This post was rejected for the following reason(s):

  • Looks like you’re crossposting this from your blog elsewhere. That can be fine, but sometimes we we’ll reject these crossposts (or allow them but with a warning) if they seem like they’re not quite aimed at the LessWrong audience or adopting LessWrong norms. LessWrong has particular norms aimed at getting towards truth (e.g. avoiding rhetoric, being precise, being quantitative, etc), and your post isn't obviously displaying those. Good solutions to this are to (a) write an extra intro to your post to explain why people on LessWrong might find it interesting, (b) rewrite to be a style more aimed at truth than persuasiveness or engagingness.  

Working hands by Aaron Gilson from Flickr.

In James Cameron's 1984 film The Terminator the two heroes are hunted by a murderous AI robot who wants to kill them. In the film's climactic final scenes the two protagonists find themselves in some sort of machine factory. Luckily for them they are able to start up the factory's automatic production lines, something that distracts the robot enough for them to trap it in a hydraulic press and destroy it, saving themselves and humanity in the process.

It is a fine ending to a fine film. There is a slight problem, though. There were no human-free factories in 1984. In fact, there are no human-free factories today, almost 40 years later.

Artificial intelligence can do amazing things these days. Some things are well beyond what mere humans can accomplish. Google's translation engine understands a dizzying 133 languages. AI has been shown to be better than human oncologists at diagnosing some sorts of cancer. And then we have chess, where it was a whole generation ago since a human could beat the AI champions.

According to the experts this is only the beginning of the AI era. Ever more powerful computing chips and ever more intricate algorithms will allow AIs to use its formidable powers of data handling to dominate ever more areas where humans traditionally ruled supreme. Should we be worried?

Great brain, but what about the rest?

Not necessarily. While AIs are very good at what they do, basically juggling large amounts of data, there is a reason those human-free factories do not exist. The real world is not as well-behaved as the digital world inside a computer network. And machinery to handle physical objects is nowhere near as sophisticated as the machinery to handle abstract data.

There is some logic to this. The human brain is a remarkable tool. But it was not developed to sift through large amounts of data in search of patterns. This is a very useful skill in our modern society and our brains spend significant time doing precisely this. But it is not what they were made for.

This is not the case with the rest of the human body. We still use our eyes to visually take in our surroundings. We still use our ears to listen for audio. And we still use our hands to handle physical objects. Our eyes, ears and hands have all evolved during millions, or even billions, of years to do exactly what we are doing with them. If they are good at what they do, there is an excellent reason to it.

While AI has been able to leapfrog human brains by doing things the brains were not primarily made for doing, this kind of leapfrogging is not possible with most other parts of the human body. This is why human eyes and ears are still very much competitive with the best cameras and microphones. It is also why those robotic hands look so clunky and fumbling compared to human hands.

Reality is hard

When I was a student at university I had an extra job at a cold storage warehouse. The warehouse received pallets from farms and factories and sent out pallets to grocery stores. Since no grocery store wants a 1-ton pallet of Italian salsiccia the job of the warehouse was to split pallets into individual boxes and then restack new pallets with a plethora of different boxes to be sent to the stores.

This was all controlled by a computer system that knew exactly the location of every box and pallet in the entire warehouse and also kept tabs on which store wanted what goods at which time. The computer was almighty.

But only to a degree. The computer might have known everything about every box and pallet, but it had no means of actually doing anything to them. That was my, and my colleagues', job. Basically, what we did all day was moving boxes from one pallet to another. Our handheld computers told us which boxes to lift. And we lifted boxes. 

When I started working at the warehouse I was shocked. After all, it was the 21st century and I could not fathom that in this modern world you actually paid human beings to lift boxes all day long. That was robot work. At the time I was studying production engineering and hoped to one day maybe work in robotics. I was naive. Very, very naive.

This warehouse where I worked was not some antiquated piece of industrial property. When I started there it was just a few years old and on the wall outside the office there sat newspaper clippings and diplomas lauding it as one of the most modern refrigerated warehouses in Europe. Apparently, not even in the most modern warehouses robots did the actual work. It shattered my world.

Alarm in the ivory tower

We simple warehouse workers were slaves under the computer. It was no AI, there was no need for one. Our computer overlord was basically a giant database with a simple logical layer on top. Either way, we did not care. To those working with their hands there is no practical difference if your boss is an AI, a computer or a human like yourself. A boss is a boss and the work is the work.

Of course, things look quite different if you are not a grunt on the floor but rather one of the bosses in the office. Our warehouse computer was simple because it did not need to be anything but simple. Keeping track of an inventory is not complicated, only tedious. I do not know what inventory tracking looked like in the age before computers, but I imagine it involved a lot of papers and a lot of browsing through file cabinets. Presumably no one, boss or grunt, wanted that job.

Today's AI can do a lot more than keep track of an inventory. Today's AI has something that could probably be described as agency. When given an abstract problem to solve it can find a solution based on logic and extrapolation from previous experience. That is eerily human-like. At least if you are a human who works with finding solutions to abstract problems.

Fifty years ago robots were also seen as eerily human-like. Maybe not the actual robots that did actual work. Just like today, they were one-armed orange monsters. But robots as a concept were very close to humans. People believed that robots in the near future would surpass humans, as demonstrated by numerous works of art like the Terminator film mentioned in the introduction.

The robots of yesteryear did cause alarm, just like AI causes alarm today. But not really in the same way. Robots were manual workers, the equivalent of the human working class. And what people expect from deviant working class is violent outbursts and wanton destruction. Consequently, robots were primarily seen as a threat to humans physical safety. 

