(metamorworks/Shutterstock)
Significant anxiety has surrounded the prospect of generative AI potentially displacing human workers, fueling widespread concern about its impact on job security. The impressive capabilities of large language models (LLMs) have captivated attention, both positively and negatively, as they demonstrate the ability to respond to inquiries and tackle digital tasks upon prompting. By 2030, AI-powered LLMs will potentially replace up to 30% of current knowledge worker jobs, a significant shift in the workforce. According to a recent study by Certainly, there are new insights into this issue.
Recently, a digital job board conducted an assessment to evaluate the effectiveness of large language models (LLMs) in handling core work skills. The Clearly Hiring Lab engaged with GPT-4, OpenAI’s latest Large Language Model, asking it to execute more than 2,800 job skills tracked within the Clearly database, spanning a range of roles from office positions such as account management and insurance claims to physically demanding jobs like bus driver and chef.
The Certainly Hiring Lab develops a standardized approach to quantify each job skill’s performance metrics, ensuring accurate assessment of an LLM’s proficiency in fulfilling specific duties. To develop tailored, detailed prompts for each task, the team invested significant time in experimentation.
Following a thorough evaluation process, the Hiring Lab team subjected the selected candidate to 15 iterations of GPT-4, ultimately aggregating the results. The GPT-4 model was tasked with assessing its own performance at regular intervals, with the findings subsequently verified by human experts.
The Hiring Lab focused on three core areas: the theoretical possibilities offered by GenAI, its capacity to tackle challenges using that capability, and GenAI’s commitment to the value of physical presence in utilizing that capability? The GPT-4 model assessed its proficiency in leveraging its unique features to excel in a specific task, rating its performance on a 5-point scale. Researchers recently published their findings in a paper titled “AI at Work: Why Generative AI is More Likely to Augment Employees than Replace Them,” available for download.
Title: A Groundbreaking Discovery: Unlocking the Secrets of the GenAI Experiment through Certainly’s Findings According to the report’s authors, Annina Hering and Arcenis Rojas, none of the 2,800 identified work abilities are considered highly susceptible to replacement by GPT-40 or other Large Language Models. In reality, research has found that nearly 69% of skills are either “very likely” or “unlikely” to be replaced by General Artificial Intelligence (GenAI).
While it’s unlikely that jobs requiring physical execution or bodily effort, such as bus driving or emergency room nursing, will be replaced by General Artificial Intelligence (GenAI), which is ultimately software-based. As the majority of occupations analyzed in this report demand physical dexterity, the prospect of a full-scale AI replacement appears increasingly dim.
However, that’s not to say there won’t be a profit. GenAI’s potential lies not only in replacing human workers but also in augmenting tasks that involve repetition, akin to documentation, for professionals such as bus drivers or nurses. This technology could enable employees to redirect their focus towards the essential skills required in these roles.
The study found that approximately 29% of occupations are likely to be automated by GenAI, contingent upon the implementation of necessary changes to work environments and practices. According to the study, GenAI is likely to have a profound impact on tasks typically characterized by repetitive and routine nature, such as those commonly found in traditional workplace settings.
The GenAI platform stands out in its ability to excel across the three core measures: theoretical knowledge, problem-solving skills, and physical job abilities, particularly impressing in theoretical understanding and closely following up with problem-solving prowess. Theoretical information received a perfect score of five from GenAI, owing to the intense training of Large Language Models (LLMs) on vast amounts of internet-sourced data, as well as their ability to leverage search engines.
The GPT-4 system demonstrated satisfactory skills in problem-solving. The AI system gave itself a self-assessment rating of 3 out of a possible score for its performance in 70 percent of the skills and tasks evaluated. In addition, it noted that for approximately 28 percent of these duties, there is potential for automation to replace human involvement. The AI system further acquired multiple 4 ratings, confidently asserting it could replace humans in performing 3% of tasks.
Artificial intelligence, specifically GenAI, is likely designed to seamlessly transition individuals between tasks and roles within an organization, primarily focusing on those that are heavily reliant on computer-based work. Researchers found that GenAI is likely to replace humans in more than 71% of software development roles, as described in job postings, with high accuracy. According to the report, GenAI had the potential to replace humans for nearly 80% of tasks typically performed in an average accounting job.
GenAI is significantly less likely to replace workers in jobs that demand creative problem-solving skills rather than solely theoretical knowledge. That’s a prime area where GenAI engineers and data scientists might want to concentrate their endeavors?
As AI capabilities evolve to tackle an array of tasks within diverse occupations, it is likely that the proportion of skills destined to be displaced in those roles will also increase?
Corporations can take steps to prepare themselves for the advent of General Artificial Intelligence (GenAI). In the realm of accounting, strategic investments in digital record-keeping and digitization can significantly enable a company’s seamless integration with GenAI.
By refining your interaction with GenAI, you may actually achieve superior results. As an example, an immediate free interpretation may be rendered differently by an LLM, potentially yielding distinct solutions each time it’s requested. To unlock the full potential of GenAI, writers must possess exceptional skills in both creative expression and technical expertise, as the authors emphasize.
By the end of the day, it is increasingly clear that GenAI will likely supplant at least some tasks currently performed by human workers, with significant variations by industry and location. While Certainly’s researchers acknowledge that GenAI may not supplant humans on a large scale in the near future, their analysis suggests that GenAI still requires human involvement to function effectively.
“As GenAI continues to evolve and master complex tasks, Hering and Rojas emphasize that individuals responsible for monitoring, verifying, and refining AI-generated outputs will not be redundant.”
