The rise of AI: challenges and opportunities in the job market
Artificial intelligence has rapidly become a major disruptor in the job market, with machines and algorithms replacing human workers at an alarming rate.
We may be witnessing an inflection point with artificial intelligence (AI), notably OpenAI’s GPT large language model replacing jobs through automation.
Last week the news and entertainment company BuzzFeed Inc (NASDAQ: BZFD) said it would use the GPT-3 engine to create quizzes and other content targeted towards teenagers and young adults, which sent its stock price soaring over 300%.
The company is struggling to remain relevant, and laid off 12% of its staff in the wake of a US$27 million loss in the third quarter.
But BuzzFeed was not the first major company to come up with the idea of replacing its staff through automation. Microsoft (NASDAQ: MSFT) fired over 50 of its editors and journalists that curated content for its Microsoft News and MSN news portals. Instead, algorithms similar to GPT will select which news stories will appear on the front pages.
The question for investors watching this unfold is, are large language models like OpenAI’s GPT-3 and Google’s Sparrow going to usher in a new paradigm of robots replacing more jobs, or is the hype and impact of the tools overblown?
The answer could be both.
Parallels from blockchain
If history repeats itself, we may see many more rallies in the stock prices of companies that announce they will use GPT-3 for some of their creative output.
This feeling of euphoria is comparable to that surrounded bitcoin in May 2020.
Back then, the coin reached an all-time high of US$20,000 for the first time and flooded its way into mainstream consciousness.
Blockchain was seen as the new paradigm back then, too.
People saw it as the end of the global banking hegemony, and decentralisation became the decade’s final buzzword.
Companies like On-line Plc, which later changed its name to On-line Blockchain Plc , saw a rally of 394% in its stock price by this act alone.
Countless initial coin offerings (ICO) were on sale, most with the intention of becoming the next big token for investors to pour their savings into.
But only a few years later, and today people investing in ICO are almost unheard of, partly due to over a 95% drop in marketing and fundraising for these investments.
This decline in the frenzy surrounding cryptocurrencies comes despite Bitcoin continuing to make higher highs, as it would later reach $65,000 in 2021.
As Bitcoin ages into a mature asset class, a corresponding amount of its glamour fades, and people are left with a realistic view of its risks and opportunities.
ChatGPT and other models like it are going through a similar adoption curve, and significant challenges stand in the way of completely automating humans out of their jobs.
Challenges and opportunities
Large language models excel at alleviating humans from doing simple, unproductive tasks. People whose jobs revolve around these basic tasks face the most amount of risk of losing them to automation.
While at the higher levels of skill, these models are more likely to augment people’s productivity rather than replace them.
Unlike machines, experts naturally mesh quantitative and qualitative data points when making complex decisions. Algorithms need to be trained on how to factor in the context and the bigger picture to give them an equal footing to human skill, and at present that’s impossible.
This problem will be why we’ll be very unlikely to see a purely AI-written book become a New York Times best-seller or AI-composed music reach the top of the billboards.
Machines are constrained in two ways that make this cost-prohibitive.
First is processing power, or the death of Moore’s law. We stopped seeing exponential gains in transistor growth a decade ago, by some estimates.
The reason is we can only pack so many of them on a limited surface area, thus leading to a perceptible drop off in processing power year over year.
Less computing power means it’s more expensive to run AI models, and as language models become more advanced, their demand for computing power increases, and thus scales in cost exponentially in two directions as time goes on.
Still, one day the commercialisation of quantum computing will mean that AI model maintainers can access an abundance of processing power and keep their costs low for the general public. This may unleash a wave of people losing their jobs to the machines.
It’s not all bad, though.
One theory is that as AI replaces humans, taxes may be levied on companies that institute layoffs to give their workers pensions or a huge severance as compensation.
The end goal could be that as more and more humans no longer need to work, society could harness AI’s economic output to give everyone a universal basic income (UBI) so that we may move towards a values-driven society and not one directed by capitalism.