The Oxford Martin Working Paper Series on - the Future of Work

Consider the following ChatGPT request by one user: “Write the complete script of a Seinfeld scene in which Jerry needs to learn the bubble sort algorithm.” To achieve this, the AI drew upon its training, on a vast body of human text that likely includes scripts, to identify the critical “features” of a “Seinfeld script”, such that, in its response, the AI assigns greater probabilities to words it. [...] People who can make their presence felt in a room, that have the capacity to forge relationships, to motivate, and to convince, are the people that will thrive in the age of AI. [...] The venture capital industry provides another case in point: when the industry shifted to remote work during the pandemic, investors sought to make up for the loss of soft information by leveraging their existing networks and collaborating with partners with whom they had prior experience.13 And the importance of human trust is only amplified by the workings of LLMs. [...] The future of work The physicist Niels Bohr supposedly once joked that “God gave the easy problems to the physicists.”26 While the laws of physics are time invariant, and apply across time and space, boundary conditions in social sciences are not timeless. [...] 11 Thus, beyond the advances outlined above, the potential scope of automation is unlikely to substantially grow merely through scaling existing models.41 In conclusion, while we expect AI to continue to surprise us, and for many jobs to be automated away, in the absence of major breakthroughs, we also expect the bottlenecks we outlined in our 2013 paper to continue to constrain our automation pos.