Wednesday 31 May 2023

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.


Tuesday 16 May 2023

Is it time for Britain to adopt "geniocracy"

There's suddenly a lot of fuss about vote-rigging with Labour and the Conservatives both bitterly accusing the other side of "gerrymandering".

The Tory press have been bitterly accusing Labour of vote-rigging over the idea of expanding the franchise to 16/17 year olds, and to people from overseas who are long-term resident and tax-payers in the UK.

Keir Starmer is such a dishonest and unprincipled flip-flopper it's impossible to know whether he'd actually go through with what he's "pledged", but the principle is sound. Giving more people the right to vote is a good thing, within reason.

Then there's the absurd Tory toff Jacob Rees-Mogg, who outright admitted that the Tory government implemented new voter ID rules in order to gerrymander elections in favour of their own political party, which is exactly what a lot of critics accused them of doing as they were doing it.

Lots of Tories have tried to argue that it's ridiculous to give votes to 16/17 year olds because they're too ignorant and uninformed about the world to use their votes wisely, however it's an absurd generalisation to pretend that the average 16/17 year old is more ignorant and unwise than the most ignorant people in older generations, who all get the right to vote, no matter how dull-witted or politically illiterate they happen to be.

This Tory insistence that youngsters are too ignorant to be allowed to vote raises the question of why intelligent and informed 16/17 year olds should be denied the vote, while the even the most stupid and gullible of over-18s get exactly the same voting power as people who actually know anything about what's going on.

If the Tories are right, and ignorance is something that needs to be combatted in elections, there are ways to do it.

People could do a simple politics test at polling stations, with the results defining the weighting of their vote, with the votes of those who score highly on the test counting for more than those who get all the answers wrong.

I'm not talking about some kind of intense and intimidating exam, just a check up on the absolute political basics:
  • Who is the current Prime Minister?
  • Who is the current head of state?
  • Who is the current MP in your constituency?
  • Which is the UK's finance minister? [Home Secretary, Chancellor of the Exchequer, Leader of the House of Commons, Speaker of the House of Commons]
  • Which is the UK's upper legislature? [House of Commons, House of Lords, Supreme Court, Houses of Parliament]
  • When is the name of legislation when it becomes law? [White Paper, Bill, Act, Point of Order]
  • Which level of government is responsible for libraries and bin collections? [The Government, House of Commons, House of Lords, Local Authority]
  • Which organisation isn't Britain is a member of? [UN, NATO, WTO, EU]
  • What is a measure of national wealth? [Average wage, GDP, Population density, Gini Coefficient]
  • Which political ideology is based on public ownership of core industries? [Capitalism, Liberalism, Socialism, Conservatism]
The first three could be set questions, followed by a random selection of (relatively easy) multiple choice questions about politics.

Tories would obviously scream that it's unfair to make people do a test before they're allowed to vote, but this would just expose their hypocrisy, because they're the ones who just introduced the principle of making people do things before being allowed to vote (acquiring and bringing the correct photographic ID).

They'd also object to the idea that those who know about politics should have more say than those who know nothing, because they know that tricking the gullible into voting against their own political interests is absolutely crucial to the success of Tory politics.

And in objecting, they'd be demonstrating their hypocrisy and venality. They want smart 16/17 year olds to be collectively punished because some of their age group are ignorant and uninformed, but they want the votes of ignorant and uninformed over-18s to count exactly the same as the votes of those who actually know what's going on!

The name for knowledge-based voting systems is "geniocracy", and with modern technology, it's easily implementable, and the data would be easy to collect and work with.

Hardcore "geniocrats" reckon that only people of above average intelligence should be allowed to vote, and only the very intelligent should be allowed to stand for public office, but there's absolutely no need to ban people from voting at all. The system could simply be operated to assign more weight to the votes of the politically informed than the politically ignorant.

If the post-election data showed that certain demographics (age, region, voting preference etc.) are lacking in basic political understanding, it would be in the public interest to help them to improve.

The data would tell us which parties had been most politically disadvantaged by the ignorance of their own voters, which would create very strong incentives for all political parties to ensure that their voters have a basic level of political understanding.

Of course it's almost impossible to conceive a system like this ever actually being implemented, because no ruling party truly wants an informed and politically engaged electorate capable of actually following what's going on, and holding them to account. But it's an interesting idea, because it's the logical solution to the problem of political ignorance that the Tories say they're so concerned about, and it's entirely compatible with their new policy of making people prove things before voting, isn't it?

 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.