Exploring the likely impacts of AI on human labor
As artificial intelligence continues to make significant progress, many people are beginning to wonder about their own job security. We’ve considered this a fair amount, and our simplest answer is: virtually every job in today’s economy has a significant potential to be automated within the coming decades.
However, there is a great degree of uncertainty involved here, and it’s difficult to predict what jobs will be automated, what new jobs will emerge, and what the net effect on employment rates will be. Still, we think there is good reason to expect that there are virtually no jobs where automation doesn’t pose a threat, that this technology may be dissimilar from past paradigm shifts in terms of the net impact on employment, and that there is much more that companies and governments need to be doing today to prepare for the impact of AI on workers.
What all can AI do?
Just this month, Jensen Huang (the CEO of Nvidia) said on stage at the World Government Summit that children should think twice about learning to code given the potential for AI to automate software development.
But it isn’t just in the field of software development where AI seems poised to soon outperform human workers. We’re already seeing AI models that can exceed human performance in copywriting, data entry, and simple research — especially when you consider the relative price of employing humans and the search and training costs involved.
To many experts’ surprise, AI is also capable of creative tasks like image and video generation, as well as writing creative fiction and poetry, which threatens to displace artists and creatives. There was once a time when people expected that the first jobs that would be automated were menial physical labor — simple tasks like sweeping floors which seemed straightforward enough to build machines for — and the final tasks to be automated (if ever) were creative tasks like painting and writing. However, recent history has clearly shown us that, if anything, the opposite may be true.
In fact, many domains that once seemed fundamentally human now — in principle — seem fully automatable. Even scientific and academic research that aims to expand the boundaries of human knowledge may be done by artificial intelligence, with the first existence proof of this being the remarkable progress made by AlphaFold at solving the previously unworkable “protein folding problem.”
As far as we can tell, there are no boundaries to what AI will and will not be able to do. However, the order in which certain jobs become automatable remains unclear, and future developments may surprise us.
What can the past tell us about AI’s impact on jobs?
In the past, new technologies haven’t had the (often forecasted effect) of reducing total employment, but instead have opened up new jobs that we hadn’t previously imagined. Some people will point to this historical trend as a way of assuaging concerns that the development of AI technology will lead to significant drops in employment in the coming years.
The industrial revolution saw a major shift from an agrarian, manual labor-based economy to one centered around machinery and manufacturing. Many predicted that the rise of machines would lead to widespread unemployment as human workers were replaced. However, over time new types of jobs were created – factory workers, machinists, mechanics, engineers. The automobile industry alone created millions of jobs for people in roles that didn’t previously exist. Though some specific trades declined, overall employment grew along with economic expansion.
More recently, the advent of personal computing and the internet during the digital revolution was expected to greatly reduce the need for certain office jobs like secretaries, file clerks, even accountants. But again, new technologies ended up creating new roles – software developers, IT specialists, web designers, social media managers. Platforms like eBay and Amazon marketplace enabled millions of small business entrepreneurs.
So, does history teach us that we shouldn’t fret about job loss caused by AI? We don’t think the answer is so simple.
There are important differences between past technologies and the development of artificial intelligence. The biggest difference is that this is the first technology that isn’t just replacing a particular tool or occupation, but rather replacing the agents that use tools and engage in occupations in the first place. It isn’t just a new method for intelligent creatures to build new things or provide new services, but it’s a fundamentally new intelligence in and of itself, capable of building new things and providing new services on its own.
Past innovations enhanced and extended human capabilities, allowing us to be more productive and take on increasingly complex work. But artificial intelligence has the potential to replicate and even surpass human intelligence. As AI systems become more advanced, they may have the ability to improve themselves and operate independently without human oversight. This introduces not just the displacement of certain jobs, but the replacement of human agency and cognition as the primary driver of economic productivity. An AI agent doesn’t just do the tasks humans instruct it to – it can set its own goals, teach itself new skills, and find novel solutions without our input. So while previous technologies expanded the set of roles for humans in the economy, advanced AI could theoretically make many roles obsolete as the locus of intelligence shifts from biological to digital. This unprecedented capability is what makes the societal impact of AI so uncertain.
How do people expect AI to impact jobs in the near future?
Among experts, there is wide disagreement about how exactly AI will impact employment, but many AI developers, economists, and forecasters expect to see at least some major labor disruptions emerge from the upcoming developments in artificial intelligence — if not mass job displacement.
OpenAI CEO Sam Altman has said that “Jobs are definitely going to go away, full stop,” but has also expressed the opinion that new forms of employment will emerge. Economic research from Goldman Sachs has predicted that 300 million jobs could be at risk due to the advent of artificial intelligence. And research conducted by ResumeBuilder has found that 37% of companies already report replacing workers due to AI, and 44% expect future layoffs from the technology.
But what are the odds that all jobs will be automated? The current goal of AI labs like OpenAI is the development of “AGI” or artificial general intelligence. There are multiple conflicting definitions for this term, but one of the most common (and the one used by OpenAI itself) is “highly autonomous systems that outperform humans at most economically valuable work.” It should go without saying that this is the sort of system which, in the right circumstances, has the potential to displace almost all jobs in today’s economy.
Forecasters currently expect that models qualifying as “AGI,” with broadly human-like capabilities and expertise in a wide range of fields, as well as the capacity to control robotics with high degrees of precision, may arrive this decade (with a median prediction of 2031). Despite the high degrees of uncertainty and methodological challenges in making predictions like this, it should be enough to make us take seriously the possibility of mass job loss in the coming years.
What can be done?
At The Midas Project, we are working to make sure that AI benefits everybody, and a crucial component of that will be ensuring that the economic benefits created by AI are widely distributed.
This doesn’t necessarily mean preventing AI automation, but it does mean making sure that nobody’s livelihood is threatened by AI automation. At a minimum, companies should work to ensure that AI automation is used by employees to improve their productivity rather than displacing employees whenever this arrangement can be beneficial. However, if AI capabilities develop quickly enough such that humans are only slowing things down, companies should ensure they are offering ample employee protections in the event of termination, including unemployment insurance.
But unfortunately, this won’t be enough. If AI continues to threaten human labor, governments will need to step in to make sure that individuals are protected from the potential negative effects of automation. The simplest solution to this would be a higher rate of taxation against the returns generated by artificial intelligence which is then spread broadly across the population in the form of something like a universal basic income. However, this is not a silver bullet — for example, if implemented by a single government, it would fail to ensure that AI doesn’t worsen international income inequality.
Most importantly, we expect that these protections and mitigations won’t come by default. Instead, employees, activists, and the public will need to make our voices heard in order to ask companies and governments to take action and prepare for the potential disruptive economic effects of AI automation.
To learn more and join our movement demanding an AI transition that benefits everybody, click the link below.