Study: Which Jobs Will Be Impacted by Tools Like ChatGPT?

Which Jobs Will Be Impacted by Tools Like ChatGPT

Are we at the tipping point of the new wave of AI tools causing a ruckus in the job market? This question is not new; in fact, it has been raised since the industrial revolution. With the emergence of artificial intelligence tools, which have been accumulating in recent months, the issue is gradually returning to the forefront.

Researchers from OpenAI, Open Research, and the University of Pennsylvania have delved into this matter, producing a study on the potential impacts of generative AI on various professions. Let’s take a closer look!

80% of jobs may be impacted by ChatGPT

In their research, they analyzed over 1,000 professions in the United States, breaking down each occupation into several tasks.

The findings reveal that a minimum of 10% of tasks in 80% of the occupations will be transformed by artificial intelligence. For 19% of jobs, the number of affected tasks increases to 50%. The study was conducted using OpenAI’s latest model, GPT-4. Upon its release, the company highlighted the significant advancements this AI model represented, particularly in comparison to human abilities.

For instance, it achieves a score comparable to the top 10% when taking the lawyer’s examination, while its predecessor, GPT 3.5, ranked in the bottom 10%.

Which jobs have a high risk of being automated?

One key insight from the study is the nature of the jobs threatened by generative AI. High-income jobs are particularly at risk of being significantly disrupted. This primarily includes professions where the use of computer software is central. More specifically, jobs that require programming and writing skills could undergo substantial changes. On the other hand, the researchers explain that science-related professions should be minimally impacted by artificial intelligence.

The study found that roles heavily dependent on scientific knowledge and critical thinking demonstrate a negative correlation with exposure to large language models (LLMs), while programming and writing skills will be impacted by LLMs.

The study provides a non-exhaustive list of jobs that are most likely to evolve:

  • Public Relations Specialist
  • Pollster
  • Legal Secretary
  • Poet, Lyricist, or Creative Writer
  • Mathematician
  • Interpreter, Translator
  • Engineer
  • Data Manager
  • Writer
  • Proofreader
  • Web and Digital Interface Designer
  • Accountant
  • Administrative Assistant
  • Journalist

Conversely, jobs based on manual labor should experience very limited impact. The study specifically mentions certain occupations that are not expected to change in the short term, which seems logical: stonecutter, heavy equipment operator, athlete, plumber, tile setter, mechanic, cook, and others.

The study should be taken with a grain of salt

It’s essential to maintain a balanced perspective regarding this study. Firstly, it’s worth noting that OpenAI participated in this research and has a vested interest in promoting the strong potential of its innovation. Moreover, making predictions on such subjects is challenging, and numerous factors cannot be considered, such as technological advancements or changes in legislation and professions.

The researchers themselves call for caution, especially concerning the complexity of jobs, which is difficult to apprehend:

It’s hard to determine to what extent professions can be entirely broken down into tasks, and if this approach systematically omits certain categories of skills or tasks that are implicitly required for competent job execution. Furthermore, tasks can be composed of sub-tasks, some of which are more automatable than others.

Lastly, artificial intelligence still has several drawbacks to overcome, and human intervention is often needed to ensure the reliability of the content produced by the tool.

On this subject, the researchers specify:

We acknowledge that there may be discrepancies between theoretical and practical performance, particularly in complex, open, and domain-specific tasks.

It’s worth noting that while in the immediate future, things are still going to be shaky with these tools, it’s more than likely that in the coming years, we will see drastic improvements.

Specifically, improvements in areas such as “hallucination” and the general ability to provide factual information. Further, as computing power (which is used to train the AI models) becomes more affordable and powerful, we will see drastic improvements in how quickly models like GPT can be retrained on new information.

Previous Post
AI Tools for Music & Audio Generation

8 AI Tools for Music & Audio

Next Post
GitHub Copilot Adapts to GPT-4

GitHub Copilot Adapts to GPT-4: A Breakdown of the New Features

Related Posts