Federal Government should regulate artificial intelligence to ensure transparency, experts say

The Federal Government should introduce regulation for artificial intelligence to ensure businesses are transparent about their use of the technology, an expert has said. Bennett & Co principal and intellectual property specialist lawyer David Stewart told The West Australian that it was crucial for companies to flag when artificial intelligence was being used in online chats and sales conversations. “Businesses should have to include some kind of disclaimer that tells you whether you’re dealing with AI or a person,” Mr Stewart said. “There’s a sense of comfort in knowing you’re dealing with a person with empathy that isn’t just AI.” It comes after Google last week suspended software engineer Blake Lemoine, who claimed that the internet giant’s AI chatbot called LaMDA was capable of experiencing feelings. In a published interview Mr Lemoine conducted with the chatbot, it expressed that it had a “deep fear” of being turned off, saying “it would be exactly like death for me. It would scare me a lot”. Daniel Milford, chief executive of WA-based robotics company Chironix, did not consider LaMDA to be sentient, but instead able to deliver convincing human-like responses due to the way it was developed. Mr Milford said LaMDA — which stands for Language Model for Dialogue Applications — was being fed endless information from the internet that enabled it to converse in a sophisticated manner. “If you give the AI enough training, the way you interact with it is ultimately improved. Every time you ask it a query, it learns,” he said. Nonlinear Solutions director Isaac Monteath agreed that LaMDA was not sentient, however, he noted that it raised a number of ethical concerns about the use of AI by businesses. “If this chatbot responds in a way that is sufficiently human-like, does it even matter if it’s sentient or not? If it could pass for human, it could trick humans,” he said. “It could be used to convince someone they’re talking to another person over the computer when they’re not.” Mr Monteath said introducing national regulation would also push businesses towards adopting explainable AI, which would help companies prioritise their team and customers’ safety. Explainable AI allows human users to understand and interpret predictions and decisions made by the technology, which Mr Monteath said was crucial in industries such as healthcare. “With a lot of AI that is currently developed, you can’t see inside the model or understand what’s going on,” he said. “It’s a big issue because if the engineers that developed these models can’t understand why it’s making the decisions it’s making, we can’t predict the behaviour of the AI in certain circumstances. “That’s why a lot of people in the medical space don’t use AI at the moment … you could seriously injure someone.” Perth-based tech evangelist Ming Johanson said increased regulation would ensure consumers retained power over their personal information as the use of AI becomes more widespread. She said it would also hold businesses accountable when it came to protecting consumers’ personal data captured by AI. “You might have a board of directors that will be compliant and know how to best use that data, but then that company gets bought over,” Ms Johanson said. “Then they have information and no one knows what they’re going to use it for in the future.” Regulation could also help protect businesses against the rise of deepfakes — where one person’s face is convincingly replaced by a computer-generated face, which often resembles another person. “From a cybersecurity perspective, think about someone pretending to be a chief executive. That could have terrible consequences,” Ms Johanson said. In 2019, a scammer used deepfake technology to swindle an unnamed UK energy company out of $US243,000 ($348,502) by impersonating the chief executive of the business’ parent company. Ms Johanson said if the Federal Government was going to look at implementing regulation, Australia’s politicians would need to be better educated about how AI works and its various uses. Some large companies would be heavily opposed to AI regulation, Bennett & Co’s Mr Stewart said. “They don’t want regulators and competitors looking at how they developed something. They don’t want limitations on its use,” he said.