WEF Jobs of Tomorrow Generative AI 2023 · Page 5

Introduction: How will 
large language models 
impact the jobs of 
tomorrow?
Labour markets are undergoing rapid transformation 
from the trajectory of growth, geoeconomics, 
sustainability and technology. The Future of Jobs 
Report 2023 found that business leaders expect 
23% of global jobs to change in the next five years.2 
In particular, generative artificial intelligence (AI) 
has undergone a profound leap in capabilities, 
embodied in products such as GitHub’s Copilot for 
programming, Midjourney for image generation and 
ChatGPT as a universal language assistant. The 
Future of Jobs Report 2023 also found that AI and 
text, image and voice processing technologies more 
generally are top of mind for businesses. The report 
found that 75% of survey respondents report having 
plans to adopt AI in their organization’s operations, 
and 62% report having plans to adopt text, image 
and voice processing technologies.3 This has raised 
questions about how this new technology will affect 
organizations and labour markets around the world.
This white paper examines the potential near-
term, direct impact on jobs of a particular type of 
generative AI, large language models (LLMs), which 
have been highly visible in public debate over the 
past year due to their human-like ability to create 
and understand language. As LLM services have 
exploded in popularity, with free services such as 
ChatGPT reaching as many as 100 million active 
users within the first two months of its debut,4 
the capabilities of these models, paired with their 
accessibility and rapid adoption rate, suggest that 
many work tasks – and jobs that emphasize them – 
could be impacted by the use of LLMs in the years 
to come. By some estimates, up to 62% of work 
time involves language-based tasks.5 Yet, artificial 
intelligence and text, image and voice processing 
technologies also have the potential to augment 
work and create new jobs. In addition, many roles 
remain wholly unaffected by these developments. 
Rapid technological change often generates 
anticipation regarding its effects on daily life, 
particularly jobs. In the aggregate, previous 
innovations have led to more employment 
opportunities, better-quality jobs and a higher 
quality of life, but they also create disruption and 
displacement.6 This paper aims to support the 
detailed analysis required to take a clear-eyed view 
around impact, opportunity and preparation.
Generative AI, LLMs  
and language tasks
The newest forms of groundbreaking generative AI 
models are created via deep learning, which is the 
process of training foundation models on very large 
sets of data. These foundation models are typically 
created in the form of a neural network, whose 
structure is inspired by the arrangement of neurons 
in the human brain. Large foundation models are 
trained on vast amounts of data and have seemingly 
super-human levels of predictive capacity, which 
can be harnessed by producing text or images in 
response to a written prompt.7
So far, generative AI models have been configured 
into a variety of different tools to serve different 
contexts, such as image, audio or video 
creation, identifying financial fraud and other 
security risks, and a host of general language 
capabilities, including the ability to generate natural, 
mathematical and computational language. While 
there is a broad range of implementations of 
generative AI, this study will focus on LLMs and 
their unique language-generat