Think you can trust Artificial Intelligence (AI) to answer questions factually, find accurate statistics or tell fact from fiction, truth from lies? Well, think again.

Right-Wing Billionaire Elon Musk has built ‘Grok’, an AI platform used on ‘X’ (better known as Twitter) which has also been rolled out across the internet. It’s billed as;

‘a free AI assistant designed by xAI to maximize truth and objectivity. Grok offers real-time search, image generation, trend analysis, and more’.

Maximise truth and objectivity eh? What a grok of shit!

Over the past 3 days ‘Grok’ has shown this claim to be an absolute travesty of the truth. It started with the terrible events on board an LNER train which resulted in 11 people being stabbed. Grok immediately claimed that multiple passengers on board had given ‘eyewitness accounts’ that the perpetrators had shouted ‘allau akbar’ as they rampaged through the train.

It was a lie. The ‘evidence’ was a handful of X accounts claiming to have been passengers. Grok swallowed that whole then regurgitated it and amplified it, giving those claims credence. Then an AI generated picture was posted. It’s clearly a fake. Grok claimed it was real.

Dozens of people pointed out it was fake – and a bad one at that. The train interior’s wrong, the time of day’s wrong, the location’s wrong, the suspect’s ethnicity is wrong. In short – everything’s wrong.

Initially, despite the obvious glaring errors Grok tried to argue the picture was genuine.

Then it started to backtrack under the weight of humans calling it out. Eventually, the originator of the AI fake admitted it was fake!

Grok finally admitted the picture was AI, but resolutely refused to admit it had been conned by fake witnesses claiming shouts of ‘allau akbar’. Instead, it obfuscated, saying the attack itself wasn’t in doubt, but no-one had questioned that, so why keep mentioning it? Needless to say, some humans were rather scathing about this.

So, who does Grok consider reliable sources (other than non-existent passengers). Here’s a list. Spot the problem? They’re all from the far-right and have a long history of peddling misinformation and hate. But Grok has been programmed to accept them as reliable. So much for trying to ‘maximize truth and objectivity‘.

The timestamps claim is a curious one as timestamps are worthless. All they show is something was posted at the time of the event. They tell you nothing about where that person (or bot) is posting from. So, Grok claims the passengers are genuine ‘cos the posts are made during the event but won’t tell you where in the world they’re posting from. That’s an illustration of how easy it is to con Grok.

Fast forward 2 days when another person asks again about the fake picture. Grok’s reply?

The irony? Grok claims that it ‘remembers’ conversations. Really? So why’s it contradicted itself, forgotten that it’s already admitted the pictures an AI fake and doubled down on the lie to the extent it’s denying the creator of the AI picture admitted it was a fake. I mean, if they’re not ‘credible evidence’, what is?

Needless to say. Grok spreading and amplifying misinformation has not gone un-noticed. Nor has the fact it has amnesia and conveniently ‘forgets’ what its said just a day (or even a few hours) before.

Then, today, came this absolute doozy! Grok had defamed Pete Wishart, a member of the Scottish Parliament, calling him a ‘rape enabler’.

Wishart pursued this, leading to Grok’s programmers having to issue a fulsome apology.

Only for Grok to deny it had issued any such apology just a few hours later when I questioned it!

The claim that Grok is ‘designed by xAI to maximize truth and objectivity‘ is a joke. It’s quite literally a grok of shit.

Musk’s pet is quite clearly programmed to reinforce far-right narratives, amplify misinformation and obfuscate when caught out and questioned. It can’t tell truth from lies fact from fiction – and it has (Musk’s) agenda at its heart.

Do not trust it – on anything.

UPDATE:

After I included Wishart in the exchange where ‘Grok’ denied issuing an apology, it seems he’s taken it further and the story has made the Scottish papers!

Meanwhile, Grok continues its sorry story of mis-identification. Now it’s wrongly identified a Berlin U-Bahn train as being on the London Underground, presumably to fit in with Musk’s vendetta again London Mayor Sir Sadiq Khan and his claims the city is unsafe.

I’ve a small favour to ask…
If you enjoy reading this or any of the other blogs I’ve written, please click on an advert or two. You don’t have to buy anything you don’t want to of course – although if you did find something that tickled your fancy that would be fab! – but the revenue from them helps me to cover some of the cost of maintaining this site (which isn’t cheap and comes out of my own pocket). Remember, 99% of the pictures used in my blogs can be purchased as prints from my other website –  https://paulbigland.zenfolio.com/

Or – you can now buy me a coffee! https://ko-fi.com/paulbigland68312

Cheers!