This research produced a list of occupations and their level of exposure with the advent of LLMs and Generative AI.
LLMs on their own is impactful, but there has been an explosion of LLM-powered software, also referred to as Generative Apps. This explosion is being fuelled by the advent of LLM specific IDEs like Stack, Flowise, LangFlow and more.
Considering this proliferation of real-world implementations, OpenAI found that:
- 80% of the U.S. workforce can have > 10% of their work tasks affected by LLMs.
- While 19% of workers will have > 50% of their tasks impacted.
- And 15% of all worker tasks in the US could be expedited considerably.
LLM-powered software will have a substantial effect on scaling the economic impacts of the underlying models.
Considering the graph above, OpenAI created five Job Zones, these zones are created based on similar criteria:
Job Zone 1
- None or little (0–3 months) preparation required.
- High school diploma or GED (optional)
- Example Occupations: Food preparation workers, dishwashers, floor sanders
- Median Income: $30,230
- Percentage > 50% exposure: 0%
Job Zone 2
- Some (3–12 months) preparation required.
- High school diploma
- Example Occupations: Orderlies, customer service representatives, tellers
- Median Income: $38,215
- Percentage > 50% exposure: 6.11%
Job Zone 3
- Medium (1–2 years) preparation required.
- Vocational school, on-the-job training, or associate’s degree
- Example Occupations: electricians, barbers, medical assistants
- Median Income: $54,815
- Percentage > 50% exposure: 10.67%
Job Zone 4
- Considerable (2–4 years) preparation required.
- Bachelor’s degree
- Example Occupations: Database administrators, graphic designers, cost estimators
- Median Income: $77,345
- Percentage > 50% exposure: 34.5%
Job Zone 5
- Extensive (4+ years) preparation required.
- Master’s degree or higher
- Example Occupations: Pharmacists, lawyers, astronomers
- Median Income: $81,980
- Percentage > 50% exposure: 26.45%
The table below shows the occupations with the highest exposure:
OpenAI estimate that GPTs and GPT-powered software are able to save workers a significant amount of time completing a large share of their tasks, but it does not necessarily suggest that their tasks can be fully automated by these technologies.
In closing, it has been estimated by human assessments that only 3% of U.S. workers would have half or more of their tasks exposed to LLMs without installing any more software or modalities.
However, taking into account other generative models and complementary technologies, it is estimated that up to 49% of workers may have half or more of their tasks exposed to LLMs.
I’m currently the Chief Evangelist @ HumanFirst. I explore and write about all things at the intersection of AI and language; ranging from LLMs, Chatbots, Voicebots, Development Frameworks, Data-Centric latent spaces and more.