AI have some thoughts.

Logic suggests that I begin by telling you a bit about myself and why you should care about my thoughts. But, as was made abundantly clear to me early in an MBA subject last semester, nobody cares what I think. Apparently, I need to earn that right. And until I do, I need to ground my opinions in credible, and preferably academic, references.

Imagine my surprise at discovering there was a speed at which my eyes could produce a thud, as they rolled into the back of my head. Anywho, let’s just get right into it.

I recently spoke with near 500 students at RMIT (N=477 for my nerds <3) to inform an assignment report on how well the university was navigating AI disruption. With my clipboard sign accounting for what I anticipated was the limit of their attention spans…

I figured 2 per minute at worst, and groups should chop that down, so 3-4 hours tops. 5 hours later I was at 234. Turns out students love giving their thoughts, in detail, if you ask for them. And the majority weren’t what I expected. That expectation being based on recent viral videos of prominent graduation speakers being booed for positively framing AI, while Ronny Chieng gets cheered by Harvard grads for repeatedly saying “Fuck AI”.

So, I went table to table through study areas across various RMIT buildings (8, 9, 10, 12, 13, & 80) to capture participants across a broad range of studies, asking three questions:

• Are you being taught how to use AI within your subjects?

• Do you think you should be taught how to use AI within subjects?

• Do you think being able to use AI well will help you land the job you want?

No preamble, straight into questions. Discussion came while they answered or afterwards. Questions were deliberately sequenced, and that sequence was vindicated by the groans and/or disapproving pressed lips from the anti-AI crowd, upon reaching the final question.

The Results and a big ol’ asterisk

36.9% said they’re being taught how to use AI, 60.5% think they should be, and 55.5% believe it’ll help them find work post-graduation (full results here). These numbers can be framed in many ways, depending on the story you want them to tell. An important caveat before I give you one, is that 36.9% of Q1 is wildly generous.

Conversation revealed that yes rarely meant they were being taught how to use AI across their subjects. For the most part, it was one or two subjects (of four), or an assignment about AI. Not being taught how to use AI within subject context, across all subjects. You could cut that 36.9% in half, shift it to the somewhat column and that might still be generous. It’s a limitation of how the data was gathered, but it’s displayed as gathered.

As for framing, you could, and should, look at that and say that when you include the somewhat group, 84% believe they should be taught, 86% believe it’ll help them get the job they want, and only 37% are being taught how to use it. Alarm bells should be ringing when you read that, and louder still when you account for the aforementioned caveat.

I’ve no doubt a program manager would frame it as 65% are being taught and invoke Officer Barbrady, move along nothing to see here. The through-line is that students already see AI literacy as a path to employability, industry is actively hiring for it (more on that later), and academia is the laggard. And it’s a laggard preoccupied with curtailing AI use rather than teaching students how to use it. I expected my student canvassing to complicate my view. Instead, it’s firmed it.

The results also showed that most students who answered no to Q2 also answered no to Q1, which may reflect a limited understanding of the technology. You might think that’s a long bow to draw, and I probably would have as well, but for a student asking whether my results would be fed into an LLM, while inputting his response into the Claude artifact I’d made to collect responses. His response to all three questions was no.

The lady doth protest too much, methinks

My favourite observation. When I approached groups, a reliable pattern emerged. More often than not, one person would dominate discussion. Launching into a passionate anti-AI soliloquy while the group silently nodded along. The environment, water, AI slop, theft, critical thinking, jobs, etc.

When they were done, I’d ask the rest of the group for their thoughts individually. And one after another, I’d get the sheepish ‘yeah I don’t think it’s that bad’, use it quite a bit, prefer to be taught response. Far more often than the group falling in line, and much to the dismay of the soliloquist.

Over and over, the majority of the silent group landed on the fence, or on the side of yes, teach it. The loudest person wasn’t changing anyone’s mind, they were making disagreement socially inconvenient.

Sociologists have a term for this, the Spiral of Silence. Students also have a phrase for it, “I cannot be fucked with this right now”. Same basic mechanism. People withhold opinions they believe are unpopular, which makes those opinions look even less popular, which silences more people, and the loud minority mistakes that silence for agreement and gets louder. Rinse, repeat, and eventually you get Ronny Chieng cheered for saying fuck AI, while Eric Schmidt gets booed.

Look at my numbers again. 84% want AI taught at some level, and 86% think it matters to their career, but spend an hour on social media and you’d swear those numbers were inverted. The anti-AI position dominates discourse through volume, not numbers. It’s a minority with a megaphone.

There were passionate responses on both sides, but the AI resistance ones tended to come when there was an audience. Not always, but noticeably often. While passionately positive responses were consistently in one-on-one conversation. And I’d almost bet there’s a vocal anti-AI engineering professor/lecturer at RMIT, after hearing the exact same ‘AI bad’ example used by 3-4 engineering students.

Shoutouts to:
The person doing an assignment on AI’s impact on the music industry, from whom I learnt that Spotify removed 75 million spam tracks over 12 months, and that there’s 50-75k AI songs uploaded daily. Oooft.

The person who, after a good minute of group argument, nuked that discussion by asking “wait, why are we arguing like it’s a TikTok clip”. That made me laugh.

And to the person who said, and I quote, “ohhhhh, I think my mum just did that” after I said I’m an EMBA student, how dare you.

Let’s get uncomfortable

EY’s 2026 Global AI Sentiment Survey, based on insights from over 18,000 people across 23 markets, plotted this chart of sentiment against adoption:

India and China top the charts for AI sentiment, adoption, and agentic use. But have a good look at that chart. Spot a trend? I’ll give you a hint. Ranked by GDP per capita, the bottom four countries here are India (USD ~$2,800), Brazil (~$12,300), China (~$14,900) and Mexico (~$15,800).

