I have been asking myself lately whether OpenAI’s GPT has positively impacted my productivity so far? To be honest, the answer is a resounding no. While GPT is an incredibly and increasingly powerful tool, I simply spend too much time goofing around, testing different approaches, and indulging my inner nerd. The negative impact on my productivity, which is working on original research, pushing my current projects forward, and responding to reviewers, among others, has obviously suffered.
But here is the thing, mastering this tool requires interacting with it – it is learning by doing. Getting a feeling for the model, which prompts work and which don’t, asking the same question in different forms and several times, observing and reflecting on changes as the model evolves is key. Sam Altman, the CEO of OpenAI, made this point quite effectively during a recent interview with Lex Fridman.
As a tenured academic, I'm privileged because learning, trying, failing, and testing are part of my job description. But what about other knowledge workers who must meet deadlines or have teammates who depend on them? It's unrealistic to expect them to explore AI during their personal time.
Thus I would like to propose a mini-manifesto (it is a beta version) for companies seeking to benefit from this new technology.
Allocate designated exploration time: Allow employees to explore AI tools, like GPT, during work hours. This will help them build confidence and expertise without sacrificing personal time.
Provide training and resources: Offer structured training sessions and materials to help employees understand the tool’s capabilities, limitations, and best practices.
Foster a culture of sharing: Encourage employees to share their experiences, insights, and tips with colleagues
Balance exploration and exploitation: Late James March used to say that exploitation of current capabilities is more seductive for organizations. It is less risky, the effects are immediate, and people know what to do. However, companies that are stuck in this mode eventually perish. Thus, last but not least, keep evaluating the impact of GPT (or similar technology) on productivity and make necessary adjustments to ensure a balance between exploration and exploitation.
One of the key principles behind OpenAI is that if we introduce those tools early and develop them slower, we will give society and our institutions time to adjust to them. But we need to invest the time so that organizations and institutions can integrate this tool into their workflows. Otherwise, all this extra headstart will be for nothing.
Comments