Killer robots. Deepfakes. Job losses.
The artificial intelligence (AI) revolution can sound dystopian. As governments and corporations pour billions of dollars into AI research projects, the long-term effects of these investments are uncertain.
But with the help of researchers from around the world, we can better understand the benefits and risks that accompany the evolution of AI. Let’s look at four positive ways AI could influence our future.
Decision making in war
You’ve probably heard of AI-enabled weapons systems, commonly referred to as ‘killer robots’. But the role AI could play in war, including whether to go to war in the first place, has been relatively neglected.
Professor Toni Erskine, from the ANU College of Asia and the Pacific, has been granted funding by the Australian Department of Defence to drill into this subject. Erskine and a group of international scholars will look at how AI could influence war room strategising. The technology could allow countries to calculate threat levels, the risk of inaction and mission cost.
While using AI in this way could bring many benefits, such as saving civilian lives, there are some disturbing risks.
“Intelligent machines will ultimately behave differently to humans,” Erskine says. “What if AI could start a nuclear war? This could challenge our understanding of nuclear deterrence and risk unintended escalations in the use of force.”
By considering the possible future effects of these technologies now, Erskine says we can create policy to guide their development and use, promote necessary education and training, and mitigate risks.
AI-stronomy
If you struggle to understand the goings-on of the galaxy, don’t worry – sometimes even scientists need a helping hand with complicated astronomy concepts.
That’s why ANU experts Professor Yuoan-Sen Ting and Research Fellow Ioana Ciucă are working with large language models (LLMs), including ChatGPT, to develop an AI tool to assist researchers.
Ting, Ciucă and their team have been feeding astronomy papers to LLMs and quizzing the algorithms on what they learn.
“Intelligent machines will ultimately behave differently to humans.”
Professor Toni Erskine
They have discovered the models have an exceptional ability to understand intricate concepts in astronomy and can interact with and learn from each other – sometimes even generating ideas of their own.
Ting hopes this way of working with AI could benefit astronomers who are finding it challenging to take their studies to stellar heights.
“By enabling intelligent literature searches and providing an AI ‘colleague’ to converse with, we believe we can make research more accessible to everyone,” he says.
That’s not the only way AI is benefitting astronomy.
A team of ANU researchers, led by PhD student Jeffrey Smith, has been enhancing big telescopes with AI. The revolutionary technology has improved resolution and image quality for stargazers.
“This technology allows astronomers to capture sharper images and obtain more detailed observations, enabling them to study celestial objects with greater detail,” Smith says.
Study buddies
Not every family can afford to hire a private tutor, but most people have access to the internet.
Working with AI, students from primary school to postgraduate levels can break language barriers and overcome inconsistent teaching quality, a lack of textbooks and their own shyness when they don’t know the answer.
ANU computing expert Dr Bernardo Pereira Nunes encourages teachers to lean into the idea of AI tutoring, equating it to the way calculators, spellcheck, the internet, search engines and wikis have enhanced education.
“AI can make personalised learning a reality, providing timely feedback and meeting specific learners’ goals while fostering a sense of independence and democratising education,” Nunes says.
AI taking the wheel
In a long-haul trucking industry facing a shortage of drivers, putting autonomous AI behind the wheel could be a welcome solution.
Totus Vehicle Control Systems, established by ANU graduate Alexander Ollman, is looking for technology-based solutions that would strike a balance between human judgement and AI efficiency.
With fast reaction times and onboard sensors that can better predict potential hazards, an AI driver could be better at avoiding an accident in many cases. However, this would require the human driver, whether remote or on board, to relinquish some control.
Instead of taking decisions out of human hands, a convoy of trucks could be piloted remotely, with one in-person driver to look after refuelling and servicing.
“It could allow truck drivers to work more flexible hours rather than spending days on the road away from family, as well as opening the door for the new drivers the industry so desperately needs,” Ollman says.
The AI co-piloting system could also be used for rescue helicopters.
With medical equipment, staff and other passengers on board, rescue helicopters don’t typically have space for a co-pilot. They can also face extreme weather conditions, from storms to bushfires. Research shows the accident rate during emergency medical helicopter missions is 28 times higher compared to commercial aircraft.
A team of researchers, led by Professor Hanna Kurniawati from the ANU School of Computing, is working on built-in co-pilots that could help make life or death decisions and reduce cognitive load for pilots.
Kuriawati says while the project presents “a big computation challenge”, she is confident her team will deliver.
Given the speed the technology is changing, concerns about AI are understandable. But with the help of these experts, AI could be used to create a better future, not the dystopian nightmare some fear.
