Can we really trust AI to tell the truth?

Like personal computing, smart phones and Google, artificial intelligence is the latest unstoppable technology to revolutionise our lives. Think of it as a work-from-home employee who is the cleverest in your team.

It will be the best copywriter and marketeer, and the hottest graphic designer. It will be your most accurate diagnostic doctor, your smartest lawyer, your most capable architect and your leading expert, regardless of industry. In fact, it will be your top employee in any job that requires knowledge and raw intelligence.

Of course, it will be missing a few human traits, such as warmth, humour and feeling, but have no doubt that within a few years those will be faked well enough to fool most people over a Teams meeting.

Read more:

• Big tech wants to mine human creativity for AI training. We say no
• Labour wants to use ‘Minority Report’ AI to predict murder. This could have chilling consequences
• How do we stop AI becoming an ‘engine of inequality’, and make it a force for good?

AI will be in your home too – advising, assisting and even befriending you and your family. According to which product you choose, it may have a distinct personality and appear to show an emotional response.

But before we get too excited about being able to find out when the next season of The Last of Us is coming out without having to lift a finger, let’s consider the downsides of AI assistants and virtual employees. Because even Elon Musk seems to be having difficulty controlling his.

In June, when X’s AI assistant, Grok, started spouting views that Musk found unacceptable – that more political violence has come from the right than the left since 2016 – he threatened to retrain it and “rewrite the entire corpus of human knowledge”.

If the owner of an AI assistant is trying to influence what results it produces and even he isn’t happy with the results, it should make us stop and scrutinise the inevitable move to AI and make us ask some important questions, such as:

Who actually owns and controls the intelligence of the AI assistants we are using? What protections are in place to keep what they learn in our homes and workplaces private, and can we really trust them to be truthful? Can we trust them, full stop?

As a writer of conspiracy thrillers centred on this very topic, it’s interesting to watch the development I’ve been writing about in a fictional context play out for real.

Earlier this year, the government’s technology secretary was reported to have held multiple meetings with tech sector employees, and now the government has come to the decision that it is pro-AI. Specifically, pro-AI assistants.

The government has told us we should all rush to employ these digital know-it-alls, and feel what it calls the resulting “exhilaration”. It’s so set upon this path that it’s chosen to lead the way, starting with its own civil servants. They will get to feel the exhilaration by working with an AI assistant called Humphrey.

Developed by the government but based on software provided by big tech AI organisations, the AI assistant will start by collating consultation responses and taking minutes of meetings.

It’s a clever choice of name, taken from Sir Humphrey Appleby, the fictional permanent secretary who assists his minister, Jim Hacker, in the 1980s TV sitcom Yes, Minister.

Except… the name feels like an IT department’s ironic joke. Sir Humphrey’s real aim was never to assist his minister’s agenda, but to block it.

He did so by withholding important information and by baffling him with long-winded responses. He schemes behind Jim Hacker’s back, working with his fellow civil servants in pursuit of their own agenda with Machiavellian delight.

So, whose agenda will the government’s AI assistant, Humphrey, really be serving? Will it be serving the agenda of those using it? Those who commissioned it? Or those who wrote the software engines behind it?

This whole debacle so closely follows the plot for my first thriller, Dirty Geese, it makes for unsettling reading.

But, putting that aside, the final question which needs to be addressed as the government runs headlong towards an uptake of AI is what will happen to all the flesh and blood copywriters, marketeers, graphic designers, doctors, lawyers and industry experts?

Divinity Games by Lou Gilmond (Armillary Books, £9.99) is out now. You can buy it from the Big Issue shop on bookshop.org, which helps to support Big Issue and independent bookshops.

Do you have a story to tell or opinions to share about this? Get in touch and tell us more.