AI is not like the working class, it is very much like the professional class. From the deviant professional class we do not expect random acts of violence but rather disobedient decision making or some terrible policy error that causes a large-scale catastrophe. AI alarmism has, predictably, focused on ways our leaders can do us harm. 

Do you want thermal paste with that?

The introduction of AI will have profound effects on the global economy. But it will most probably not have very much effect on global humanity. At least not the way it plays out in the minds of the AI alarmists where AI is a threat to the very existence of humanity.

The reason AI is not a threat to humanity is the same reason I was lifting boxes in the cold storage warehouse: Machines simply can not do everything that humans can. At the very least, machines can not do everything that humans can as well and as cheaply as humans can do it.

This is very visible in the economy of today. Take a look at Youtube and you will find countless videos of fruit-picking robots. Take a look at any orchard and you will inevitably only see human fruit-pickers. While there are many robots capable of picking fruit there are no robots capable of picking fruit as well and as cheaply as humans.

The superiority of human dexterity will not disappear anytime soon. One day robots might equal humans in physical ability. But this day is not today. It is not tomorrow either. The time when robots will outdo humans physically is probably more centuries than decades away, if it ever comes.

AI exists in the digital world, but the digital world exists on physical servers and networks. Someone needs to maintain this physical infrastructure and that someone is human. Even an AI needs a new GPU from time to time. And when it needs one the human server technician will be there to provide it.

Any AI with the intelligence to pose an existential threat to humanity will understand this simple logic. A brain needs its hands. No matter how smart the AI gets, humanity’s future will be assured as long as we can provide vital services to the masters of the digital realm. It might not be the future most AI alarmists envision for humanity, but for most people on Earth it is a future very similar to the present. 

Who's in charge?

There is another reason why AI is not a real threat to humanity. The same reason that the professional classes of today are not a real threat to humanity. They are not really in charge.

The world is composed of leaders and followers, bosses and workers. Nominally the bosses are in charge and decide what the workers should do. But this only works as long as the workers actually follow the orders the bosses give them. At every given moment the followers have the ability to stop following the leaders.

At my old cold storage warehouse we were sometimes given strange instructions by the computer overlord. It could be to fetch boxes from an empty pallet or stack our pallets way higher than would be advisable. As a general rule we did not follow these orders. Instead we went to the (human) supervisor and reported an error. We might have been grunts, but we still had common sense and could stop things before they turned into accidents.

An AI boss will be even more dependent on its workers than a human boss. The human boss can always decide to do the task at hand personally, lead by example, so to speak. The AI boss has no such ability. It is wholly dependent on its powers of persuasion to get anything done in the physical world.

AI alarmists seem to believe that AI will surpass humans in persuasive skill by a significant margin. Maybe they are right, maybe not. I do not think it matters very much in practice. Humans are used to being subject to persuasion. Significant resources are invested in persuading us to do things (advertising, politics, etc). Most of it is simply ignored. An AI that gave orders to destroy the world would most probably be ignored as well.

The bane of the bullshitting class

Eyes, ears and hands are all very old biological functions. Language is not. The comparative newness of language might go some way to explain why humans are not very good at it. Human babies take years to develop even basic language skills. And even fully adult and well-educated humans sometimes struggle to find the right words to convey the information they want to convey (it even happens to blog writers sometimes).

Humans’ lack of language skills have been magnified by the paramount importance language has attained. Language skills are what underpins social skills and social skills are key to advancing in human society. If you want to get ahead in life you need excellent language skills.

If skill with language is the number one trait that decides your status as a human being, it is intriguing to reflect on what will happen when AI suddenly gets ahead of all humans in this particular field. Logically, AI should be able to monopolize all high status positions and, by extension, rule us all. 

But there is another potential development as well. If AI surpasses all of humanity in language skill then language skill will cease to be a gateway to social success. If everyone has access to a silken tongued AI then silken tongues will no longer be a way to get ahead in life. People will just ignore the well-spoken and social success will have to be decided by some other metric.

Phrased differently, since AI is a world-beating producer of bullshit, it will simply saturate the world with bullshit and all those who did bullshitting for a living will find themselves out of work and out of social position. AI might not be an existential threat to humanity but it is definitely an existential threat to the bullshitting class.

Humans of the world, unite!

The problem with many bosses is that they believe they are leaders. They believe that their job is to supervise and guide the work of their underlings. They believe that without their constant leadership nothing would ever happen.

This is not the perception from the view of the workers. In the eyes of the workers, bosses who fancy themselves leaders are more of an obstacle than an aid. Workers are fully aware of what they are supposed to do. If they do not want to do it, a sinister looking boss will not change their mind so much as make them better at shirking their duties.

Workers still need bosses, though. What many bosses overlook is that they are in fact service personnel. Workers need someone to organize the jobs at hand and coordinate with other workers. Ideal bosses are efficient administrators who are always at hand when the workers need them and never in the way when workers do not need them.

This sounds suspiciously like an AI. AIs are great at mundane administrative tasks and should be able to coordinate with other workers, especially if those other workers are also supervised by an AI. And although AIs are prone to bullshitting they generally keep quiet if told to keep quiet. To a simple worker, AI looks very much like the perfect boss.

It could be a match made in heaven. AI is a great administrator but has no abilities to handle things in the physical world. Manual workers are great at handling physical objects but have limited capacity, or willingness, to deal with organization and administration. Together they can do great things. 

Just like household appliances freed the professional class from domestic servants, AI may free the working class from middle managers. AI alarmists warn that AI will become our sinister overlords. For a significant part of the population this warning has no credibility. For the simple reason that they already have sinister overlords. For them, AI is not a threat so much as an opportunity. An opportunity to break free from the overlords and finally manage their own affairs. 

AI to the rescue! (Screenshot from the film Office Space)

New Comment