Slow AI

Another week, another study exploring AI’s role in automation. However, unlike previous ones, this MIT study by Svanberg et al. didn’t just ask whether AI could perform a task, but also if it’s affordable to use AI for it. This is a more complicated question to answer, hence the decision to only study computer vision because of more information about its costs and applicability.

 
At current costs, the study shows, most businesses would not consider automating vision tasks. It was found that automating these tasks would be cost-effective for only about 23% of the wages currently allocated to them. This indicates that while the technical feasibility of automation with computer vision exists, it would be economically viable in only a quarter of cases.

The study suggests that AI will spread slowly, giving us more time to handle its impact on jobs, as mentioned in last week’s newsletter. The authors showed that the pace of vision task automation depends on how fast costs decline. Even at a 50% annual cost decline, adoption would be slower than the rate of US job loss between 2017-2019. Overall, the fear of AI replacing jobs, especially in computer vision, seems exaggerated.

But LLMs are different. A foundational language model can generalise across tasks more easily than an image model. For example, I don’t have to fine-tune ChatGPT to produce ad copy for a marketing campaign. But for vision tasks, you need to tailor it to specific jobs, like spotting defects in a product. Another difference is that text data for fine-tuning LLMs is often cheaper and more available than images. One of the authors said that: While AI systems are certainly rolling out quickly, their improvements are remarkably predictable, as work in our lab and others demonstrate.

Συνέχεια εδώ