It’s a pattern that’s hard to not see. The wealthiest, most comfortable, most institutionally protected countries are the most hostile to AI. And the countries where opportunity has historically been rationed are grabbing it with both hands.

The word privilege gets thrown around a hell of a lot when it’s convenient, and markedly less so when it isn’t. If you live in a country where the status quo has treated you well, AI looks like a threat to something worth protecting. If you don’t, it looks like a ladder.

If I inverted that chart, labelled the axes social and economic privilege, and went around campus asking if it looked accurate, we both know what the response would be. But that’s not what I did, I asked whether they thought they should be taught how to use AI. And have a wild guess at how the array of accents over those 477 students aligned. Because that’s what made me remember the EY report.

Those people are here for opportunity, the Hunger Games for AI literate graduates is only just beginning, and that chart suggests they know it better than most. Demand for AI skilled workers in Australia has more than doubled in twelve months. Don’t ignore that opportunity to appease a vocal minority. We’re already standing on the platform outside, and the sixty second countdown is well underway. Which leads me to…

AI wasn’t eating grad jobs, and Victoria may be legislating what was

An opinion piece without a spicy prediction based on synthesis of limited data is no fun, so let’s go out on a limb and come back to this in 12-24 months. If you don’t like that sort of thing, please refer to the end of the Margot Robbie bath scene in The Big Short.

We may have the whole AI destroying entry-level roles thing ass backwards.

Now, I’ll concede the evidence usually cited on this is real. Stanford’s Digital Economy Lab (and a host of others) found employment among early-career workers in AI-exposed occupations dropped sharply after ChatGPT launched, graduate hiring numbers have looked grim for a few years, yada yada yada. You’ve heard all this before.

A recent study from London School of Economics researchers, titled The Broken Ladder, examined 243 million hires and 407 million job postings, over 9 years, across the US, UK, Canada, and Australia. They found that the jobs most exposed to AI also happen to be the jobs that went remote post-pandemic, and previous studies didn’t control for working from home. When this one did, the AI effect collapsed and working from home emerged as the better predictor of the early-career hiring decline. And it’s not remotely (pun intended) close.

If we just take a step back from our desire to work from home for a moment, the mechanism is obvious when stated. Junior employees need supervision, mentorship, and the thousand informal learning moments that happen when experienced people are physically nearby. Remote work makes juniors more challenging (and expensive) to train, while senior hires don’t have that problem. So, companies naturally reduced hiring at the bottom of the ladder while remote work became a cultural third rail, championed by people farther up the ladder.

The New York Fed has since reached a similar conclusion, calculating that remote work can explain roughly 64% of the rise in unemployment among young college graduates. Sixty. Four. Percent. We may need to have a look at Victoria’s proposed work from home Bill again.

But wait, there’s more. After years of decline, a prominent survey (NACE) shows US employers expect to boost new graduate hires by 5.6% for the class of 2026, a survey by ZipRecruiter found that nearly a third of employers plan to hire more entry-level workers this year, and unemployment among 20-24 year olds with bachelor’s degrees and higher dropped from 9.7% in November to 5.8% in May, the lowest May figure since 2023.

But wait, there’s still more! The latest US payrolls data just made a mockery of estimates. 172,000 against a forecast of 80,000. Overperforming by so much that prediction markets went from pricing a 25% chance of Fed rate hike to 52%, within a week. Economies in the grip of mass technological unemployment don’t generally force central banks to contemplate raising rates because hiring is running too hot.

Here in Australia, Indeed’s Hiring Lab 22nd April report showed that after three consecutive years of decline, graduate job postings stabilised in early 2026. March quarter postings were up 6.4% year on year and sat roughly 50% higher than 2019 (pre-pandemic) levels.

One might also argue that suppressed graduate hiring over the past few years had quite a bit to do with inflation, but I appreciate that it’s much easier to just blame the thing you don’t like instead. In the end, it doesn’t actually matter what you attribute it to. What matters is (spicy take time) that we may have the whole thing ass backwards, and AI may be the thing that saves entry-level roles.

I’m old enough to have seen people enter the Communications space as “social natives”, replacing an older workforce that refused to adapt to it. And that older workforce replaced another which refused to adapt to the internet. It wasn’t long ago that people refused to send emails because they preferred handwritten letters. History often rhymes. The only thing that’s different now, is we have a vehicle to amplify collective resistance (social media), and a CEO who stokes the fires whenever his company is doing an equity raise. I see you, Dario Amodei.

Large companies have always, and will always, hire to adapt to technological change because it’s faster and cheaper than retraining a resistant workforce. I think the thing currently shouldering blame for breaking the ladder might be quietly rebuilding it, and I’d bet that it does so increasingly rapidly. AI natives are a scarce resource, and RMIT’s own research shows that only 7% of Australian workers show advanced AI literacy, so…

A microphone-shaped divining rod

Don’t completely ignore this minority with a microphone. Use them as a divining rod for what you should be doing, especially the academic ones. Try to break everything (legally) with AI. Being able to do that may prove more valuable to your employability than much of what you’re being taught in (most) classrooms. You’re there for a shot at a better career, not to tick rubric boxes.

Second, next time you see an anti-AI tirade on LinkedIn, look at where the author sits in the pecking order and compare their position to yours. There’s a fair chance they’re verbally pulling up the ladder behind them, while telling you how terrible it is up where they are.

And lastly, if you believe there’ll be no jobs left because of AI, you’re betting that for the first time in human history, people won’t find new ways to exploit labour for profit. I’ll take the other side of that bet.

References

How good is it being able to hyperlink text instead of pretending like we’re still referencing scrolls with quill and ink.