I now believe AI has the potential to bring down civilisation

[UndercoverElephant]

So I have had a bit of a wakeup call this morning. Until now I have viewed AI as part of the general shitification of the internet and the world. It is superficially impressive, useful for summarising stuff quickly, but soul-less and prone to mistakes. Not good, but not catastrophic either.

This morning I have been banned from reddit by an AI which completely misinterpreted what I said. It was in a thread about Boris Johnson saying how bad it would be to stop arms sales to Israel, somebody else had said that Israel's actions in Gaza was creating more enemies than it was destroying, and my comment was "And at this point you might as well extend that to the whole population of Gaza. Everybody in Gaza now hates Israel about as much as it is possible to hate them." This was first misinterpreted by a human moderator of ukpolitics subreddit to be hate speech, and I replied to that message saying "I think you've misinterpreted what I said. Who is it hateful towards? Can't be the Palestinians, because I'm on their side, and it can't be my hatred of Israel because I am saying that Israel is making Palestinians hate them, and I'm not Palestinian." That was last night. This morning I have got a message saying that an automated system has decided that my comment was inciting violence, and I've been issued a permanent sitewide ban by an AI.

Don't get me wrong -- I was expecting to be banned eventually anyway, and my life won't be poorer for not having reddit in it. The wakeup call is about the potential for AI to make really bad mistakes while people are ever more trusting of its flawed behaviour. I can only assume that the AI has actually reversed the actual meaning, because it didn't understand the context. It has analysed my post and has wrongly decided that I meant something like "All Palestinians are like Hamas now. All Palestinians are hate-fuelled terrorists."

AI is suddenly everywhere. It is writing books, documentation, reviews of products, it is driving vehicles, pretending to be people. It will replace online help centres that have already replaced call centres. It is probably already making decisions about how to deploy weapons. Pretty soon it will be flying military aircraft and operating nuclear power stations. Given that everybody is also currently trying to save money, there is a huge incentive to replace humans with AI in all these areas and every other corner of modern society. It's super-fast, it looks impressive, and 99% of the time it get things right. The problem is that 1% of the time it gets things wrong, and occasionally it gets things really badly wrong.

I think we are heading for major AI-related problems. It takes years for our culture to respond to advancing technology -- we've seen this in other areas already. We need a whole new corpus of law placing restrictions on what people can use AI for, always keeping in mind this combination of looking impressive, saving time, saving money, but with no means of preventing occasional catastrophic errors. My fear is that by the time we've learned how serious the consequences of AI mistakes are, it will be too late.

[Ralphw2]

Of course, the sector of the global economy that will adapt to AI fastest is the criminal one. It will make scams, on line and on phone, ten times harder to spot and a thousand times more frequent. Banking fraud will be an obvious target. I suspect Ukraine is already using AI in its drones to select targets at Russian airfields, as it is immune to GPS and other jamming methods.

I fully expect AI to be used in front line medicine, to filter out who gets to see a real doctor, and by the police and security services, all done with no oversight and no recourse to appeal, because the computer has been selected to operate the algorithm, even if there is no identifiable algorithm to operate.

[UndercoverElephant]

Yes to all that. It takes "computer says no" to a whole new level of sinister.

https://www.youtube.com/watch?v=x0YGZPycMEU

[Catweazle]

I have recently almost been misled by an AI.

I searched online for statistics of harp sales, market research.

I found a well written summary report predicting increased future sales, current sales, links to all the major manufacturers etc. Also, a link to allow me to buy the full report.

Naturally, I was suspicious, so I repeated my search substituting harp for mandolin. A very similar report appeared amongst the Google results.

Then I searched for banjo. There was another similar report.

I suppose it's possible that harp, mandolin and banjo are all going to have increased demand because of a range of factors from COVID to the global economy, but I'm not convinced.

https://medium.com/@krish.reportprime/b ... 22f1a16e32

[UndercoverElephant]

I have recently almost been misled by an AI.

I searched online for statistics of harp sales, market research.

I found a well written summary report predicting increased future sales, current sales, links to all the major manufacturers etc. Also, a link to allow me to buy the full report.

Naturally, I was suspicious, so I repeated my search substituting harp for mandolin. A very similar report appeared amongst the Google results.

Then I searched for banjo. There was another similar report.

I suppose it's possible that harp, mandolin and banjo are all going to have increased demand because of a range of factors from COVID to the global economy, but I'm not convinced.

https://medium.com/@krish.reportprime/b ... 22f1a16e32

Yes. That is a perfect example of how AI is popping up, and being used to cynically make money by people who could not care less about the accuracy of the information that is being produced. And right now, this isn't even illegal.

Whole books on foraging for fungi, written by AI, and (predictably) containing serious mistakes:

https://www.theguardian.com/technology/ ... tten-by-ai

Mushroom pickers urged to avoid foraging books on DODGY TAX AVOIDERS that appear to be written by AI

Sample of books scored 100% on AI detection test as experts warn they contain dangerous advice

[Catweazle]

As an author you must find this worrying. It's not difficult to feed an AI your whole book, ask it to re-write it in a different style, substitute the photos for open licenced versions, and publish it.

The detail is scary, in one report I saw is an analysis of how the Russia Ukraine conflict will affect the sales of carved top mandolins

[UndercoverElephant]

As an author you must find this worrying. It's not difficult to feed an AI your whole book, ask it to re-write it in a different style, substitute the photos for open licenced versions, and publish it.

This isn't actually a threat. It isn't possible to produce a book that could compete with mine in this way, simply because the suitable open licensed versions of photos of the fungi just aren't available. Not good enough illustrations at resolutions high enough to use in the book. That's why I had to walk for thousands of miles, and why it took me 6 years to produce. I had to actually find the fungi. Where I've had to rely on free images for things I couldn't find are some of the weakest pages in the book.

However, that doesn't make it any less worrying from a the POV of the potential damage caused to people who actually fall for it.

[Ralphw2]

A few years ago my other half wanted to add something to an (dodgy tax avoiders) order to trigger free postage, so she bought me a book on cataracts, as I had been diagnosed with them. It was a small print to order book and was clearly an automatic google translation of a few random articles originally pubished on the web in probably Hindi or other Asian languages. It was of course a complete waste of the very cheap paper it was printed on. Nobody who had even a basic grasp of Engish could mistake it for anything else.

The world has moved on a long way in a few short years.

[mr brightside]

This morning I have been banned from reddit

I wouldn't be too miffed at that, all clouds...

[UndercoverElephant]

This morning I have been banned from reddit

I wouldn't be too miffed at that, all clouds...

I am not bothered, no. I knew I was going to get banned at some point, I just didn't expect it to happen quite like it did. In the past I often flirted around the edge of the rules because I wanted to say something but couldn't. But recently I actually made a real effort to abide by the rules, and it was still impossible to avoid getting banned.

[BritDownUnder]

When we each have our

I'm sorry Dave. I'm afraid I can't do that

moment then we will realise.