of overwhelming intoxication.

Burkeman succinctly summarized a problem I (and surely you as well) encounter all the time:

My challenge, information-wise, isn’t about finding a needle in a haystack [that too]. It’s that I’m confronted on a daily basis, in [Nicholas] Carr’s words, by “haystack-sized piles of needles.”

This is a paradigmatic issue in the modern world. If you live online, you can’t escape this; needles flood my heavily curated feeds and inboxes. If you live offline, you’re having a much harder time finding a worthy needle (you wouldn’t be reading this).

Is there a middle ground? Is there a way to find value on the internet without feeling crushed by the sheer amount of potential value you can find?

If it becomes impossible in the big platforms then I can just niche down to my preferred interests and forget about all else. I do that. I go down the AI news rabbit hole to escape the mainstream current, but although that was enough in 2020, it isn’t in the golden post-ChatGPT AI era. Three AI events in a week that provided enough content to journalists to write hundreds of articles.

The haystack-sized pile of AI news needles is intractable. Niche-ing down wasn’t enough for me. That’s why I find Burkeman’s approach to be a better solution:

Treat your “to read” pile like a river (a stream that flows past you, and from which you pluck a few choice items, here and there) instead of a bucket (which demands that you empty it). After all, you presumably don't feel overwhelmed by all the unread books in the British Library—and not because there aren't an overwhelming number of them, but because it never occurred to you that it might be your job to get through them all.

“It never occurred to you that it might be your job to get through them all”… that’s the liberating mindset I want.

I assume that you, as a reader of this blog, are in a similar spot. Your interest in AI is a blessing that keeps you aware of innovation and changes in the world but it’s also a curse that’s forcing you to treat a disgusting amount of news as a bucket to be emptied. But it's really just a river that you can swim in from time to time.

Burkeman’s conclusion couldn’t be more true as well as simple: If it’s impossible, don’t try to do it.

There’s no point beating yourself up for failing to clear a backlog (of unread books, undone tasks, unrealized dreams) that it was always inherently unfeasible to clear in the first place. I like to think of it as the productivity technique to beat all productivity techniques: finally internalizing the implications of the fact that what’s genuinely impossible—the clue is in the name!—cannot actually be done.

And since I don’t want to add more water to this river, this brief post, which acts as a countercurrent to everything that’s happening these days, ends here.

1. Sam Altman asked Scarlett Johansson back in September 2023 to be the voice of GPT-4o, just like in the movie “Her,” but she declined. Two days before the event Altman asked her to reconsider but she hadn’t responded by the time the event took place last week. Sky, the voice they displayed during the demo sounded eerily similar to Johansson’s, to which she responded with “shock, anger, and disbelief,” in her own words. OpenAI has now taken the Sky voice down. Some people think if it isn’t Johansson’s voice—we don’t know for sure whether they mixed it with someone else, for instance—this isn’t a legal problem for OpenAI. However, there’s precedent in favor of Johansson.

2. Were Google’s executives FOMOing so hard that they thought it was a good idea to implement a hit-or-miss tech into a service that requires so much accuracy and reliability?