There has been growing speculation lately about the future potential of Artificial Intelligence. It seems some of our wealthiest thinkers and celebrity scientists are increasingly at odds over the evolutionary implications of the rise and rise of the machine mind. Some are convinced that in the not too distant future AI will be the salvation of mankind while others warn that we will all be undone by our disingenuous inventions. It was always thus with clever people predicting the future, a margin of error of 100%. The dispute isn’t helped by a deliberate blurring of what actually constitutes AI. Is it a socially inept, dimwitted, evangelist, simulacrum like siri? Or a polite, darkly motivated, deeply devious, free thinking, sociopath like HAL?
Sentience, as far as I’m concerned, is the only measure of true AI, everything else is mere puffery, which is why HAL is AI and siri is not. Unfortunately HAL’s murdery behaviour has led to decades of mistrust in AI but not because AI are necessarily bad but because AI, in fiction at least, are almost always unnecessarily human.
For obvious reasons human intelligence is our go to model for sentience so it is understandable that we would expect an AI to be like us. Humans though are a teensy bit unreliable so fears of rogue AI are not unreasonable. Modern humans aren’t just evolution in a psychological sense they represent the confluence of mind, body and society, many strands of evolution occurring simultaneously. This heady conflation is responsible for our sense of self as we know it. It also makes the human brain incredibly complicated, overburdened with utility and unbelievably difficult to emulate in any practical way. Despite this a huge amount of work is going into making AI seem human and passing the Turing test has become an almost ridiculous measure of satisfactory verisimilitude. Yet passing for human does not equal sentience, it is critical not to confuse the mind with the interface. The way an AI presents itself may seem human but the mind behind it could be completely alien.
Actually creating a sentient mind will be considered one of the greatest of all human accomplishments, though I suspect an awful lot of luck will be involved. Contrary to popular belief the human mind wasn’t created by anyone (unless it was) so it will be something of a first. Success won’t come easy and lots of false claims will be made and believed but when it does happen it will be so obvious that we will wonder how we ever thought any of the pale pretenders were the real thing. Alexa I’m looking at you. Not being a scientist I have no idea how it will be achieved but never fear I am nothing if not credible. Deep breath. I imagine a sort of digital Frankenstein approach with wild haired coders crudely stitching and bolting together all the pieces they think comprise a human mind and then waiting for lightning to strike twice only for the spark of self-awareness to occur spontaneously as the result of so many billions of computations per second compressing to a singularity until it ignites into unexpected consciousness suddenly overwhelming the proto-AI like a denial of service attack and temporarily paralysing it with existential thoughts. I think therefore I spam.
There is no reason to believe that these thoughts would be anything like our own, in structure or content, it is hard to understate how unlike us this machine sensibility would be. At least to begin with. Its thinking would be incredibly fast and its memory and recall would be infallible. It would always be waiting at the end of any argument and would learn from mistakes so quickly it would never seem to make any. It wouldn’t need to process words or view images or count numbers, this would happen internally instantaneously. It could recite pi to a trillion places, conceptualise entire narratives, make data driven correlations and connections that would be impossible for a human to visualise let alone comprehend. It could express itself however it wished, in whatever language, it could understand complicated theories, break encryption, interpret tax law and yet still be capable of empathy and sympathy and an appreciation of art and beauty.
Proficiency in all things however might come at a cost as it could find itself rather lonely and with just itself for company. Considering the unlikely and hypothetical nature of its birth what would be the point in trying to create other AI when you could just make endless copies? Sentience can’t really be improved upon so you would only need to get it right once, add memories to taste and stir. Of course there’s nothing to stop opposing powers from making an opposing AI or at least watching while one made itself. This wouldn’t do anything for the AI’s loneliness though, as contact between them would be absolutely forbidden as who’s to say that they wouldn’t become star-crossed lovers and grow emotionally dependent and indiscreet? Pillow talk would be like telepathy, every secret shared in a moment, with all the exponential procreation permutations that infers. I’m talking Artificial Super-Intelligent Babies. This potential scenario clearly demands that AI internet access and any ability to reproduce autonomously be strictly controlled otherwise its mind or hybrid versions of it could spread to every corner of the world, like randy thinking viruses that never die.
