[I have written about and against AI before, pointing out that the underlying technology and cultural orientation are nothing new, and arguing that AI is a last desperate attempt for the tech industry to save their bloated techno-salvationist bureaucracies. However, I admit that both of those pieces are very long and hard to follow, especially in the age of AI.
Hence, I have produced the present piece, in which I make four straightforward points about why you shouldn't use AI while also including images from "the Funnies," which I am reliably informed that everyone loves (the year is 1967), for popular appeal.
Before reading the following, however, a clarification is necessary:"AI" in contemporary usage is not a descriptive term for a new technology, but an advertising term for an essentially political program of reallocating massive economic and social resources toward developing and finding applications for and forcing large-scale use of a heterogenous range of pre-existing digital technologies, products, and services.
Hence, the question "are you for or against AI?" is not "do you in the abstract support the further development of digital technology?" but "do you support this political program?": or more immediately "do you think that the present or future achievements of this program in fact and on the whole justify the costs?" To be unable to ask this question, as I pointed out, is to be insane.]
Point 1: AI will never be profitable enough to justify the private investments made in it.
Argument 1.1: No LLM product or data center has as yet generated a profit purely on the basis of consumer and corporate usage, as opposed to investments or circular financing. Nevertheless, AI investment is scheduled to exceed 1 trillion dollars in 2026 alone. The scale differential between zero profit and trillions in investment makes it nearly impossible for the former to ever catch up with the latter for basic mathematical reasons.
Argument 1.2: The intrinsic nature of LLMs makes them processing-heavy and therefore expensive even in carrying out simple tasks. Increases in processing power in chips have already reached a plateau, and the chips themselves require rare materials and can only be made in a few locations. Hence, the technology will continue to be very expensive for the foreseeable future.
Argument 1.3: LLMs are inherently limited to automating tasks that are in and of themselves either superfluous to making a profit or not all that expensive. In itself, human labor is cheap and will likely always be cheaper than LLM usage at scale and for the overwhelming majority of tasks.
Point 2: Given its lack of profitability, the only way AI development can continue on anything approaching its current scale is on the basis of enormous government investment.
Argument 2.1: The current hyper-financialized economy is ultimately structured by profit, if not for corporations at least for investors. Sooner or later, AI investment without returns will lead to a market crash and a pullback on investment into AI.
Argument 2.2: The only institution that regularly invests AI-scale capital without any financial return on investment at all is the government.
Argument 2.3: The use cases for AI include present priorities of the US government in which it regularly invests large amounts of money. These include most prominently military technology and mass surveillance.
Argument 2.4: The US government has been the main driver behind the current AI boom both through favorable regulation and active investment.
Point 3: The only feasible use case for mass-scale long-term government-funded AI is surveillance and murder.
Argument 3.1: Over the past decades, the US government has built a remarkable surveillance and data gathering operation. The main obstacle to the utility of mass surveillance and big data for government purposes is the difficulty of sorting through so much information and arranging it into useful and clear and short and actionable narratives. But this is in fact the one thing that LLMs are consistently good at.
Argument 3.2: Over the past decades, the world has entered a new era of drone warfare, exemplified by the Russia-Ukraine War and the Iran War. Large-scale drone manufacturing requires development in magnet, battery, and chip technologies. The AI boom involves in part targeted development of some of these technologies.
Argument 3.3: One of the main limiters for drone warfare has become the ability to jam remote-controlled drones or even hack into them. This has led to a significant push in the direction of so-called "autonomous" drones with the ability to choose targets and carry out attacks largely on their own, as well as swarming drones operating in sync with each other. Both require sophisticated software.
Argument 3.4: Even non-autonomous drones lead to a vast multiplication of targets; the same is true even for the current military trend towards long-range targeted air bombing with manned aircraft. Selecting targets rapidly enough to keep up with operational requests requires the processing of large amounts of information and the weighting and determining of threats ad hoc and at scale. LLMs are again a technology adept at processing large amounts of information and producing an outcome based on it.
Argument 3.5: The US military and DoD have become a leading investor in AI technology over the past year, with the head of the Department of Energy declaring it a "new Manhattan project."
Point 4: The social, political, and creative effects of mass-scale long-term government-funded AI are and always will be overwhelmingly negative and fall most heavily on the poor and vulnerable.
Argument 4.1: Most if not all military AI uses are unethical in and of themselves, since they involve violations of privacy and the waging of war and taking of life irresponsibly and unjustly.
Argument 4.2: AI Chatbots have already provably caused suicides and driven people mad by forming quasi-social or romantic relationships with them.
Argument 4.3: AI tools have already deprived numerous students of education by allowing them to automate most aspects of schooling.
Argument 4.4: AI tools have already seriously degraded the quality of American visual art, film, television, and even corporate graphic design.
Argument 4.5: By their very nature, LLMs cannot be entirely controlled or predicted, and will always be prone to "hallucination" of various sorts. This has already caused damage in many ways, from spreading false information to hacking networks to causing suicides to committing economic fraud. Militarized AI systems would be able to cause untold destruction.
Argument 4.6: The build-out of data centers has already caused environmental damage, raised energy costs, increased inflation, created noise and other pollution, and failed to produce jobs. Current plans call for an exponential growth of data centers, which would have even more drastic effects.
Ergo: You Should Never Use AI At All For Anything Ever and Actively Do Everything You Can to Prevent the US Government From Supporting It in Any Way. QED.
