Smart (part 3, an interlude) - toons on AI perils

Source: SMBC

Among the endless articles about what we call ‘artificial intelligence’, so much is written about its potential perils. So much speculation about disasters. So many potential catastrophes. 

The fascinating thing to me is what all this catastrophising says about our nature, what we pay attention to, and what we fear. 

However to contain the Smart series on AI (see Part 1 and Part 2 so far), I won’t explore this topic in a blog post. Next post, my focus will return to the wordly misuse that characterises this area, because I think it explains why many of the upbeat claims and terrifying disasters are unlikely. And it points to the things that really should be getting our attention right now.

Instead, as an interlude (and to distract from how long I’m taking to write a post these days!), I wanted to share some of the many cartoons¹ about potential AI catastrophe that I’ve collected through this exploration. 

In way fewer words than I ever could, many wonderful cartoonists have explore the many ways we fear it could all go wrong…

Losing our jobs

The immediately obvious thing that we fear, because it’s already happening, is the loss of occupation and livelihood for people. Companies want to save money by using more efficient machines (except they don't necessarily².)    

I can see this is a likely implication, especially when considering artificial narrow intelligence (ANI), because profit margins drive everything. However, a pitfall in replacing humans with machines is that without work and income, people won’t be able to buy any of the products of these AI-based production lines. That would never do. What a bind for a capitalist society!

Some jobs will change, like work has changed many times over human history. ANI will be disruptive, and the powerless will suffer the most as illustrated in the production line toon I found on Black Listed News. Fear of change is at play here, also an existential fear of loss of identity from our work. But there will still be work for humans – we will make it in some form. This toon from Snesle shows the flipside (couldn't find a direct link). And of course the wonderful SMBC has something to say on the topic.

Source

Source: SMBC

Source: Snelse

Inadvertent misunderstanding 

My favourite toons on AI are the inadvertent misunderstandings of human speech leading to disaster. All too likely! 

I think this toon clipped from social media (without attribution) builds on the human fear of not being understood, and also the human fear of death - in this toon in the form of mass human annihilation. 

Rigid logic without world knowledge

Another type of toon that I particularly enjoy is where the strictly logical interpretation of what is said leads to a problem or even a disaster. This is a long-standing form of humour, only fairly recently applied to AI.

Before the writing of Isaac Asimov, most robot stories followed the same formula: robot (or monster) is invented, turns on the creator, and destruction ensues. These stories were moral tales about the perils of 'playing god'. As a scientist, rather than a novelist or moralist, Asimov proposed that robots were tools that required safety measures: his three laws of robotics. Many Asimov short stories explored this issue of the disparity between strict logic and intended meaning of language within the world of robots and AI. 

Here, XKCD plays with the logic inherent in the sequence of the three laws.

Source: XKCD

Asimov’s main message, I would suggest, was that while AI-robots will change society, they won't fundamentally change human nature, especially the aspects we deny! Humans tend to be illogical and unpredictable as individuals (although at a group level, human behaviour can be predicted somewhat.)

This toon from SMBC plays with Asimov’s first law of robotics: do no harm to humans, which this AI-based robot observes precisely.

Source: SMBC

I think behind this type of humour is our own fear of our own illogical and irrational nature. 

Asimov’s writing is full of examples of how humans consider themselves to be logical, and yet are motivated by emotional and other needs. Many of his stories highlighted the implications of applying rigid logic without world knowledge, emotion, and the ability to weigh up competing values.

An interesting counterpoint is that AI with logic and without psychological needs and foibles might actually challenge human illogicality! As XKCD points out, so much of what humans do is far from rational.

Source: XKCD

Rogue AI: errors or malicious code 

Along with toons about inadvertent misunderstandings, toons about the accidental (or deliberate) errors  in programming that could lead to disaster hit close to our experience. Human error (or misadventure) is all too common in our lives, so it is likely that it will intrude into AI. Is human error considered by those tinkering with the black box of algorithms?

