Deepfakes and a robot's warning: Why AI leaves more questions than answers
Rachel Younger reports on the advancement of AI and why it leaves more questions than answers
For me, it was personal. The moment I really understood AI’s potential and its pitfalls came when I first saw myself on screen, speaking words I hadn’t spoken, doing things I hadn’t done.
Worse still, my boss, who was passing my desk, couldn’t tell the difference.
“Looks good,” he said, as he glanced down at my laptop.
My deepfake didn't convince him for long. But for an uneasy moment or two, I wondered if, in attempting to explain AI (artificial intelligence), I had done myself out of a job.
It wasn’t difficult. For 20 minutes I was recorded reading a book before having a quick photo taken.
It took just half an hour for this deepfake to be made - this is how it was done
Then, using freely available tools online, we cloned my voice and manipulated my image, before asking the new version of me a question or two and filming the answers.
Journalists don’t shock easily, but none of us could believe the result.
Not because my AI replacement was perfect – she wasn’t – but if we could make a vaguely credible deepfake in just over an hour, what could the technology be used for in the wrong hands?
As a reporter it is my job, whatever the story, to try to establish the truth. But the more I learned about AI, the more I realised how much harder telling the difference between fact and fiction is going to become for all of us.
Just look online. Deepfake technology is already causing chaos and confusion.
Social media is awash with fake images of Donald Trump’s arrest, showing him being manhandled by police and then running away from them.
That didn’t happen, nor did President Volodymyr Zelenskyy of Ukraine ever tell his troops to walk away from the war against Russia, but you can find videos of him online appearing to do just that.
One of the UK’s leading experts on the development of AI is Professor Michael Wooldridge, a director at the Alan Turing Institute. He believes AI has the potential to disrupt democracy.
“I’m really worried that these programmes are going to be used to industrialise the production of misinformation and fake news,” he told us.
“There are clearly people that would want to use the technology for that and the technology clearly can be used for that, so I think it’s a really imminent concern.”
He isn’t the only expert to worry that we might be losing control. The past year has seen game-changing developments, like the arrival of Chat GPT or Google's Bard.
They are computer programmes, powered by artificial intelligence that can both generate human-like text and interact with us.
Trained on vast amounts of information from the internet, they are so advanced their impact has been compared to the discovery of electricity.
The machines learn using artificial neural networks modelled on our brains, but develop their learning so quickly there is a real fear they could one day start taking decisions on their own.
Earlier this year, over a thousand tech leaders, including Twitter’s Elon Musk, signed a letter calling for a pause on the development of sophisticated AI until we work out what the consequences might be.
But for everyone worrying that artificial intelligence might be moving too far, too fast, we found others already making the most of what it can offer.
We met Aidan Cramer, an entrepreneur, who uses Chat GPT like an assistant. It takes care of all the jobs that slow down his productivity.
“I use it for coming up with ideas, finding domain and company names, writing legal contracts or job posts, the list is endless,” he said.
Aidan Cramer says he uses AI to help his business
The problem is, what Aidan does at the touch of a button is what some people do for a living.
Admin workers, lawyers, accountants and call handlers could find some or all of their jobs being automated.
Emad Mostaque, the founder and CEO of Stability AI, thinks the economic impact could be bigger than the pandemic.
“The only question, is in which direction?” he said.
“I think we need to move as a society to make sure that we mitigate against the potential losses and create new roles… sector by sector, we have to see it coming.”
Over the weeks we spent exploring AI, we saw how it promises to change many areas of our lives.
From NHS hospitals, where its ability to interrogate data and images makes it a great tool for doctors and could eventually generate new drugs, to community centres, where virtual companions might be used to engage in conversation with the lonely.
I ended my journey by meeting a robot called Ai-Da, who creates spectacular paintings using cameras in her eyes, AI algorithms and robotic arms. Created by scientists at Oxford University, she was designed both to illustrate AI’s power and to serve as a warning.
Ai-Da was taller than me, with a sharp bob, an extensive vocabulary and an unnerving way of maintaining eye contact.
Interviewing her was one of the most surreal moments of my career. But after all the experts we’d spoken to, it felt rather apt that she gave the best answer to the most important question of all.
“Will AI ever become a threat to us? Are you dangerous?” I asked.
'What do you want from your futures?' - the critical question Ai-Da asks
“Me, the robot? No. But the technologies I represent… potentially, yes. These could be dangerous.
“So this is really a question for you. What do you want from your future? Are you aware of what AI might bring?”
The speed all this is moving at is breathtaking and there is no putting the genie back in the bottle.
When we set out to make this episode of ITV's Tonight show, we wanted to explore what all of us need to know about AI.
We found some interesting answers, but it’s the big questions surrounding this remarkable technology that matter and the need to keep asking them, every step of the way.
Watch A.I: What You Need to Know on the Tonight programme on ITV 1 and ITVX at 8.30pm on Thursday, May 25
Want a quick and expert briefing on the biggest news stories? Listen to our latest podcasts to find out What You Need To Know...