Add a Comment   Back to OJB's Blog Search Page

An AI Apocalypse

Entry 2364, on 2024-09-27 at 12:25:18 (Rating 2, Computers)

I have talked about AI a bit recently. Why? Well, it is the most important "big new thing" we need to be aware of currently. In fact, it might be the only thing that really matters, and there might not ever be anything more important, even in the future.

Am I exaggerating this idea, just a tiny but? Well, maybe. After all, they say that predictions are difficult, especially those which involve the future! But I am in good company, because many knowledgeable, intelligent people have been issuing similar warnings for a while now.

So what is my concern? Well, in depends on how I project out these changes and whether I assume they follow the changes from previous revolutions (agricultural and industrial) in the past.

As new technology has become available societies have changed to fit. Tech isn't the only thing that drives social change, because politics, religion, and other mechanisms also affect it, but technology is arguably the most important.

There is significant debate at the moment about just how intelligent current AI systems are. Some people deny they are intelligent at all, but the biggest problem here is defining the word "intelligent". It seems that every time we come up with a definition and test of intelligence, AI passes the test (often very easily), so we just change the definition. It's one of those things that "we can't define, but know it when we see it".

In a blog post titled "The Right Priorities" from 2024-07-01 I discussed the more extreme reasons that AI could become a problem. I got some disagreement on that, and fair enough, I was discussing the more extreme, but more uncertain, attributes of AI which are certainly questionable.

But the more widely accepted results from the widespread application of AI involve replacing humans with machines in the workplace. Truck driving is often used as an example of one of the occupations which AI will take over first, and finding alternative employment for those drivers is often cited as a major problem, but I don't think it will go that way at all, at least not initially.

In my opinion it is the more professional class of jobs which will go first. This includes jobs in law, accounting, writing, acting, and programming. Yes, it is the professionals who can be replaced most easily, not the "lesser" semi-skilled jobs which (ironically) are a lot harder for an AI to replicate reliably.

So instead of unemployed truck drivers roaming the streets and causing havoc we might have a pile of lawyers and accounts out there causing havoc! And yes, it hasn't escaped my attention that a significant skill I have (programming) is also on the list of jobs due for replacement!

So here's an example of what I think we might have in 10 or 20 years time. Imagine I want to watch a movie about an AI taking over the world. Instead of choosing an existing movie I will enter a brief description about what I want to see and AI will do the rest...

An accounting AI will calculate the costs involved and bill me, it will send it to a legal AI to make sure it's legitimate from that perspective, then my description will be sent to a script writing AI which will trigger a movie AI to create the movie, including all dialog and acting, with no human intervention at all. After a few seconds my movie will start playing. It will be unique and utterly convincing.

Does this sound crazy to you? If it does, maybe you haven't been paying attention to what is already happening, because all of this is already possible, although in a relatively primitive form, which is why I specified "in 10 or 20 years" above.

It sort of appeals to me that people in allegedly highly skilled jobs like lawyers can be replaced, but those in "lesser" jobs like truck driving are a lot more difficult. I'm sure that automated trucks will arrive too, but pure information processing (which is what many of those professional jobs are really all about) is currently a lot easier for AI to handle than interaction with the real world.

Already we are seeing a lot of professions using AI to help with their jobs. Many media companies use AI to write articles, for example, with varying results. Currently the AIs often make mistakes and have trouble distinguishing fact from fantasy, but so do many humans, and yes, remember that 10 to 20 year thing? At this point a human really needs to check the material being created but that is only temporary. Eventually another, independent AI will fact check all material.

So is this good or bad? Well, it doesn't really matter, because the people who thought other technologies were bad had little influence on how those technologies were deployed. There will be no choice, because a conventional movie costs tens of millions to produce where I would expect an AI movie to take just a few hundred dollars at most, and possibly be free.

Are there any jobs which are safe? Well, probably not. It might be possible for all the work currently done by humans to be done by AIs and robots controlled by AIs. Humans might not need to work at all, and remember that work in the form we have now is just an invention of the industrial revolution, there is nothing inevitable or natural about it.

Humans might be able to get on with more rewarding things, like social interaction, consuming content, engaging in games and sports, and just generally enjoying themselves doing what they want to do instead of what they have to do.

I'm trying to put a positive spin on this, but it is going to be tough. Maybe it will cause major societal breakdown. Maybe it really will be an AI Apocalypse.

Comment 1 (7724) by Anonymous on 2024-10-01 at 16:09:31:

I cannot understand why people like you think AI is such a big problem. All they do is write sentences using what they have seen in the internet. How can this be a threat?

Comment 2 (7725) by OJB on 2024-10-01 at 22:41:39:

Yes, a friend who is a computer scientist said something similar. But I think that view is ignoring the speed the technology is improving at, and also the unintended consequences and emergent properties this sort of technology might have. Have you read two other relevant blog posts I've done on this: "The Right Priorities" from 2024-07-01 and "What Could Go Wrong?" from 2021-03-25?

Comment 3 (7726) by EK on 2024-10-02 at 16:19:20:

very timely discussion. Let's hear more of this. Beyond job losses, is it going to improve the quality and quantity of knowledge, does it break open difficult topics (like in theoretical physics, cosmology) for humans to get a cognitive handle on? In other words, can it potentially enable humanity to take another evolutionary mental step (provided AI remains under human control).In a futuristic society unemployment presumably can be addressed (see "The Future Shock", cant think of the author's name at the moment).

Comment 4 (7727) by OJB on 2024-10-02 at 18:17:40:

It should be the best thing ever for extending knowledge. Already it has enough knowledge to pass medical and legal exams at a level equivalent to above average humans. Surely it is only a matter of time before it is vastly better than any human. Whether it will allow humans to take an evolutionary step in the biological sense is doubtful, but maybe the future is more likely to involve humans with AI enhancement. Elon Musk is already working on computer to brain interfaces ("Neuralink") with some success, so again, it is only a matter of time. Maybe the next evolutionary step involves physics, not biology.

Comment 5 (7728) by EK on 2024-10-02 at 22:21:18:

Just for completeness' sake. The author's name is Alvin Toffler. The book was a big hit. Among other things it predicted that work will be replaced with increasing leisure time filled with self-improvement.
We seem far away from any knowledge improvement by AI. Maybe I expect too much, but just for fun I have used chatbot a few times to explore some topics. The result: absolute mind-numbing crap. Uninformed nonsense, misinformation, contrary to good sense (or what is conventionally regarded as such) in short no reason for optimism. At best we seem decades away from anything useful.(Funny on the side: asking it about myself, it told me that I had died years ago.)

Comment 6 (7729) by OJB on 2024-10-03 at 09:20:44:

Yes, Toffler's book is very well known. Don't think I've read it, but I have certainly heard of it. That prediction is an interesting one, it has been predicted for a long time. Will it happen? With AI, maybe.

Yes, there are many stories of AI failures out there, yet there are also the successes with it passing medical and legal exams, writing quite competent programs, and producing accurate and succinct summaries. I think the next stage needs to be some sort of verification process to check output is accurate. It only produces what it was trained on, but the same thing could be said about humans!

You can leave comments about this entry using this form.

Enter your name (optional):
Enter your email address (optional):
Enter the number shown here:
Enter the comment:

To add a comment: enter a name and email (optional), type the number shown, enter a comment, click Add.
Note that you can leave the name blank if you want to remain anonymous.
Enter your email address to receive notifications of replies and updates to this entry.
The comment should appear immediately because the authorisation system is currently inactive.