After Comyn’s AI trip, the CBA boss can compute impact on banking

Profound change on horizon

Comyn is still absorbing the implications of this. He clearly does not have all the answers. But as he wrestles with the incredible advance in computational power, he realises the impact will be widespread.

There will be geopolitical implications. It will reshape the global technology industry and business landscape in profound ways yet to play out.

“Now, I have a window into all the things I don’t understand, so will approach this with due humility,” he tells the Summit in Sydney on Tuesday.

What is at stake is “the production of intelligence”. Jobs will inevitably change. Generative AI will provide the “ability to plan, to reason and to undertake longer-running tasks and multi-sequence steps”, he says.

In this world, AI will not just assist workers, but will become an agent – able to do a lot of work on its own.

CBA, like other banks, has been touting its AI capabilities for several years. But Comyn is now more interested in foundational issues: how models will become more powerful, how governments should be thinking about industry policy to ensure Australia can attract proper skills and global investment.

This involves complex policy questions, from providing renewable energy supplies for data centres, to data policy and regulation to protect users from a multitude of risks, without stifling innovation.

He says computer power and investment in AI is roughly scaling at around 10 times every year.

It cost about $100 million to train the early generative AI models; the latest versions cost in the billions of dollars. (Meta will be use 350,000 graphics processing units by the end of this year to power its new LAMA 3 model, costing $US10.5 billion ($15.8 billion).

Capital expenditure is surging. Microsoft, Google and Meta will each spend around $50 billion in the next year on their AI capabilities.

As Tesla chief executive Elon Musk told US chief executives last year: “GPUs are at this point considerably harder to get than drugs”.

AI in geopolitics

Comparing AI with the industrial revolution is reasonable, Comyn says.

What he tells the Summit shows how CBA is thinking about experimenting with several large language models offered by the tech companies. He doesn’t want to put all the bank’s eggs in one basket, but nor does he want to spread them too broadly.

“We’re not going to be training our own large language models, but there are data sets that will be running over them,” he says.

“I think every enterprise is very much thinking about data security and protection, and that’s an important consideration for us.”

Ultimately, the design of engineering teams and software should ensure “you can move out different models and replace them in and out,” he says.

Improving customer service and the efficiency of its engineering teams are the two initial use cases at CBA.

He also wants Australia to consider AI as a geopolitical issue. Globally, big tech is thinking about migrating data centres to where energy is cheapest and are looking to vertically integrate power generation.

Microsoft and Amazon are exploring using nuclear reactors to power some of their data centres. The Middle East is keen to host big tech AI data centres with renewable energy.

“I certainly think it’s something we should be thinking deeply about. I do think for every country, there are a number of fundamental questions associated with this,” Comyn says.

“The production of intelligence requires a lot of energy.”

Industry Minister Ed Husic tells the Summit the government will provide $17 million to assist small and medium-sized businesses adopt artificial intelligence, under the Adopt AI Program. The federal budget involved $40 million to support the adoption of AI in “a safe and responsible manner”.

Impact on jobs

But Comyn’s comments suggest Australia could be more ambitious.

In the US the CHIPS Act of 2022 has triggered a wave of capital expenditure, up to $US260 billion ($390 billion) to 2030 by some estimates. Congress has allocated $US54 billion to rebuild domestic chip making.

The UK is spending £1 billion ($1.9 billion) on developing AI capability, including acquiring GPUs to create an ability to compute for issues of national interest, such as financial services, education or healthcare.

“You could imagine some use cases in and around defence and sovereignty where that might be important. You see a number of different countries who are very actively developing and promoting national champions,” Comyn says, pointing to Canada, France and Germany.

He also flags a looming and significant impact on jobs, although he adds that it is too early to be specific on numbers.

Last year, researchers at OpenAI found 80 per cent of the US workforce could have at least 10 per cent of work tasks automated by AI; for 19 per cent of workers, it would be at least 50 per cent.

“Whether it’s the agricultural or industrial revolutions, there’s always concerns, understandably, about what will the impact be on jobs,” Comyn says.

“And we’ve seen, obviously, over that time big changes in the way economies might be arranged and the various impacts. So this is going to continue to be a significant issue and concern.

“I don’t know that it will be large in the near term. I think over the medium and longer term, there will definitely be a reshaping, and we are thinking about that in terms of our workforce, but also the skills that we really need for the future that we don’t have at the moment.”

No wonder AI has become one of the most important areas for CEOs to understand. The potential of the technology is unclear. But Comyn’s message is bosses need to attempt to wrap their heads around it, or risk missing out on the benefits it is set to deliver.