For at stake in the episode of the Golden Calf is not simply whether the Israelites will live by God’s law, but whether they will succeed in authoring a radically revisionist history of the Exodus. In effectively declaring that they freed themselves, the Israelites erase all debts to God and Moses. And to their slavish minds, freedom means licence; the Israelites’ implicit self-deification authorises the next day’s feasting, drinking, and sexual play. The fitting symbol of the people’s self-exaltation is a mass of gold — a precious metal that, even in Pharaoh’s day, was convertible into the primary objects of appetite, including power, honour, comfort and pleasure.
Today, the World Economic Forum imagines that AI will lead us to a less primitive “utopia”, a 21st-century Promised Land in which people will “spend their time on leisure, creative, and spiritual pursuits”. A safer bet would be drugs and sex robots. Ninety years ago, John Maynard Keynes prophesied, with what looks like eerie accuracy, that machines would make labour obsolete within a century. The prospect filled him with “dread”, because very few people have been educated for leisure.
Judging by the bad behaviour of the wealthy, an “advance guard… spying out the promised land of leisure for the rest of us and pitching their camp there”, Keynes found the outlook “very depressing”. And to those who, freed from labour, looked forward to doing nothing but listening to songs, he replied: “it will only be for those who have to do with the singing that life will be tolerable and how few of us can sing!”
In 2018, an article in Scientific American predicted that advanced AI will “augment our abilities, enhancing our humanness in unprecedented ways”. This Pollyannaish prognosis ignores the fact that all human capacities tend to atrophy in disuse. In particular, AI is inexorably changing the way we think (or don’t). Students now use ChatGPT to do their homework for professors who perhaps rely on it to write their lectures. What makes this absurd scenario amusing is not just the thought of machines talking to machines, but that intellectually lazy people would employ a simulacrum of human intelligence for the sake of mutual deception.
Compared with the natural endowment of human intelligence, the artificial kind is an oxymoron, like “genuine imitation leather”. AI is a mechanical simulation of only one part of intelligence: the capacity of discursive thinking, or the analysis and synthesis of information. Discursive thinking deals with humanly constructed tokens, including numerical and linguistic symbols (or, in the case of AI, digitally encoded data). While human intelligence can compare these tokens with the things they represent, AI cannot because it lacks intuition: the immediate cognition of reality that roots us in the world and directs our energies beyond ourselves and the operations of our own minds. It is intuition, for example, that tells us whether our nearest and dearest are fundamentally worthy of trust. (Needless to say, intuition is fallible, like any other intellectual operation.)
AI has no direct and concrete ties to the actual world, to which it relates only through the medium of binary notation. Self-enclosed in the electronic ether, it dwells nowhere, fears and loves nothing, and has no individual point of view. Does it make sense to grant autonomy and agency to an intelligence that has no natural connection to human needs?
The strength of sophisticated AI is its capacity to sort through massive quantities of data, aggregating and disaggregating discrete bits of information in potentially meaningful ways. This is a promising capability with applications in multiple fields from medicine to transportation. But AI’s productions are artificial regurgitations of material skimmed from vast but shallow pools of digital content and manipulated in ways limited, at least in principle, only by the constraints of programmers.
This can be enormously useful when it comes to detecting patterns of information that would otherwise be invisible to the human eye. Many problems, however, cannot be meaningfully approached by mining Big Data. Asked questions of an ethical or political nature, AI can either refuse to give a definitive response, or it can scour databases for opinions and return what it calculates is the most likely answer. But whether any answer generated in this manner is just or wise can only be a matter of happenstance. This is due, in part, to programming bias, including over or under-weighted data sets. When ChatGPT or Google’s chatbot Bard are asked to evaluate Biden and Trump, for example, their Leftward slant is obvious. (Try asking ChatGPT to write essays comparing each of these presidents to Stalin.)
A more fundamental problem is that machine learning is simply not equipped to sift information according to ill-defined qualitative measures like justice or wisdom. This would be the case even if just or wise perspectives were common on the internet, which they are not. Although no one fully understands how advanced AI works, the old saying applies no less to it than to the simplest computer programmes: “Garbage in, garbage out.”
To the extent that AI remains within the limits of its capabilities, it is because programmers have intentionally constrained its activity. What happens when, for their own all-too-human reasons — the desire for power, honour, and wealth; national pride; or simply the fear of losing their jobs — they remove these constraints? Or when, having well and truly lost the habit of thinking for ourselves, people in general are willing to grant AI authority over matters it is not equipped to handle? Would anyone be surprised if tomorrow someone launched an AI-driven Justice App that promises to settle practical issues of distribution and retribution on the spot? Or if, taking our cue from a well-known software company (NYSE: ORCL), we were someday soon to treat that App — or some other algorithmically-generated distillation of aggregated opinions — as though it were an oracle?
That day is fast approaching, if it is not already here. AI is now substituting for clergy in religious rituals and ceremonies, and Catholics can even utilise a Confession Chatbot. A recent article that sees a use for AI in writing sermons nevertheless observes one limitation on the pastoral employment of machines: “speaking God’s word to a congregation or to an individual requires [personal] relationship.” But AI is incapable of any direct relationship with human beings, including one that is open to the possibility of faith. How could a congregation trust a religious leader, much less a God, that cannot reciprocate this trust?
The Israelite experiment with idolatry ended in disaster. After the episode of the Golden Calf, Moses ordered the Levites to take their swords and purge the camp of wrongdoers. “Slay every man his brother,” he commanded, “and every man his companion, and every man his neighbour.” Should we not expect a similarly bloody consummation when, heeding the utterly irresponsible voice of AI as though it were the Word of God, we once again reach peak idolatry?
Join the discussion
Join like minded readers that support our journalism by becoming a paid subscriber
To join the discussion in the comments, become a paid subscriber.
Join like minded readers that support our journalism, read unlimited articles and enjoy other subscriber-only benefits.
Subscribe