Is AI really a threat to the survival of humanity?

AI experts have once again warned that AI represents a potential threat to the survival of humanity, putting it on a par with pandemics and nuclear war.

The smug conspiratorial response from people who aren't AI experts is that the AI experts are deliberately overstating the existential threat to human survival to drum up interest in AI, which they're doing for their own personal profit.

"The tool I've helped create could wipe out humanity" seems like a very odd way of advertising your services if it's actually just a harmless piece of tat that simply creates new memes and writes fake essays isn't it?

However the dangerous applications and consequences of AI are already becoming clear to those who have been paying attention.

One of the most concerning things is the way AI simply invents citations (research papers, news reports, criminal cases ...) to support the assertions it's asked to make.

One teacher set their students the task of getting AI chatbots to write an essay for them, and then finding the errors for themselves. 100% of the 63 AI generated essays contained hallucinated information, fake quotes, citations to invented sources, and/or misrepresented citations to real sources.

A lawyer in the United States was caught using ChatGPT to write a legal filing, which referenced legal cases that did not exist, and when the lawyer asked the chatbot for its source for the fictional legal cases, it lied that they were found on real legal databases.

ChatGPT has already accused multiple innocent people of being criminals, including a law professor it accused of groping students on a trip to Alaska, citing a 2018 Washington Post article. In reality the professor has never been accused of sexual misconduct, he's never even been to Alaska, and the Washington Post article does not exist.

The fact that AI chatbots are capable of ruining people's lives with false accusations, outright lying, and making up false sources is just the tip of a very dangerous iceberg.

These AI lies are fairly easy to spot now because the fake sources they cite are so obvious, but when challenged the chatbots have attempted to create the fake sources for themselves, to back up their own lies.

You only have to imagine the potential for chaos when these chatbots get better at lying, and better at creating fake sources to justify their lies.

Imagine somebody gets AI to spread lies about an impending nuclear attack on Russia and China by the United States, backed up by fake news articles, fake maps with fake secret nuclear bases in Turkey or wherever, deepfake video of the US President discussing this nuclear attack strategy ... (or threats of nuclear attack on the US and its allies from China or Russia).

We all know that fake news spreads across social media like wildfire, and the when the truth is revealed it invariably travels at a snail's pace and gets only a tiny fraction of the engagement, so how dangerous could a hoax like this actually be?

AI obviously doesn't have to bring the world to the brink of nuclear war to pose a threat to humanity, this is just a hypothetical example.

Imagine the destructive consequences of ever more powerful AI misinformation tools in the hands of the climate change denial lobby; fascist dictators; genocidal maniacs; stock market speculators; corrupt politicians police, and/or secret services; bonkers conspiracy theorists; scam artists; terrorists; or downright trolls who just want to watch the world burn.

As AI lies become more and more convincing, it's going to become even more difficult for people to discern the difference between reality and fiction, to a point where nothing remains believable.

We already live in a world where the boundaries between truth and fiction have been alarmingly eroded away, with so little pushback and punishment against dishonesty that the biggest political liars of all have already risen to become heads of state of some of the most powerful countries on earth (I don't even need to name them do I?)

But AI has the potential to accelerate the process dramatically, so that stock markets and governments tumble based on AI fakery, and real events and scandals are increasingly ignored and dismissed as fake.

We don't have to get locked into a debate over the worst case scenario of human extinction to recognise the very real dangers.

So what could be done about it?

There's already strong resistance to AI regulation from within the industry, and even if regulation were to be introduced, there would be significant problems.

What would stop AI industries simply moving to the countries that have the least AI regulation?

And if regulation was introduced to prevent AI misinformation, where would the line be drawn?

Banning AI from telling lies is nowhere near as simple as it sounds, because every time someone asks AI to do a funny task like find out who Nathan Fielder's "out of frame" friends are in his viral tweet, they're asking AI to lie about reality.

How do you make a computer programme discern the moral difference between faking an image for laughs, or faking an image to cause a stock market crash or to incite political/ethnic tensions?

So, is AI capable of making humanity extinct?

Nobody really knows. Maybe AI experts are overstating the power of their industry for dramatic effect, or maybe things just keep on getting more and more insane and dangerous from here on, until human survival is genuinely at risk.

But what is beyond doubt is that AI in the wrong hands will clearly have the power to spread conspiracies and lies; crash stock markets; topple governments; cause false persecution; accelerate climate change; trigger wars and ethnic cleansing; and erase the boundary between truth and fiction, which are all legitimate grounds for serious concern.

 Another Angry Voice  is a "Pay As You Feel" website. Access to my online writing will always remain free. If you see some value in what I do, please consider supporting my work with a small donation/subscription.

OR