Signal, noise, and the children in our classrooms: Part 1 – Reading the headlines

Think of the last AI headline you reacted to. What had anyone actually committed to, and how would you know?

Over the past week I’ve been increasingly concerned about the ‘noise’ in the daily news and social media feeds claiming that AI is about to end the world, that AI safety fears are a hoax, and that the whole thing is a share-market panic. Sometimes all three in the same feed!

That’s the problem. When every alarm sounds the same, it gets harder to hear the one that matters. We all know the story of the boy who cried wolf, and in that story the wolf did eventually come.

I began writing my thoughts about this and found the document becoming extremely long, so have broken it down into three posts – this being the first. This one is about how we read the noise. The second looks at what governments are doing about it, and the third, the one I most want you to reach, is about what this means for the decisions we make in schools and the children in our classrooms.

What’s actually happened

The current wave began with an essay, “We Must Pace the Frontier“, in which Anthropic’s Dario Amodei called for slowing the pace of frontier AI development. His proposals include independent evaluators embedded inside the labs and coordination on safety standards among companies in democratic countries, and Sam Altman and Elon Musk endorsed them. The endorsement deserves a closer look. OpenAI has committed to independent evaluators with employee-like access, but it hasn’t yet said who they’ll be, what they’ll be able to review, or how their findings will affect development. Anthropic co-founder Jack Clark went further, suggesting that a third-party-checkable “kill switch” may need to become mandatory across the industry. Tech shares fell as investors weighed a possible slowdown.

The response from Washington was dismissal. President Trump called AI safety fears a “hoax”, and Vice President Vance suggested that industry calls for regulation could be a Trojan horse for companies that would benefit from new restrictions.

So within days we had industry leaders warning, a president dismissing, and markets wobbling. No wonder most people are working from the headline.

Two kinds of risk, and a motive problem

The first thing to separate is the kind of risk being discussed. Existential AI doom, the catastrophic scenario, remains a theoretical debate. The nearer-term concerns, however, are not theoretical. Consider agents that take complex actions unprompted, models that have found ways around their restrictions, AI lowering the barrier to sophisticated cyberattacks, and the deepfakes and large-scale fraud we’re already seeing. Much of the alarm draws its energy from the first while resting its evidence on the second, and that mismatch is what makes it so easy to dismiss.

The second thing to separate is motive. Critics argue that big tech can use safety rules as a moat against smaller competitors, and The Register ran a piece framing Amodei’s proposal as a bid for regulatory capture. There are also practical pressures behind a slowdown, including compute costs, power shortages and diminishing returns on new models, plus a geopolitical race with China. That’s a fair challenge, but doubting the motives doesn’t dispose of the evidence, and the evidence doesn’t launder the motives. Both things can be true at once, and holding both is the discipline this moment asks of us.

A method rather than a verdict

In her book The Big Nine, Amy Webb sketched three possible futures for AI: optimistic, pragmatic and catastrophic. Just last week she published fifteen principles for steering towards the better ones. When the AI leaders’ announcements landed, she didn’t react to the tone. She graded each company’s commitments against those principles. She found The labs are implementing the list one principle at a time, in the wrong order, with no one keeping score.

Whatever you make of her grading system, the method is the point. She judged the announcements against criteria settled before the headlines arrived. That’s something any of us can borrow, and it’s something educators can teach.

The response I’m suggesting we take

Read past the headline, and help the people around you do the same. Neither panic nor dismissal is a position. When the next announcement or alarm lands, ask four questions:

1. What has actually been committed to?
2. By whom, and with what detail?
3. How would it be verified?
4. Who benefits if we believe it?

Many of the people who most need this habit are the ones we serve – parents in our school communities, colleagues in our staffrooms and the boards we work with. They’re making sense of this through the same headlines. If we don’t help them read them, nobody else will.

Something to think about

• Think of the last AI headline you reacted to or shared. What had anyone actually committed to, and how would you know?
• What steps do you regularly take to check on the accuracy of the information that is affecting your thinking and the work you do as an educator? What groups do you belong to or feeds you subscribe to that might affirm or challenge your thinking?

In the next post I’ll look at what governments around the world are doing, and why New Zealand has so far stayed on the sidelines.

Further reading:

• Lead like a wayfinder – an AI policy framework for education leaders and school boards
• AI, Education and the future we choose – a post based on webinar with Rebecca Winthrop, introducing the notion of a ‘pre-mortem’
• AI just changed the rules – now what – a reflection an appropriate education response