Generative AI like LLMs has been touted as a boon for collective productivity. But the authors say relying too much on hype could be a mistake. Productivity assessments typically focus on the level of the task and how well people could use and benefit from LLMs. Use these results to draw general conclusions about business level performance could prove costly. The authors argue that executives need to understand two fundamental problems with LLMs before adopting them company-wide: 1) their persistent ability to produce convincing falsehoods and 2) the likely long-term negative effects of use LLMs on employees and internal processes. The authors present a long-term perspective on LLMs, as well as the types of tasks LLMs can reliably perform.
Large Language Models (LLM) have been presented as a boon for collective productivity. McKinsey boldly proclaimed that LLMs and other forms of generative AI could increase corporate profits globally by $4.4 trillion per year, and Nielsen announced a 66% increase in employee productivity through the use of these same tools. Such projections have made finding ways to use these tools – and maximize productivity – a top priority for many businesses over the past year. While we are intrigued and impressed by this new technology, we advise cautious experimentation rather than mass enterprise-wide adoption.
