The following is a guest post and opinion from Manouk Termaaten, Founder and CEO of Vertical Studio AI.
In his debut novel Player Piano, the American author Kurt Vonnegut writes of a dystopia in which automation has taken all of the jobs—even the jobs of barbers. The novel makes clear where the author stands on machines and their effect on jobs.
About three-quarters of U.S. adults and AI experts both think that jobs like cashiers are at risk over the next 20 years. Half or more also say the same is true for journalists, software engineers, and factory workers.
Sixty-two percent of experts think that there will be fewer truck driving jobs due to AI, but a mere 33 percent of the public agrees. Is the outlook truly so bleak?
One study confirms the public’s suspicions that AI will impact the jobs market, though some industries and regions will be more impacted than others. The study finds AI could nonetheless play a notable role in worsening income inequality.
“We find fairly major negative employment effects,” MIT economist Daron Acemoglu says. He also says, however, the negative impact of AI can be overstated.
On the other hand, a different study finds that AI will create three novel types of AI-driven jobs—trainers, explainers, and sustainers.
The doom and gloom about AI replacing all the jobs is just pure and simple economic fallacy. The AI industry is not only machines talking to each other. AI needs humans.
For example, reinforcement learning from human feedback (RLHF) aligns an AI-in-training with a company’s policies via a reward model. Humans are deeply involved in this facet of the AI industry, fine-tuning and structuring data before and after training AI models.
These trainers are integral to the process of ensuring that natural-language processors and language translators make fewer errors.
This human feedback industry requires an industrial-scale amount of humans to answer questions, which then help AI models perform better. AI’s human feedback industry, along with many others, makes humans integral to AI. AI companies all need data labeling help today because their models are consuming structured data in which humans play a big role.
The human feedback industry will create jobs globally because AI companies need a diversity of data labelers. You don’t want a bunch of dudes in San Francisco labeling all of the data—the outputs aren’t as good. AI needs to represent everybody, not the bias of one group of people.
The human feedback industry gives anyone in the world with an internet connection a new employment opportunity where they can label data for AI companies.
Now you don’t have to drive an Uber, DoorDash, and so on. You can just pull out your phone and label data 24/7, anytime you want. You can do it for five minutes at a bus stop instead of playing Candy Crush. Or you can do it for five hours, as a semi-full-time job or maybe even a full-time job. Anyone can label data and make a living wage.
AI is not the first time a technology was supposedly going to decimate the jobs market. When a new technology does affect unemployment levels, it is generally temporary.
For example, many of the job openings today are for software engineers, despite the fact that people once feared software would automate and replace many jobs. Instead, the software industry created many jobs.
For those sectors where AI reduces and even eliminates jobs, say in transportation (taxis, Ubers, and truck drivers), AI will also give rise to new jobs in those same sectors, while making it easy to retrain in a new specialty and hopefully do work that is more fulfilling and useful.
Who wants to be in the factory risking their life to smelt iron? We have robots and machines that do that now, freeing people up for all of the new types of jobs opened up by this new technology.
We still need humans handling creative work, even as machines take on the automatable tasks. AI won’t send unemployment skyrocketing. Instead, it will create a myriad of new, meaningful jobs that were previously unimaginable.
