When researching intentionality in relation to artificial intellgence (or writing about it here), I analyze my findings or hypotheses from the lense of our human perception and reality. The effect of seeking and comparing qualities under this premisse, has the implication that we sometimes interpret features into AI or compare them with something that seems similar, but isn’t (Nagel/Searle). However, even before any anthropomorphizing can happen, we seem to introduce another bias into the eqation, which is that AI is on a way to become more like humans or to potentially surpass us (AGI). What would happen, if we exclude this bias from our assumption? What if we were to invert this argument and consider, if humans will become more similar to AI than the other way around? When toying with the idea that humans will sometime have the interfaces and means to edit their own neurological processes or personality, I always arrive at the conclusion, that we will optimize ourselves towards a being that is very similar to what we currently perceive as AI. In what follows I will try to explain how I arrive at this conclusion and what indicators I see.

Imagine the following for a moment: While working on a project, something doesn’t work out the way I expected it to, which frustrates me and angers me so much, that I can’t concentrate anymore. Luckily, I can use an app that is directly connected towards my neural interface and dial down the anger attribute, while at the same time overclocking my focus for the next half an hour. If I then were to present my project, but the reception turns out to be bad, I can simply couter that by bumping up my happyness metric. Why not afterall? What benefit lies in being upset over something I can not change anymore? It would be more helpful to focus on what I can indeed change, which is getting back to work and bettering myself. Stressed out? Adjust it. Can’t sleep? There’s an app for that. Heartbroken? Mute it.

If you read an essay or two of mine, you hopefully come to the conclusion that I try to avoid subjectivity or matters that can’t be quantified or measured as best as possible (except for my human lense), simply because in practice, this will water down any academic debate into a matter of opinion. Hence, to discuss the matter in a meaningfull way, as objective as one can hope, it seems to make the most sense to me to analyze popular arguments brought up in disucssions of incompatibilism, reverse them and see where this may lead us.

I assume:

• AI processes information, it fundamentally doesn’t understand it. Furthermore, it has no independent intent on its own. It acts, because it is told, not because it wants to do something.
• In only being fed digital information (for example image + text descriptions), AI lacks the peripherals to experience and potentially build an understanding of the world it perceives.
• The AI doesn’t have a self, as in an own personality. It can be argued that it has no concioueness too, though impossible to prove.

So, what would be, if AI was never meant to become more human-like, instead we are evolving as a species to become more similar to AI? What, if we were to reverse each claim and see, if we could approach such a state for ourselves?

The question of understanding and processing

Imagine we’re in an office and we are handed a task. The content of the task doesn’t make any sense, as if written in language we don’t speak or because we can’t be bothered to look into it. However, we have a very eager intern, who’ll happily take over for us. The customer reports that their expectations are satisfied and the task is closed. What we’re talking about isn’t the chinese room experiment, but me paraphrasing a recent paper called “Thinking-Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender”. Well, actually it is the chinese room somewhat. The researchers found out that people tend to increasingly consult AI while reasoning with some interesting implications.

Everyone who ever programmed something knows this initial spark of confusion wondering “How could I solve this?”, “How do I obtain the variable?” or “Where could the bug possibly be now?”. Personally, I often feel that this is a moment of shortcoming; I fail to understand the framework or programming language as in I’m not good enought yet. However, solving these problems is the very heart of what differenciates coding (akin to an author knowing grammar and spelling) to programming (creating a new book). At best, I don’t throw the first solution that comes to mind at the problem, instead, I try to understand and reason what options I have and if there could be potential better candidates for my application. However, as outlined in the study, I could also skip the understanding part completely, trust that the AI has a better soluton at hand and simply implement the process. I wonder, how much different that is from an AI simply processing information? The study specifically speaks of “cognitive surrender”, meaning that the moment friction (or developing an understanding / researching) arrises, people tend to hand over to AI.

The biggest difference in the way we therefore process things compared to AI, we could argue, is probably the inherent intentionallity that we have. Naturally, I don’t programm for programmings sake, but to solve a problem. Perhaps to earn money, status or to make something cool. This interconnection seems to be the bastion that holds the difference to a machine that simply enacts orders it receives.

But how does that hold up in the light of us being able to edit ourselves? As mentioned, in my thought experiment, instead of acting in the goal of obtaining money, expanding my portfolio or to stop my boredom, I could simply dial these attributes down, if I were given full writing access towards my neorological processes. Perhaps, with this surrounding nosie out of the way, I could set myself to nobler causes, which I simply weren’t capable of before, like curing a deadly illness or solving a math theorem. Yet, if I were not to change my toolset of having an AI doing the heavy lifting for me, only managing a surface level understanding that is just enough to hopefully verify, if my work is correct or not, how is that not comparable to processing? Idealogically speaking, without going much into detail as this isn’t my field of expertise, dogmatic thinking, (if I were to borrow populistic scientific terminology: brainwashing) has shown to “overcome” reasoning and holistic understanding. What makes us so individual and resistant to becoming a mere drone? If our mind is to be understood like a muscle that can be trained, it can also loose its strength if retired. And this oversees only the implications of manipulating the “environment variables” such as motivation, anger, concentration. How would I react, if I were given the choice, to do my work completely on autopilot, while daydreaming? Admittedly, this sounds very sci-fi-ish, still, when I exchange the programming task with my way to the super market, I could argue that I do 90% of my shopping on autopilot anyhow. Perhaps I actually have to think through what I want to eat for dinner, but I don’t mindfully, intentnionally, step one foot in front of the other. It therefore could be argued, that when intent is to be understood as a gradient, at least the ability to completely switch it of, is not out of the question, given the right hardware and motivation.

