16 May 2023

Is the public service really ready for AI?

| Chris Johnson
Join the conversation

AI is presenting the APS with a number of dilemmas. Can it handle them? Photo: File.

How will artificial intelligence be regulated in the Australian Public Service? Good question. And it seems lots of people are asking it inside the service and receiving very little back by way of an answer.

AI is here and taking over the world – one departmental pod at a time.

There is a lot of good to embrace with AI. It will make so much about how we all work easier, but there is also a great deal to learn about this brave new world rushing right at us. It is shaping up to be a game changer on the scale of the industrial revolution.

Putting aside that whole other debate about whether it will soon do most people out of a job (not to mention the one about whether it will control humanity one day), there are many other questions longing for answers.

Like, is ChatGPT my new boss?

There is a lot of encouragement in the APS right now to get on board the AI train.

Nothing wrong with that.

READ ALSO Too many public servants and too many gambling ads, says Dutton in budget reply

Technological advances have traditionally been instrumental to the progress of government work and service delivery.

Anything that can streamline processes and hasten outcomes has to be a good thing, right?

Well, maybe.

With all new technologies come teething problems, hiccups and the potential for disaster.

That’s OK (kind of) if it’s the trial of a new software suite or the upgrading of old systems.

When the genuine aim is to reduce the instance of error during a period of trial-and-error, hopefully, a positive outcome is the end result.

But when the technology is regarded as the absolute truth and left to rule, catastrophes await – hello, Robodebt.

With AI, however, a significant level of control is lost to a bot.

It presents the public service with an ethical dilemma.

How will the embedding of AI affect the development and delivery of policy?

After all, policy decides much of how we live, so it’s important to get it right.

READ ALSO Beginning of the end? Budget delivers a whole new meaning to APS consulting

AI can and will conduct research, present scenarios, flag issues and even write whole theses.

That might sound like a perfect day for a policy wonk stuck behind a computer with relentless deadlines crashing in on them, but it’s also the recipe for a perfect storm.

Will the service soon be relying on AI to make decisions?

It’s easy to say “of course not” but … once again … Robodebt – a perfect example of how easy it can be to let technology make decisions, take out the human element and insist everyone live by those decisions.

Is the APS even able to advise the government about AI?

Can it procure consultants to do that? Will the government allow the APS to procure consultants? Do they even exist?

Will the onset of the age of AI be much like the digital age, where it left so many behind the eight ball for so long?

It’s a bit of a cliche to suggest AI could be used for good or evil, but when it comes to being used by governments, those opposing realities must always be front of mind.

The alarms ring even louder now that even some creators and pioneers of AI are publicly warning that embracing it wholeheartedly is fraught with risk.

Harness it, yes. Capitalise on its benefits, certainly.

But when it comes to the vital work of government, which permeates everyone’s life, working out how to regulate its use must be the first priority.

Can someone send someone to a course on that, please?

Join the conversation

All Comments
  • All Comments
  • Website Comments

Ugly reality is that it seems the powers that be are expecting significant drop in global population, hence the need to get AI to do the jobs of those disappearing people.

What if Deagel.com is wrong? It predicted a drop of more than 50% population for Oz. How might that happen? Plandemic No. 2 perhaps?

So in a trial in New York, AI did better than senior lawyers with contract analysis. I would expect 25-50% of lawyers to be out of work in the next 10 years.
IBM said they avoided hiring 8000 people because they could automate many jobs with AI.

HiddenDragon7:39 pm 16 May 23

Sooner or later someone somewhere in the world will come up with what will amount to a more elaborate version of Asimov’s Laws of Robotics which the APS will be able to cut and paste and adopt with a bit of panel beating for its particular circumstances.

In the meantime, budget estimates of major indicators such as inflation and economic growth could be produced with AI – it couldn’t do a worse job than the Treasury.

TruthinMedia2:39 pm 16 May 23

None of what anyone anywhere is doing is artificial intelligence. Back when I studied it in the late 80s it was called expert systems, a much better title. There is no sentience and no chance of this becoming sentient. It is just another computer program but with a twist called fuzzy logic. This has two major implications. The first is that it you can pre-load it with lots of synonyms e.g. if you say how do I get back to my pad / condo / shack etc it can infer you mean home which is the database entry holding your current address and then provide directions. It has inferred your intent. It can also use probability to suggest and answer to you. Over thousands of transactions of people saying pad/shack etc and the program saying I’m not sure what that is and us saying our house/home it will ‘learn’ that pad/shack etc mean home with a high probability. The problem is that this machine learning can be supervised (where we correct it) or unsupervised where the program webscrapes information and no one tells it is wrong and this is how problems such as racial or gender bias arise. Simple’s yes!

Balance needed1:22 pm 16 May 23

I asked ChatGP to provide me with 10 recent peer-reviewed journal articles on the subject of “female emotional aggression towards other females”.

Here are the first 2.

1. Hedges, S. J., Hirsch, J. K., & Marmion, S. M. (2016). Informational support seeking and friendship quality among adult females: The moderating role of relational aggression. Journal of Social and Personal Relationships, 33(3), 347-367.
2. Tung, Y. H., & Ruppert, T. (2016). Victim–perpetrator dynamics in female college students: Relationships among relational aggression, perceived control, and psychological distress. Journal of Interpersonal Violence, 31(3), 496-519.

The references look impressive. The problem is, chatgp simply made them up. Made up all 10 in fact. Those articles do not exist in any form. Upon repeated queries, chatgpt finally admitted that none of the articles existed. This certainly raised my concerns about AI.

Capital Retro4:57 pm 16 May 23

Having read that it’s easy to see how climate change alarmism got legs.

I mean 10,000 climate scientists agreeing with each other that we are doomed yet there is not one climate scientist listed in the Yellow Pages.

Capital Retro10:42 am 16 May 23

I’ve heard a lot about AI lately – it was often discussed on ABC TV Landline.

I was thinking why is everyone suddenly become obsessed about Artificial Insemination but I can now see that it has a place in the APS so please carry on.

Daily Digest

Want the best Canberra news delivered daily? Every day we package the most popular Riotact stories and send them straight to your inbox. Sign-up now for trusted local news that will never be behind a paywall.

By submitting your email address you are agreeing to Region Group's terms and conditions and privacy policy.