On Valentine’s Day, Dr Jon Whittle’s partner gave him two poems. One was written by her, the other by ChatGPT. “I could tell straight away which one was which,” the artificial intelligence (AI) expert says. “From a technical point of view, the ChatGPT one was better – it was more grammatically correct and the rhymes and meter were better – but I couldn’t see her in it.”
Whittle is the director of CSIRO’s Data61, one of the largest AI research institutions in the world, but he doesn’t overstate the capabilities of the technology. In articles he’s written in The Conversation and on his podcast Everyday AI, he emphasises that “AI is not a panacea for everything”.
For example, Whittle has just finished writing a novel, and he’s now working on a children’s book. “I wondered, ‘How can ChatGPT help me?’” he recalls. “The conclusion I came to was, ‘Not at all!’ For the children’s book, I tried to get ChatGPT to write a chapter, after I gave it bullet points of what to write, and it did it precisely, with proper grammar, but it didn’t have any humanity in it. There was no emotion or imagination.”
Similarly, Whittle asked ChatGPT to write his most recent article for The Conversation, uploaded talking points via voice recordings for the 800-word article. “Initially, I thought it was amazing – in 5 minutes I had a finished article,” says Whittle. “But the more I looked at it, the more I didn’t like it. ChatGPT is really good at producing mediocre copy. I ended up only keeping one sentence, but it gave me the confidence to know that there was an article there to write.”
Making the most of AI
A major platform of Data61, which brings together roughly 800 staff and affiliates, is driving AI’s development and adoption in Australia. Work on robotics, cyber security and other digital technologies are also key areas of research.
Whittle claims that Australians have nothing to fear from AI, either in terms of losing their jobs in the not-too-distant future or going to war with them (a lá The Terminator or The Matrix movies). “There is absolutely no evidence that AI will present an existential threat any time in the near future,” he says. “But, having said that, I wouldn’t hook up today’s AI to the nuclear codes and let it do its thing.”
In order to get the most benefits from AI technology, the industry needs far more diversity, says Whittle. “Diversity in AI is an area where Australia could lead the world,” he says. “Something like 80% of AI practitioners globally are male, and that needs to change. In Data61, we’ve made major inroads in to that, with 55% women hired in the latest recruitment.”
It’s well-known that those who design information technology systems tend to design for themselves, even subconsciously, says Whittle. “So, if 80% of the people developing these systems have a lived experience of being male, then the results will reflect that.”
As an example, Whittle describes an AI system that had been developed to sort CVs for a job placement. The AI was biased against candidates who had taken time away from the workforce to look after children, because it was programmed to mark uninterrupted employment as a desired characteristic.
Australia is actually at the forefront of nations when it comes to responsible or ethical AI use, says Whittle. “Australia was one of the first countries to come up with an ethical framework,” he explains. “Data61, in part, has been leading in how you translate those principles into practice.”
Robots for good
On some metrics, such as numbers of published research papers, Australia is in the top 10 countries for Ai research. “In some ways, we’re more innovative in this space than other countries. We really punch above our weight,” says Whittle. “We’re very strong in computer vision and robotics, and the applied uses for AI, such as AI for the great outdoors, including environmental uses, mining, climate change and natural hazards, such as predicting the path of bushfires in real time. But it’s a familiar story in terms of commercialisation, in that Australia lags a little bit behind, in terms of number of start-ups and other factors.”
Whittle is enthusiastic about finding ways of developing AI for social good. “The challenge is how do we design these technologies so we get the benefits, and not the downsides – so we get the society we want and not one that occurs by accident,” he says. “At the moment, the design of these systems is largely driven by economics. I think that could be turned on its head and we could say, ‘Let’s put social responsibility first.’”
Key to ensuring AI technologies are designed for social good in the future is education, says Whittle. Right now, most computer science students get very little exposure to the study of ethics and social responsibility, maybe doing one, compulsory, disembodied subject during their 3-year degree, he explains. “You come away from that course and you forget everything that you’ve learnt,” says Whittle.
During his former role as dean of the Faculty of Information Technology at Monash University in Melbourne, Whittle looked at embedding ethics into the curriculum of every computing subject, to ensure that it’s part of the fabric of nearly every assignment and discussion. “That’s the only way to make a change,” he says.
Will AI increase productivity? Whittle is not so sure, according to his latest article in The Conversation. “If you drop advanced AI into a dumb organisation, it won’t make it smart. It will just help the organisation do dumb stuff more efficiently,” he writes, pointing out that there have been many cases where “using AI has been costly and time-consuming and failed to generate the desired result”.
The reality is AI could actually reduce productivity in some circumstances, says Whittle, generating more bureaucracy, more noise and more poorly designed systems. “Or it might increase productivity in one place, but at the expense of productivity in another place,” he says. “For example, a manager’s job might become much easier, but everyone else’s becomes harder.”
“If you look at large organisation, there are a lot of mundane tasks that people have to do.
So, the theory with AI is that it will take away those mundane jobs and give people more time for creativity,” says Whittle. But that’s like putting a Band-Aid on a large wound, he argues. “Maybe you shouldn’t have all those mundane jobs in the first place.”
Ultimately, Whittle sees himself as an optimist when it comes to AI technologies. His hope, which is implicit in all of the work that he and his teams do at Data61, is that AI will be a net-positive for society. “I think it comes back to this question of, are we using this technology to create something we want?” he says.
Follow Jon Whittle on LinkedIn
Story by Ken Eastwood
Comments are closed.