No one asked me. But here's what Australia needs to survive AI

Apologies we’re a couple days late. I bit off more than I could chew trying to cover something more substantial than last week.

Let me know what you think, and it’d be ripper if you could subscribe at tokeneffort.io

It’s safe to say the 2010’s were a little more laid-back than the 20’s have been so far. Even the Orange Man’s first term in the White House didn’t come close to the fireworks we’ve seen this time around…

It was the same for AI, especially in Australia. Our AI projects included things like the CSIRO working on bushfire predictions; and the bank's “AI modelling” things like credit applications. The Government even published a nice little 8 point voluntary AI Ethics Framework in 2019

Then we crossed into the next decade.

In a real “hold my beer” moment trying to upstage COVID, ChatGPT arrived in November 2022 and the AI revolution really got cracking. In the 2.5 years since, we’ve heard lots about AI in the US and China, but how is it impacting Australia? Are we ready? Is this a good news story or bad?

Definitively answering that question is above my pay grade, but below are four things Australia needs more of, if we want to ride out of the AI storm.

More workers

Globally, the demand for AI expertise is already going f*cking bananas. Mark Zuckerberg is now throwing $100Mil USD sized individual signing bonuses at top tier AI talent to get them to join Meta.

In the words of my least favourite former footy commentator. “Boy oh boy. Wowee.”

PwC came out a couple of weeks ago and said “6.9 % of all job ads now ask for explicit AI capability” (although they’re unlikely to come with a 9 figure signing bonus) that’s doubled since 2023, and most importantly we don’t have the people to meet the demand. 

The unfortunate flow on effect is that according to The Tech Council of Australia our economy could lose out on up to $26 Billion each year to other economies that can meet the demand we can’t. 

That’s bad.

More compute

If anyone’s wondered how the Artificial Intelligence sausage is made, the Intelligence part is powered by “compute”. Compute is the crunching of numbers for AI, the physical data centres that do it, and the power this consumes. This infrastructure is incredibly in demand and Australia relies mostly on stuff owned by off-shore companies. 

According to this study, Australia’s data centre power requirements are going to more-than-double by 2030 and that power alone will cost roughly an extra $26 billion.  

Just this week Albo signed a deal with Amazon for $20Bil worth of data centre infrastructure, so the money is coming, but the problem is that most of it is foreign.

While this admittedly pretty boring we should still care. Because even though it’s based in Aus, the access to the compute isn’t under our control. 

When Australian businesses try to tap into this infrastructure in our own backyard, because demand is so high, we can get left in the dark. This has already started happening, and in future it could kill our economy.

That’s bad.

More laws

I wonder what the road toll would be if speed limits operated on an “if you feel like it” basis…

Despite everything, Australia still doesn’t have AI specific laws. There are some recommendations and guidelines, but it’s all voluntary. (*The exception being some light touch obligations for Federal or NSW Government departments)

Voluntary isn’t good enough, for two reasons. 

The first is obvious, in that if AI is unregulated it can move into areas it shouldn’t, and when it does there’s not much that can be done. Case in point, Clearview AI stole peoples images off social media, and tried to sell it to the police. . 

The second is that if governments and the public sector don’t need to adopt AI at all - then a huge amount of them won't. People resisting new technology at work isn’t a new thing, in businesses, they’re the people who lose jobs. 

But the impact of the public sector burying heads in sand is that the whole country moves too slow, misses opportunities, wastes money, and moves backwards in the global economy.

That’s bad.

More clarity 

While we needed AI laws yesterday, I’ll admit they’re not easy to write. 

Early chat on Australia’s regulations are pointing towards adopting a “risk-tier” approach similar to parts of the EU’s model. (Groups AI into 4 risk buckets. Prohibited, high-risk, limited-risk, minimal-risk. The more risk, the stricter the laws.)

While there’s lots of good about what Europe is doing, the temptation for lawmakers is to play it safe when classifying technology that moves this fast. This helps stop AI being used for bad stuff, but could also mean Australia doesn’t get the good stuff and/or discourages companies from building anything here at all.

Safety is important, but so is access. If we don’t get it right, we become the bubble-child economy of the AI world. Never gets sick, but is useless at everything.

That’s bad.

So far, I’ve just been whingeing. So here’s what we should be doing instead.

Firstly, let’s pull our finger out and create an AI specific department of government.

Tim Ayers is the freshly minted Minister for Industry and Innovation and Minister for Science, He’s going to be a busy man. Included within that hefty portfolio are the bulk of AI specific responsibilities, but he’s got a lot of other stuff on his plate.

We need a Minister for Artificial Intelligence and with a portfolio driving AI specific progress in Australia, ensuring that they include the following:

Create laws based on a blend of the EU, US and UK - Safe AND progressive. 

• National Rulebook. Don’t let the states/territories build a birds nest of AI laws. It’s too complicated and costly for businesses to figure out, and it slows progress.

• Risk Based Model. Makes it easier for laws to keep up with change.

• Innovation First. Don’t get heavy handed with the risk classification because it’s easier, things like LLM’s shouldn’t be automatically classed as high risk.

Cash in on Green Compute - Maybe trillions of dollars of future upside.

• Build sovereign data centres. Lots of them. Either government owned or more likely, joint ventures. That’s massive potential for a data centre construction boom creating high skill jobs, on projects that will make us money.

• Green Power.  Power them with renewables. Admittedly easier said than done, but with advances in tech and the requirements specific to a data centre it’s looking more and more possible.

Find and create AI talent - AI will take jobs, but it can give them back too.

• Create an AI skills plan. Then pay people to learn. Supercharge something like Microsoft’s Data Centre Academy to create a national curriculum. Or provide tax deductions for companies that fund training programs.

• Bring talent with special AI skills shortage visas. Pretty self explanatory really. Find great talent and bring them in to help build out industry, particularly outside of Sydney & Melbourne.

Look I don’t reckon any of that is groundbreaking, we just need to move faster and double down on AI. If we don’t we’ll cop all the bad of AI and none of the good.

The upside though, is that Microsoft and the Tech Council estimate responsible generative-AI adoption could add $115 billion to our GDP EVERY YEAR by 2030 

If we capture even two-thirds of that number, the value of AI would overtake coal as an Australian export. 

But if we sit on our hands, the cost will be very clear. Talent and money leave the country, as Australia struggles to keep up. 

If you’ve managed to read this far, I’m pretty impressed at your staying power. Despite the fact it’s bloody important, AI legislation isn’t exactly blockbuster entertainment.

As mentioned, I bit off more than I could chew with the topic this week, so let me know if there’s something in Artificial Intelligence I can have a crack at, that isn’t going to be so heavy duty.

Subscribe at tokeneffort.io if you haven’t already.