There is a saying that everything old is new again. Avron Barr and Edward Feigenbaum edited The Handbook of Artificial Intelligence in 1981; Newell’s Intellectual Issues in the history of Artificial Intelligence appeared in 1983.
It seems silly now, but when I was a young business planner in the early 1990s, I was full of expectation that ‘expert systems’ and ‘artificial intelligence’ were going to make a big impact on strategic planning by being able to rapidly consider the consequences of multiple potential pathways.
I wasn’t the only person to be disappointed, as Henry Mintzberg noted in 1994:
The failure of strategic planning is the failure of systems to do better than, or even nearly as well as, human beings. Formal systems, mechanical or otherwise, have offered no improved means of dealing with the information overload of human brains; indeed, they have often made matters worse. All the promises about artificial intelligence, expert systems, and the like improving if not replacing human intuition never materialized at the strategy level.
On this website two years ago I noted that suddenly AI was everywhere, the year before I had noted the failure of a decade of Digital Economy strategies and pondered whether the AI strategies proposed for that election would be any more successful. More recently we have been assailed by more frequent columns on AI, including calls reported by this site’s Editorial Director for a national AI strategy. The specifics in this latest call include the noble (funding for core AI research) to the bizarre (establishing a domestic semi-conductor industry).
In his column, James Riley made the observation that “Decisions that get made today about industry development strategies and productivity will have a big impact in the decades to come.” This statement is true of many decisions made by government; decisions that range from tax policy, general research policy, immigration policy and foreign investment policy.
As one example consider the decision by Amazon to invest $20 billion over five years in data centre infrastructure and renewable energy. Sandy Plunkett described this as light on detail, noting, “No clear technology transfer, sovereign capability development, or industrial spillover built into the deal; no requirement that Australian firms, universities, or workers be structurally embedded in Amazon’s AI supply chain.”
But James is right to link the question of AI to the other current hot topic, productivity. AI in of itself will not contribute to productivity, we are highly unlikely to build an export industry on the technology alone. An AI strategy will contribute to productivity by the extent to which it can grow real output using the same level of labour input.
That clearly points the application of AI to the services sector where so much of our labour force is employed. We need AI in health care, aged care and education.
The other place it points is value adding to existing sectors. I was struck by a tweet (what do you call a post on X?) this week of a bag of potatoes with the label “Grown in Australia, Sorted in China”. They turned out to be macadamias not potatoes, but the point remains.
The logic of this story would be that primary produce is often visually inspected to grade and sort it. That is more cheaply done in a low wage country, and since a lot of the produce in this case would be going to other countries not just Australia, sending it in bulk for sorting to China makes economic sense.
To think more about this question, I want to use two stories from this week’s edition of Landline. The first was a story about a small business using AI to grade lentils. Lentil grading is done by taking a small sample from a load and testing it for seed damage, insect damage and foreign seeds. Manual inspection by trained examiners can take up to twenty minutes to grade a sample which becomes a major bottleneck in processing.
The second story was about a farmer who makes maximum use of a small plot, raising ducks, pigs and cattle. The relevant part of his story is that to make his farm profitable he has integrated downstream to gain more of the margin before the product s retailed or consumed. He has to outsource beef processing, but butchers and packages beef himself. He carts his ducks to his high-end customers.
The application of AI and robotics can transform Australian food production, resulting in more value for Australian goods without the use of additional labour. The six Australian robotics start-ups fighting for an opportunity to go to Boston. However, far simpler, perhaps even boring, projects have the ability to make substantial additions to productivity.
In 1987, in the middle of the last wave of the ‘technology changes everything’ boom, the economist Robert Solow quipped “You can see the computer age everywhere but in the productivity statistics.”
A domestic large language model might well help make the bots that drive the artificial agents in online customer service, but they are no substitute for better designing products and services so that consumers don’t need to visit the service queues at all. Let’s not build an AI strategy for the sake of an AI strategy, let’s ground it in meaningful productivity improvement.
Do you know more? Contact James Riley via Email.