Tomorrow, the «AI Useless Class»: a recent MIT study tends to prove the contrary

What if the further artificial intelligence advances, the more humanity becomes its collateral victim? Contrary to alarming projections, the most recent MIT (Massachusetts Institute of Technology) study suggests that AI may not be as disruptive to employment as previously anticipated. What is the efficiency of AI in performing tasks compared to humans? How costly is it for businesses to replace human labor with AI? By focusing on these key factors, researchers aimed to provide a clearer picture of AI's potential impact on employment.

What if businesses were taking more risks using AI?

Their findings present a more tempered view of AI's role in job displacement. Currently, computer vision AI, a subset of AI technology, can automate 1.6% of tasks achieved by worker wages in the US economy, excluding agriculture. However, only 0.4% of these tasks are economically advantageous for businesses to automate instead of hiring human workers at current costs.

This revelation challenges previous estimates by institutions like Goldman Sachs and McKinsey that suggested a higher percentage of jobs could be automated shortly. Moreover, the University of Pennsylvania, NYU, and Princeton established potential impacts of models like ChatGPT on jobs, painting a more alarming picture. A good reason to be anxious?

Not yet, says MIT. Neil Thompson, a research scientist at MIT CSAIL and a co-author of the study, highlights a key takeaway: the potential AI disruption might occur more gradually and less dramatically than some commentators have suggested. While the study shows the limitations of current AI technologies in replacing human labor, it also underscores the need for further research to understand AI's full impact comprehensively.

ChatGPT was not taken into account

However, the MIT study acknowledges its limitations. It primarily focuses on jobs requiring visual analysis and does not explore the broader implications of text- and image-generating models like ChatGPT. Not to mention the fact it does not consider scenarios where AI could augment human labor or the potential cost savings from pre-trained models like GPT-4.

Amidst these findings, regulatory developments regarding AI are also underway. The European Commission's plan to establish "AI Factories" aims to stimulate AI adoption in strategic sectors. While international cooperation tries to regulate AI responsibly, legislation addressing AI's societal impact is still in progress. The European Parliament is set to vote on the new AI Act deal, but significant changes may not take effect until at least 2025.

Meanwhile, the UK's hosting of the world's first AI conference signifies the global recognition of one purpose: the need, in the near future, to pursue our dream of not being irrelevant.