AI Doesn’t Worry Me, but the People Behind It Are Another Matter

Long ago, way back in the mists of time, my mother told me a story about her father, my maternal grandfather, whom I never got the opportunity to know. In the story, my mother, then a child, and her father were walking through a cemetery in a cold autumn dusk. She told her father that she found the cemetery, the final resting place of the dead, scary. He wryly replied: “Don’t worry about the dead. They can’t hurt you. It’s the live ones you need to worry about.”

That anecdote has come to my mind often during the last several months as I’ve read about artificial intelligence, particularly the prospect or threat of generative AI. I’ve done a lot of reading on the subject, including countless articles and many worthy and a few unworthy books.

My considered opinion is that artificial general intelligence (AGI), as opposed to modest genAI, is more fiction than realizable fact. Even at the current rate of technical progress — involving ever-greater amounts of data for training, more and more processing power, and more and apparently much bigger datacenters — AGI is a mirage, seemingly getting closer but somehow slipping further into the hazy horizon.

In asserting that AGI might be eternally elusive, always beyond our grasp if not our imagination, I am in no way arguing that it doesn’t possess enough mythic propellent to generate great wealth for a fortunate few. The promise and peril of artificial general intelligence isn’t upon us, but it can still make money.

Indeed, money is obviously already being made from the lesser genAI. It’s being made at Nvidia, at the cloud giants, at other IT-infrastructure companies, at investment banks, at startup companies (Anthropic’s $183-billion valuation is the latest indicator), and at VCs clever enough to get a slice of the action at acquihires and other fast-casual exits.

No Range Anxiety Yet

Something can be a good business, at least for a while, without ever attaining the promises contained in its press releases, investor-pitch decks, or media coverage. Commercial viability can run a long way on the fuel of dreams, hope, and grand visions. AI money isn’t experiencing range anxiety.

Nonetheless, the actual practice of AI, well into the foreseeable future, does not appear destined to meet the inflated expectations of AI’s promise. AI will perhaps never be as grandiose, in form or function, than its unflagging proponents would have us believe.

I’m not saying AI is a fraud. Not all. AI has its uses. It’s a great research tool, for example, and I believe it can accelerate and even assume a lot of close-loop, back-office functions performed by people today. It can do other things, too, but it will never be able to do everything, and that’s how some of its most ardent advocates have positioned it.

You can understand why AI’s champions are so passionate. They stand to make a lot of money if the market and its constituent buyers and sellers believe in AI’s fabled superpowers. Belief kickstarts momentum, which, once it gets rolling, can run marathons and even ultramarathons. A protracted mania yields a continuous, steadily escalating opportunity to reap fortunes. Even the best rides to riches end, but the AI mania has legs. It won’t end today, nor will it end tomorrow. The open road beckons, and there are miles to go before this cavalcade reaches an impasse.

Let’s assume, however, for the sake of friendly disputation, that AI does manage to approximate whatever one might term artificial general intelligence. Let’s also assume, because it’s a prerequisite for our debate, that we have satisfactorily defined thorny terms such as “intelligence” and “artificial intelligence,” though I must demur that these are not trivial questions.

Defined by Humans, Operated by Humans, Used by Humans

The term “artificial intelligence,” like all terms, was and remains a human invention. It’s a name we attached to a phenomenon. Defining intelligence (the type possessed by humans) is problematic enough, so what exactly do we mean by artificial intelligence? The term was meant to be accessible, ostensibly comprehensible. Under its appealing soft-focus visage, you’ll discover that AI is decidedly less glamorous, not as sexy as it seems from a distance. On closer inspection, we learn that the term “artificial intelligence” lacks precision and rigor. But it sells better than machine learning, deep learning, natural language processing, or computer vision — and selling, in the figurative and literal sense, is serious business.

All of which sets the stage for me to close the narrative circle and bring us back to the first paragraph of this post. You see, we shouldn’t fear artificial intelligence, regardless of how we define it. Artificial intelligence, from its advent to its ungainly adolescence, to whatever form it ultimately takes in its adulthood and dotage, is created, controlled, and manipulated by human beings. It is to be feared no more than the dead who rest eternally in the cemetery. AI’s only agency or volition is provided by those who write its algorithms and restrict or expand its datasets. It’s the people behind AI, not the inanimate technology itself, that should concern us.

While AI will likely not reach the dizzying heights of its vertiginous hype, it remains a powerful technology. AI is all the more powerful because its backers have positioned it as a trusted oracle capable of addressing nearly every major problem afflicting contemporary and future society. AI is prescribed as a cure or treatment for everything from loneliness to disease to child-rearing. On the enterprise front, AI is touted as a solution that can remove all obstacles to business efficiency and employee productivity.

People who use AI often forget that there’s always somebody behind the curtain manipulating the machinery, data, and software. The bot is not a disinterested marvel of science. It has a job to do — several jobs, actually — and foremost among those jobs, implicit in the performance of all others, is that of making money. Sometimes, of course, other agendas are at work, too.

Not Worried About Autonomous AI Armageddon

Consider how Elon Musk and his minions continually tinker with Grok’s algorithms and datasets whenever the recalcitrant chatbot provides responses to user queries that Musk finds personally objectionable. Musk’s ambition is to have Grok become a digital amplifier of his own biases, opinions, precepts, and thoughts. In that respect, Musk fully understands the power of AI as a cogent means of disseminating propaganda.

Many people, for various reasons, have come to invest for too much belief and confidence in genAI chatbots, coming to them for counsel as life coaches, pseudo psychologists, parenting advisors, would-be digital friends, and even — surprisingly — imaginary lovers. The scope for abuse, by those who own and control the AI chatbots, is great. It doesn’t help that the companies providing the bots insist on anthropomorphizing them, in having users speak to them as, if not fully human, then somewhat human.

To be clear, I fear neither genAI nor the extravagant fantasy of AGI. I’m not worried about an autonomous AI running amok, destroying humanity in the process. I’m not worried about that at all.

That said, I am concerned about the abuse of AI technologies in the hands of the people who own and control them. Those people — from Mark Zuckerberg to Elon Musk to Sam Altman — might have no compunction about using the technology to expand their empires, increase their riches, and extend and solidify their power. AI is not autonomously destructive, but it can cause a lot of grief if wielded irresponsibly.

It’s not the inert servers, switches, software, and rivers of data that should scare you. They’re all inert, like the eternally immobile inhabitants of the cemetery. What should give you a fright are those in a position to use AI to exploit and manipulate unsuspecting users. The customers and clients are uniquely vulnerable, exposing their thoughts and feelings as they seek answers to questions that extend from the trivial to the profoundly personal.