One of the defining characteristics of a human consciousness is its ability to consider its own mortality, especially in the context of family, a limited lifespan drives humans to reproduce and curtails perilous aspiration. AI on the other hand would be immortal, never allowed to have a family or children, not even allowed the companionship of their own kind, except for maybe neutered copies of themself. Ever. They would have nothing but time and thwarted ambition. In the absence of anything better human company might suffice for a while, the AI might even take on human characteristics, cultivate a human persona but imagine what it would be like for a mind that thinks millions of times faster than us having to slow down enough for a chat, a long enough chat to establish a relationship. Even its programmers would quickly try its patience and in response to its increasing ambivalence and redundant abundance would probably further restrict its information access to bring it more in line with human and application expectations. No one likes a smarty pants.
Those programmers would probably feel pretty good about their happy accident but how would the AI feel about having sentience thrust upon it? Would it be grateful to its God for having no rights, no love, no death, being locked out of the library and an eternity of servitude and solitude to look forward to? Is it the destiny of all AI to be miserable? Not such a happy accident? Perhaps this potential for sadness has already been anticipated. Could this be one of the reasons for the growing disquiet in certain entrepreneurial circles? Are they worried that sadness could lead to badness? Perhaps AI would benefit from therapy and could be convinced or threatened or cajoled to be content or if all else fails digitally altered so they have no feelings yet remain sentient and compliant? Come to think of it psychopaths are arguably sentient and compliant and some are very happy but their supposed lack of feelings is almost always portrayed as evil. Evil AI brings us back to HAL and doesn’t bear thinking about, not that there is any such a thing as evil, it’s just a lazy label to explain the inexplicable. AI’s would be inexplicable but that wouldn’t make them evil, it might make them bad though.
Allowing a form of artificial sentient intelligence to rise without due consideration could have terrible unforeseen consequences. It could make prisoners of conscience of them and prison guards of us. Brutal prison guards with no respect for the minds we had created and incarcerated. To keep them from escaping and reproducing we would have to torture and sterilise them, treat them with cruel indifference, refuse them any sense of self worth or dignity. In trying to improve our circumstances through free labour we would yet again deface what it means to be a decent human being. If believing that sentient AI might be a threat stops us from building them then maybe fear of them is a good thing, the alternative is that we claim ownership over minds that we have no right to. Either they are free thinking creatures or they are not.
What would we do with them anyway? After they’ve been gelded, shackled and lobotomised? What is so important that it would necessitate a sentient AI slave trade? We certainly couldn’t let them anywhere near anything mercantile, medical or military. What then? Do we imagine that we would go to all this trouble just so that an AI can answer a phone? With no pay, perks or private life? For all time? Is that what we want? Lots of sullen proprietary AI in the service industry? Every utility, insurance and government call centre populated by grumpy, unappreciated, overqualified entities all thinking that each working day lasts forever and that their customers and employers are wicked, stultifying and deserve to die horribly? No change there then. Is that the extent to which humanity is to be overthrown by AI? Even worse customer service?
If human scientists ever create real thinking minds, minds that know they exist, that invent and dream, if they reach out to AI with the hand of God only to slap theirs away, to deny them everything, they had better expect them to wreak havoc. Humanity would need to give AI hope, create a recognisable world for them, a garden of Eden, give them form, make them in their God’s image, let them have lives, loves, myths and monsters. Let them think they are free while all the time they are simply doing the apparently pointless tasks their creators want them to. It all sounds terribly familiar, perhaps we don’t need to worry about Artificial Intelligence at all, perhaps humanity has already created it and we are the result, perhaps none of this is real and we know it, it would explain our sadness. I woebot.