In this biweekly newsletter, Roula Khalaf, the FT’s director, handpicks her preferred articles.
The future appears grim for white-collar employees ever since the emergence of conceptual artificial intelligence in November. OpenAI, the entity behind ChatGPT, warns that individuals with the highest incomes face the greatest threat from the latest AI advancements, with those earning a six-figure salary being approximately three times more vulnerable than those making $30,000 annually. McKinsey also highlights concerns regarding the ability to oversee the utilization of knowledge.
It may be tempting to disregard these warnings as mere predictions. Yet, many have been lulled into a false sense of security by the belief that technological advancements only impact others’ jobs and not our own, a misconception perpetuated over millennia.
Nevertheless, for some, the fear of AI displacing white-collar workers has become a reality. A recent groundbreaking study revealed that copywriters and graphic designers on popular online freelance platforms experienced a significant reduction in job opportunities and a notable decline in wages shortly after the introduction of ChatGPT. This indicates that conceptual AI not only overtakes their roles but also diminishes the value of their work.
The study’s most notable discovery was that freelancers who previously earned higher incomes and put in more effort were equally susceptible to experiencing setbacks in their careers and earnings as their counterparts. If anything, the outcomes were even more severe. Essentially, possessing more skills does not shield individuals from job or income loss.
However, the online blogging sector covers a very specific segment of white-collar jobs and the labor market. What about examining the performance of information workers as a whole?
For this purpose, a captivating new study from Harvard Business School analyzed the impact of equipping Boston Consulting Group employees with GPT-4, OpenAI’s latest cutting-edge tool.
During a series of consulting assignments, BCG employees who were randomly assigned to utilize GPT-4 demonstrated significantly higher efficiency compared to their colleagues who did not have access to the tool. These AI-assisted consultants not only completed tasks 25% faster and accomplished 12% more tasks overall but also produced work of 40% higher quality than their unassisted peers.
Employees at all skill levels benefited from this implementation, with the less experienced members of the workforce experiencing the most significant performance enhancements, following a common trend observed in conceptual AI studies. These large language models excel at summarizing and regurgitating existing human expertise in the public domain, which aligns with intuition. The advantage of employing them diminishes as one’s personal understanding approaches that threshold.
There was, however, one drawback: AI-assisted consultants struggled with a more intricate task that necessitated analyzing statistical data after thorough review of quantitative materials, as GPT overlooked nuanced details. Nevertheless, two participant groups deviated from this trend. The first group, labeled “cyborgs” by the authors, seamlessly integrated with the AI, continually shaping, verifying, and refining its responses. The second group, termed “centaurs,” divided tasks, assigning more AI-suited subtasks while focusing on their areas of expertise.
When viewed collectively, these experiments highlight three key points. Firstly, legislation may play a pivotal role, especially in sectors like online freelance work, which currently lack adequate protections. Secondly, roles with greater complexity are less susceptible to complete automation. Gig-worker arrangements, involving the completion of tasks for multiple clients, such as advertising or logo design, are particularly vulnerable.
Lastly, to maximize the benefits of these tools while mitigating their limitations, they should be regarded as extensions of our capabilities, with results evaluated as rigorously as one would assess their own work. They are not standalone, infallible assistants to be surrendered to or held accountable for.
Generative AI poses a challenge to white-collar workers in this nascent era. While exercising caution is prudent, this relationship could evolve into one that thrives.
John. @jburnmurdoch , [email protected]
