In healthcare, there are high hopes for AI. A US study published in the Journal of the American Medical Association found that ChatGPT on its own demonstrated higher performance than physician groups in improving diagnostic capabilities. However, the study also revealed that doctors sometimes stick to their initial diagnosis even when ChatGPT suggests a better one, and that some physicians don't know how to best exploit the tool's capabilities. Another study at MIT found that AI assistance in material science research led to the discovery of 44% more materials and a 39% increase in patent filings, but also a sharp reduction in job satisfaction among researchers.
Key takeaways:
- Drew Breunig has categorized AI technology into three use cases: gods, interns, and cogs. "Gods" refer to super-intelligent, autonomous AI entities, "interns" are supervised AI that collaborate with experts, and "cogs" are machines optimized to perform a single task extremely well.
- AI "interns" are already in widespread use in many industries and occupations, augmenting human capabilities. They represent the first generation of quasi-intelligent machines with which humans have had close cognitive interactions in work settings.
- A study published in the Journal of the American Medical Association found that the AI tool ChatGPT did not significantly improve the diagnostic capabilities of physicians. However, ChatGPT on its own demonstrated higher performance than both physician groups.
- Collaborating with AI can have an impact on job satisfaction. An experiment at MIT found that while AI assistance led to the discovery of 44% more materials and a 39% increase in patent filings, it also resulted in a sharp reduction in job satisfaction among researchers.