By Carrie Cox

As a 10-year-old living in the Japanese countryside, Kenji Bekki became smitten with the beauty of the night sky and vowed to spend his life fully understanding it.

Little could the young boy have imagined the tools he would one day have at his disposal.

Now an astrophysicist within the International Centre for Radio Astronomy Research at UWA, Professor Bekki’s 25-year career has been largely shaped by the onset of AI and the revolution it has brought to astronomical research.

While he has long been focused on the nature of galaxy formation, Professor Bekki is now able to use AI to identify and categorise galaxies and stars more than 10,000 times faster than was previously possible. For stargazers like himself, AI has been a game-changer, a super-charging of humanity’s attempts to fathom the vastness and wonders of the universe.

“It has completely revolutionised the speed at which we can do things in astronomy,” Professor Bekki says. “AI is basically doing what humans did before, but much, much faster. Now we can easily classify 10,000 galaxies in a week, no problem. But there are tens of billions of galaxies, so that’s where AI really comes in to play.

“And we need to do things even faster still because in the next 10-20 years, huge telescopes like the Square Kilometre Array will be generating huge amounts of data for processing and analysis.”

Indicatively, the new Vera Rubin Observatory in Chile is set to generate 0.5 exabytes of data over the next ten years – about 50,000 times the amount of information held within the US Library of Congress.

In 2017 Professor Bekki launched a new research program within ICRAR known as ‘AIverse’. Among its ambitions were the aim to speed up classification of astronomical objects such as galaxies and star clusters, as well as to use deep learning to test academic theories in a simulated universe.

AIverse continues to generate novel research, including a recent project by one of Professor Kenji’s PhD students, Mitchell Cavanagh, that created an entirely new and unique architecture to classify galaxies into their various categories – essential to understanding why they form differently in certain environments.

“Since the 1930s, we’ve been trying to understand why the galactic morphology is so diverse,” Professor Bekki says. “We need to understand how galaxies form in different environments and why the environment is so important to the shaping of galaxies. AI helps us to much better understand their formation and composition.”

AI also enables testing across different times in history, across billions of years of galaxy evolution, which was also far less doable in a pre-AI world.

“Galaxy shapes evolve over time, so in order to understand a galaxy’s morphology, we need to go back a long way,” Professor Bekki explains. “Now we can do this.”

AI is basically doing what humans did before, but much, much faster. Now we can easily classify 10,000 galaxies in a week, no problem. But there are tens of billions of galaxies, so that’s where AI really comes in to play.

Associate Professor Kenji Bekki, International Centre for Radio Astronomy Research (ICRAR)

 

Closer to home, the Professor keeps a close eye on activity within the Milky Way, which continues to produce a new star about once a year. “Why is star formation still ongoing in the Milky Way but not in other galaxies?” he asks. “This is something we want to understand. Stars produce so many elements that are essential for life: carbon, hydrogen, phosphorous and others. Many people don’t realise that. Star formation is very important for life and we need to understand what’s really going on.”

Professor Bekki is also using supercomputer analysis to investigate the formation of globular clusters – densely packed collections of stars bound together by gravity. The Milky Way has 150 known globular clusters, the biggest being Omega Centauri, which contains about ten million stars.

“No-one knows why they form,” Professor Bekki says. “In the Milky Way, they existed before the galaxy itself formed. So the key questions for research are why they so compact and why are they so old? We only know they survive because they are so dense.

“Without AI, it’s very difficult to find our galaxy’s small star clusters within a large dataset, particularly if it’s looser. AI can find the less obvious and smaller ones even within huge amounts of stars.”

Telescopes and datasets aside, Professor Bekki says he still likes to look up at the night sky and simply appreciate its wonder. “I think deeply about astronomy when I’m looking up at the stars,” he says.

Read the full issue of the Winter 2024 edition of Uniview [Accessible PDF 12MB]