You’ve probably tried it: the new AI tool that writes more articulate essays than you did after a late-night rager at university.
The app, known as ChatGPT, is fuelling excitement, speculation and investment. Artificial intelligence companies have moved quickly to take advantage, and many are turning the hype dial to maximum.
“We will build the descendants of humanity and launch them off to colonize the universe,” OpenAI CEO Sam Altman told the Wall Street Journal recently.
“We are trying to build one intelligence that is smarter and more capable than humans in every way…if we’re successful, I think it will be the most significant technological transformation in human history.”
Altman’s proclamations are a vintage example of what critics call “AI Hype,” a phenomenon characterized by drawing “a straight line from an existing accomplishment to a limitless future.”
Hype or not, Canada’s governments have invested heavily in AI, and the sector is booming.
As of March 2021, an estimated $1 billion in government funding has flowed to AI research, academic programs and business ventures. The formation of national “AI institutes” in Toronto, Montreal, and Edmonton has led to over $3 billion in private investment, hundreds of new AI firms, and tens of thousands of specialized AI jobs.
Canada’s AI startups are developing automated customer service bots, robots for a variety of manufacturing and service roles, medical diagnosis through machine learning, and as their boosters tout, “insights” to “reduce workforce costs.”
And as billions in speculative capital flee deflating bubbles like cryptocurrencies, many investors are susceptible to a pitch for a new technology that will transform everything.
While hype may be overshooting the technology’s actual potential, institutional actors are responding with oceans of cash and are expecting big returns. Making apps like ChatGPT and DALL-E accessible is a central driver of the current wave of coverage and speculation, and is costing OpenAI millions per month.
The AI hype is fuelling a dizzying array of profound disruptions to workplaces, employer-worker relationships, and business models. Even if the most overblown promises fall short, what lies ahead are drastic shifts in power that are bound to favour corporations and the ultra-wealthy.
Automating the workplace
Riffing on Marx, some critics have characterized AI as “undead” labour—a zombified copy of actual labour, codified in the machine and living on in the brains of human workers.
The enthusiasm for this form of labour relations is accelerating.
Last month, Canada’s federal government announced an additional $30 million investment in robotic automation.
Automation is one of the key drivers of AI investment. For any given application, developing and training a robot AI is likely to be much more expensive than hiring a human in the short term. However, capital outlays can pay off over time with reduced overhead.
In 2015, a McKinsey Digital analysis concluded that “very few occupations will be automated in their entirety.” However, “certain activities are more likely to be automated…and jobs performed by people to be redefined.”
AI can perform relatively simple, repetitive tasks, albeit at a high computational cost. But to achieve proficiency, the software has to be “trained”—given very large, high-quality datasets or constant feedback from human operators.
Machine learning software created to analyze CT scans, for example, is reportedly able to complete a task in 20 seconds that would take a radiologist 20 minutes. But to have trustworthy results, each task requires extensive—if not permanent—human supervision.
The result: dwindling ranks of human employees, as AI takes on repetitive and simple—but increasingly complex—tasks.
Structuring work to serve the bot
But as machine learning becomes more central to the workplace, human work is being restructured to support, correct and extend the capabilities of the software.
This includes the untold thousands of offshore workers who are paid $2 per hour to moderate graphic, violent or offensive content or to civilize apps like ChatGPT—preventing it from being racist, for example.
But the category increasingly also includes skilled labour in the Global North.
A recent essay in N+1 describes how grad students and people with PhDs in the United States were recruited to be “human fallback” for a customer service bot known as “Brenda.” When Brenda would fail to understand a human’s chat queries, the employees—many with degrees in English or performance studies—would step in.
By working for an AI company, one does not just sell one’s labour, but the imprinted pattern of that labour, which can be used in perpetuity without the worker.
“The operators got paid better than they would as adjuncts,” author Laura Preston writes, “and Brenda became more likable, more convincing, more humane.”
But such alliances are temporary at best. By design, machine learning software will increase the scope of the activities it can do without human intervention, learning how to accomplish increasingly sophisticated tasks, eroding the amount of “human fallback” required.
