Why we don't trade with ants
...and why hypothetical AI:s won't, either.
Summer 2023
In a fairly recent text, Katja Grace writes about a hypothetical AI and why it would or wouldn't trade with humans. The AI entity described is a superintelligence - presumably orders of magnitude cleverer than humans - and neither inherently benevolent nor malicious.
The argument Grace wants to refute is that such a superintelligence would likely not trade with humans because we'd simply be too weak and stupid. That is, if we've got something of value to the AI, it could easily take it from us using manipulation, coercion or force - but, argues Grace, it would be just as likely to trade with us.
This is all hypothetical of course, but I find it interesting regardless - not least because it tells us something about how the magic promises of AI seems to blind us to facts that are easily observable all around us in the real world.
Honey, honey, honey, in a rich man's world
Grace spends some time arguing the usefulness of ants, but I don't really think that's what's on dispute. Grace seems to have overlooked this, but the fact that insects (and even more primitive creatures) can be of use to humans has been common knowledge for millennia. For example, I recently battled a fungus gnat infestation using nematodes. It was very successful - for me. The nematodes kill off the gnats by feeding on their larvae, which means that when the job is done, the nematodes all die of starvation. Not exactly what I'd describe as a trade, but a good example of the usefulness of what we consider to be a very primitive species.
Honey bees are another, perhaps better example. This is clear and apparent wealth extraction, and one could argue that beekeepers provide the bees with some kind of safety, and that this constitutes a trade. However, it's not - it's trickery. The bees are fooled into doing our bidding and we can fool them because we're smarter. Are we actively cruel to the bees? I'd say no, but wealth extraction without outright cruelty doesn't constitute a trade.
The concept of trade
Grace writes that if we could tell an ant colony living in our home that we won't kill them if they simply move elsewhere, this would be beneficial to the ants. This is true. As is her claim that it would be even more beneficial to them if we offered them food instead of a death threat - but the death threat would of course still be there, implicitly, if the ants for some reason refused to move. What's described here is not a trade, but coercion by means of lethal force. I'm sure I'd pay handsomely for a smelly old sock if it was offered to me at gunpoint, but I wouldn't exactly call the transaction a trade.
What, then, is in fact a trade? One simple and uncontroversial definition is a voluntary exchange of goods and/or services resulting in some kind of benefit to all involved parties. As such, trade implies active decisionmaking and mutual consent. If one party isn't happy, they might choose to not trade at all, or find a trading partner that can offer a better deal.
The article briefly mentions that ants might lack the concept of trade, but then goes on to claim that we don't trade with ants because we can't communicate with them. But that of course begs the question - why can't we talk to ants?
On insect intelligence
A popular example about insects that sometimes pops up in discussions about free will or intelligent agency is that of the female digger wasp.
After laying eggs in a hole in the ground, the wasp brings a food stash (usually a paralyzed grasshopper) into the hole and leaves it there for the larvae to feast on once they've hatched. The wasp observes a small ritual when doing this, which consists of the following few steps:
- Bring food to hole entrance.
- Enter the hole to look for enemies, leaving the food outside.
- If the hole is clear, go back and bring the food inside.
If the food is moved a little bit when the wasp is inspecting her hole, she can be fooled into a lengthy repetition of this ritual. The story about this behavior usually claims that an infinite loop can be provoked, but most recorded experiments report between five and twenty repetitions before the wasp gets wise to the prank.
Even if triggering repetition was impossible, this ritual tells us something about insect intelligence. If had I left a box of food outside my dwelling, gone inside to fix something, and then returned to find the box had been moved, I wouldn't just haul the box inside and start offering my children food from it. It would be far more pertinent to identify the reason why the box had been moved, and ascertain that it hadn't been tampered with in some way.
The wasp, of course, doesn't realize that a powerful superintelligence is subjecting it to a cruel experiment. But, more interestingly, the thought of something being wrong with the grasshopper and the outside surroundings - parasites, a predator, insufficient paralyzation - is simply beyond the scope of wasp reasoning. No matter how many times the procedure is repeated, the grasshopper will eventually be dragged into the hole and stored together with the precious wasp eggs.
In short, insects don't have a concept of trade because they're not sufficiently clever. This is also why we can't communicate with them in a meaningful way.