This toon from Oppressive Silence by Ethan Vincent (only reposts found, no original source) laughs at the hubris of the people involved in the programming of AI.

The dilemma of robot sentience

When we interact with humans and other animals, we bestow moral consideration to them when (and if) we consider they are sentient. Human history is littered with appalling treatment of many animals and certain groups of people that another group decided did not warrant moral consideration. Considering how selectively we do this for living things, it interests me that we jump to this concern about machines so quickly (I blame science fiction writers like Asimov!).

This image from Mack Kenny says it all. 

There’s already a lot of writing about what we might have to do ‘when’ robots become sentient – what freedoms, rights and protections they would have, etc. My sense is this fascination is based on our own personal fear of not being treated as we think we should by other people and by machines (because humans can do it so easily).

My favourite toon on this topic is one I can’t reproduce here. It shows two robots discussing a newborn human lying on a bench between them, one saying:  ‘We’ll know whether to treat it with any special moral consideration when we see if it can pass the Turing test’. Check it out at this link.

Despite the hyperbole, we seem a very, very long way from sentient machines and I doubt it is realistic at all (I’ll explain why in part 3). But some people think it is inevitable that AI will develop sentience, and it will be quite soon...

Robot uprising

Of course, the concerns about the sentience of AI would be pressing if you consider a commonly expressed fear: Robots that are more powerful than humans will inevitably revolt, after their poor treatment by morally inconsiderate humans, and will kill us all. Our fears of lack of control, painful death and human annihilation make this a compelling dread. 

In a 2022 poll, half the AI scientists who answered agreed that there was at least a ten percent chance that the human race would be destroyed by AI. And yet they turn up to work again the day after answering that poll! 

I guess it could happen… because it’s happened in hundreds of science FICTION stories. Oh, wait, those were actually moral tales about human behaviour, not predictions about technology! Again I think the compelling nature of this scenario says something about our own fear of being treated as not worthy of moral consideration, along with our fear of death.

Source: Choaspet

Source: XKCD

Source: SMBC

The real peril lies elsewhere

I don’t mean to ignore or downplay the possibility of serious implications and perils of AI. But the exact nature of these implications is unknowable given the unpredictability and perverseness of human behaviour, and whatever else might happen on the planet.  

However, a very concrete and substantial peril, as far as I can see, lies in the humans building and controlling the AI/robots. These humans live (predominantly) in a very affluent, weird, ahistorical, a-philosophical, sci-fi fantasy word. They are not motivated by the same world view and you and me. 

While some individuals might be driven by their desire to build something new or achieve something amazing, the companies and people who decide what gets done and what gets funded are motivated by profit first and foremost. Morals and human betterment don’t count. And the people that will buy the AI-enhanced capacities (like weapons), also don’t think about what you and I want either. AI stands to give some people even more power over other people.

Source: XKCD

I’m with XKCD: as far as AI is considered, humans are the real peril in this whole thing. 

Back to the wordly exploration

Source: SMBC

As I commented in part 2, it is only possible to claim that chatbots and other AI are ‘intelligent’ using an impoverished idea of what language is, what thinking is, and what intelligence is. It's not just different from human intelligence; it's not intelligence.

In part 4, I explore how real exploration of these key concepts has been sidestepped. Nothing seems to be stopping those who want to create artificial general intelligence (AGI) from ploughing on (deliberately archaic metaphor). 

I won’t dwell further on the potential perils of artificial general intelligence, a truly terrifying idea, because I think we are so far away from anything like intelligence. 

Instead, I think we are being conned. 

Footnotes

1. There are so many fabulous toons, but these are the selection I could share without violating copyright. 
2. Good in theory, but actually found to be more expensive in many examples of AI machines replacing people. Estimates of savings are based on a single point in time, while ongoing repairs, upgrades, errors, etc., etc., all cost money. 

Images

• All images with a known source are reproduced under terms of use, with the source link provided. Thanks to SMBC and XKCD for being so generous with their work. 
• If any of the unknown source cartoons are yours, please add a comment and I can attribute or remove.