By working for an AI company, one does not just sell one’s labour, but the imprinted pattern of that labour, which can be used in perpetuity without the worker.
Fine-tuning manipulation
“Of all a company’s functions, marketing has perhaps the most to gain from artificial intelligence,” an article in Harvard Business Review suggests.
Social media apps are most notable for using large datasets to constantly make their feeds more addictive, manipulating the weaknesses of the human brain. Machine learning will make these techniques increasingly sophisticated and specific by leveraging both large datasets and information gathered on individuals.
Early applications of machine learning in marketing are used to reduce friction, connecting people to products or services they want to buy. Others rapidly generate customized marketing messages to hook various audiences.
But less-trumpeted marketing applications could involve manipulation, misdirection or underhanded persuasion.
Facebook’s in-house AI, after all, was able to predict mental illness diagnoses 18 months before individuals received them. The leverage that large datasets create for persuasion and manipulation is massive, and can extend far beyond the self-understanding of those it analyzes.
And whether they’re using their own tools to optimize marketing or not, OpenAI has proven effective at creating global conversation with its public-facing applications.
Controlling the investment firehose
The repetition of the narrative of imminent machine intelligence elevates what is undeniably a remarkable technology into some kind of culmination of the history of technology, designed to awe casual observers—and get investors to open their wallets a little wider for a turbocharged Next Big Thing.
The same hype method has been deployed repeatedly by Silicon Valley; it involves a sophisticated infrastructure built around the tech press and its symbiotic relationship with venture capital.
Uber claimed to be changing the future of transportation—it wasn’t, but it generated enough capital to disrupt the livelihoods of thousands of drivers with the illusion of creating something better. Cryptocurrency investors claimed to be creating an alternative to fiat currency—but used the hype cycle to bail out rich speculators by getting working class people to buy in just before the crash.
The function of the hype machine is to choose where investment goes and where it doesn’t. The more money goes to Uber, the more lobbyists will argue for de facto privatization of public transit.
The function of AI hype is to project our public, collective hopes onto speculative—and currently nonexistent—technology over proven solutions.
Altman has vague-but-reassuring promises to accompany his hype claims. “We will discover new jobs,” he writes of the disruptive effects of AI. “[W]e always do after a technological revolution–and because of the abundance on the other side, we will have incredible freedom to be creative about what they are.”
By several measures, Canada is a leading contributor to these disruptive effects. The country is reported to lead G7 countries and China in the number of AI patents per capita, and Toronto is said to have the densest cluster of AI startups in the world.
A new bludgeon against the working class
Long before Altman and ChatGPT took centre stage, corporate actors were using the threat of AI to intimidate workers.
During a push for higher wages for fast food workers, a former McDonalds CEO threatened that automation would replace uppity workers.
“Automation is both a reality and an ideology,” Astra Taylor wrote in Logic in 2018, “and thus also a weapon wielded against poor and working people who have the audacity to demand better treatment, or just the right to subsist.”
Altman’s plan—based, of course, on his wild claims about the future of AI—is to create a small dividend from the anticipated AI windfall that those replaced by automation can live on.
AI derives its power—to automate, manipulate, predict, and innovate—from collecting vast quantities of data, from concentrating massive computing power, and surrounding it with human trainers.
Given enough of each of those, some believe it could become self-teaching, drastically accelerating the upward trajectory of its sophistication.
In short, it empowers those with the largest concentrations of wealth and capital.
And despite spending billions to train a generation of AI experts and conduct the basic research that fuels the industry, governments have so far given away their ability to determine how the technology is used.
What Altman and his ilk leave unsaid is that if even part of OpenAI’s narrative comes to pass, we’ll be depending on the magnanimity of the billionaire class and elite financial institutions to dole out enough to meet the basic needs of the newly disempowered and unemployed.
Neither tea leaves nor AI prediction engines are required to see if that scenario will come to pass. Days before it announced an unprecedented $10 billion investment in OpenAI, Microsoft laid off 10,000 workers.
If AI is eventually used for the benefit of humanity, it will be because someone wrests control of it from corporate actors and billionaires.