The usefulness of unintelligence
Grace writes that "being unable to communicate probably makes a creature more absolutely useless than if it just lacks skills", which is, as proven above, wrong. I have no way of communicating with nematodes, and yet they've proven extremely useful. I didn't even have to tell them to get out of the way - I happily exploited them and waited for them to die once their purpose to me was fulfilled. We're less callous about the lives of honey bees, but in the end, the situation is the same; they - and all other pets - live and die by our hand, and they do so because we can extract some use from them.
We can communicate with some animals in some ways - mostly by encouraging rote learning through positive reinforcement. We also have a decent understanding of their behavior and psychology and can oftentimes determine their mood, feelings and level of happiness. Most commonly, these are animals that do provide us with useful services, such as dogs, horses and other pets not kept for food. Still, there's no trade going on here. Living with friendly humans may be extremely beneficial to the dog, and yet not even the happiest and most loved dog in the world is part of a consensual exchange, because dogs lack the agency to choose another human to trade with. No matter how much we like to pretend otherwise, they are, in the end, our property.
A monkey could do that job!
What about even smarter animals? There's a fairly well known photo of a male bonobo who's not only holding sugar canes, but also sporting an exposed erection. His intentions are clear: he wants to pay for sex. This behavior is common among chimpanzees, and tells us that they have some concept of mutually beneficial trade between consenting individuals. It is, in fact, a very good example of trade: The females are free to choose if they accept the food for sex, or decide to find food elsewhere and sleep with a male they find attractive regardless of how much sugar cane he's holding at a particular moment.
Despite this, there are (to my knowledge) no accounts of actual, mutually beneficial consent-based trade between humans and chimpanzees - sexual or otherwise.
There are a few examples of monkeys performing actually useful human work - but these are curiosities and still lack the element of informed consent. Take Jack the baboon, for example. His amputee owner trained him to help out with his work on the South African railroad. Eventually, Jack was even formally hired by the railway, receiving payment in both money and a bi-weekly beer. But not even this constitutes a trade: Jack didn't know about the many groups of baboons freely roaming the savanna, and he couldn't trade his services to another employer willing to pay him more. Like all other species doing our bidding, he was tricked - however lovingly - into compliance, because he was less intelligent than his master.
Intelligence isn't everything
Mutually beneficial trade is a well known concept to us humans, yet there are plenty of times when trade isn't even considered. This may seem surprising, but it shouldn't be. Apart from land annexation through mechanized warfare, trade is probably one of our most costly and least efficient methods of wealth extraction. It can only exist if a kind of power balance is upheld - externally or internally - between the entities involved, whether they are nations, peoples or individuals. Ironically, it also has the ability to shift the power balance in favor of one of the parties. China, for example, is now effectively undermining the previous unipolar hegemony of the US, and is capable of doing so because of skillfully leveraged trade.
Put another way: Ants, even imaginary ones that we could trade with, are simple creatures. Would we trade with an ant colony if we knew for certain that this trade would make the ants numerous and powerful enough to simply take what they wanted from us with brute force - or would we prefer to somehow keep the power imbalance in our favor?
Hence, even when dealing with our peers, trade isn't always a preferred method of wealth extraction. Unlike what Grace claims, we don't have to enter an "imaginary zero-sum world" to realize this. Malthusian inclinations notwithstanding, reality is usually an excellent source of information. Upon inspecting it, we find that it's rife with pillaging, theft, slavery, trafficking, child labor, colonization, debt bondage, serfdom, indentured servitude, sweatshops with suicide barriers, and other forms of exploitation.
Far from being just historical cruelties, many of these practices are still prevalent in various forms. We have very little of this in the west, which is a good thing, but the wealth and power balance enabling this situation stems mostly from outsourcing the unsavory bits of wealth extraction to other parts of the world. In fact, due to the recent decline of western economies, such practices are once again on the rise even on our home turf.
Conclusion
The world is full of intra- and extraspecies exploitation. Trade - as we'd recognize it - implies consent, and mutually beneficial trade is often an inefficient form of wealth extraction. It requires not only parties of a certain intelligence, but also the ability to uphold an overall power balance. A hypothetical superintelligent AI will be driven by incentives vastly different from biological creatures and is easily capable of outwitting us. Even if it could communicate with us, it is unlikely that we, in such a situation, could make the informed decisions required to constitute a mutually beneficial, consensual trade.
While the AI might not kill us or even be cruel to us, it probably won't trade with us either. But hey - being a dog in a loving family isn't half bad, after all.