The question of personality and self

While a computer always processes something excactly the same, human outcomes differ a lot. We are subjective, have our ups and downs, edges and oddities; some are more neurodivergent than others. Personallity or the self, has always been a strange subject for me, because beyond the implications of Platos cave allegory, I find it hard to justify that it exists at all. Naturally, I can’t prove that I have a personallity, it is always implied, but I can’t measure, smell or taste it. To say, that it doesn’t exist because of that is a bit shortsighted admittedly, especially because “personality” and “self” have a variety of definitions. In a similar way, my personality itself doesn’t have one definition either. If we were to ask my friends to describe me compared to my landlord, the definitions would vastly differ, perhaps overlap here and there, but in the end, never equal. Hence, the way I see myself or the self I define mysself as, simply doesn’t exist except for me. While, as pretty much always in philosophical discourse, this doesn’t provide us with a definitive answer, it already casts heavy shadows over the claim, that personality, or the self, is something that sets us appart from computers, because the concept itself isn’t understood.

Perhaps then, we could argue, personality isn’t something static, but always in motion; a complex network out of things that are always subject to change. Afterall, I as a person, am vastly different than my 12 year old self. While this constant change can also be said about personality traits we might interpret into an LLM, it would process the same input exactly the same, given that every node is exactly triggered in the same sequence as before, which is something that can’t be said about us. A student who trained on a completely identical equation during studies, isn’t guaranteed to replicate the same result in a test environment, or a year from then. Something that seems to also be unique to us is that we may reject something, because it isn’t to our subjective liking. Unthinkable for a computer and circling back to intent.

But then again, what would happen, if I were at a deadend job and we’d have to suffer eight hours a day through a shift, simply to make ends meet, if we could edit ourselves? If I’m already there for eight hours I might as well be having the time of my life. Perhaps I’ll even get a raise out of it, because it’s such a pleasure to work with me? What good would it be, to sit there every day in misery? Maybe, it is a bit farfetched that people were to abandon their intrinsic reaction the moment they are given the choice to do so, however, in the light of people prefering immediate satisfaction through completion of tasks over a longer and often harder autonomous studying process, it doesn’t seem completely unlikely. Additionally, this isn’t a trait that would be unique and novel to neural interfaces. How often did we want to say something unprofessional but opted for a diplomatic approach? Did we ever withhold information while bargaining over something or did we straight up say that we’re willing to pay any price, because we want something that much? Ever lied to not hurt someone? We override our emotional response, our personal wishes, all the time to fit in better into society. We’re already editing ourselves, just rather inefficiently compared to the implications that an interface were to give us. What makes us so sure that, if we gain the abillity to edit outselves, we wouldn’t strive for the most comfortable life, if easily accessibly, even if in turn, this makes us loose all uniqueness and therefore becoming more similar to a machine? By that I mean that we gain the abillity to reproduce exactly the same result similar to a method that works in a binary capacity.

The question of reality

With our senses we perceive the world, or what we think reality is, in different subjective shades and connect our experience richer than a computer could derive from digitally fed data, since it lacks the peripherals for senses we have. Similarly, I can only try to imagine how a bat experiences the world through echolocation, but never truly understand the difference of perception an open field or small room makes in sensations and subjective experiences I derive from that just as I struggle to explain the concept of blue to a blind person.

At least for now. In the world of biohacking, a very common modification is to implant magnets into our hand. After a while, our brain adapts to this new sensation, giving us an additional sense for perceiving magnetic fields and electric devices. While it might take a while for echolocation to become widely available consumer grade upgrades, this discussion point is the one where I think both sides could approach eachother rather than humans becoming more similar to AI. Computer vision supported by neural networks is an old hat, such is sound, temperature and spatial awareness of course. While on the other hand, smart glasses/AR could be led as example of supplying a stream of digitized information directly to a person (even though vastly underperformant compared to computers processing such data). In that, I fail to see why a neural interface would not be able to have an interpreter layer someday that allows me to receive digital data and perceive it with my brain. If mapped correctly, an image file of an apple could be send to my brains visual areals, making me dream up an exact image of such. Similarly, I don’t see a reason why it should always remain impossible to revert this process, making an interpreter for human thoughts that translates into the realm of binary processing. While Searle raised concerns that these two worlds cannot be seen as equal, similarly to a wooden combustion engine not being able to perform exactly to one made out of metal, this argument offers itself as a simple answer that rules out that a new paradigma of mental processes is impossible to adapt for our brains. If breaching our own perspective would at all be impossible, empathy wouldn’t exist and while not the answer to everything, Wittgensteins attempt to convey philosophy through logic and symbols wouldn’t exist either. Perhaps a fuzzier argument, because it differs from the perception of our peripherals a lot, but culture and language could also be considered a factor that shows fragments of different mental models. Hence, I don’t see the argument as exhausted, that it is the computer that becomes more like us, rather than us adapting to the thought structure of a computer, simply because it seems to be an incompatible model at first. While a magic interpreter layer that acts as a translator between the two systems seems like a solution, a new language that would offer itself as common denominator seems also very plausible, given that both parties approach each other at the same time.

In review

If we step back from this exploration, as often with AI and philosophy, it feels like a walk through a colorful garden of ideas that is glazed with unsurmountable flowers of things we do not understand yet. Because of that I’m used to the feeling that this essay feels rather superficial, echoing copies and variations of ideas of other people. Still, perhaps a reader might catch an idea or two or connection I made, that wasn’t present on their mental horizon (personally I’m just glad I breached the border of vague idea to structured post). Therefore the only conclusion I can offer is that our technological race surrounding AI isn’t a one way street perhaps. Will AI become like us? Maybe? Do we evolve to become more similar to the machines we build? Potentially? Or is it that we step towards each other and find a common denominator becoming an algamation of both in the process? The only thing I know for sure, is that I don’t know.