What is going to get automated in the limit. Everything? At what rate? Will the rate be consistent given type of work? I think these are all interesting questions that deserve some level of meditation. Lets first classify a MECE grouping of work. I have come up with the following three highest level buckets:
IMO this is first to go. This will have the fastest rate to achieve 90% automation, that I think happens in 10 years. AI has made significant strides in automating decision-making processes already. Machine learning algorithms can analyze vast real-time datasets to make consistent, accurate, and swift decisions. AI never shows up drunk or hungover, is never going through a divorce. AI is remarkably consistent and improving. AI-driven systems can predict consumer demand, optimize supply chains, and assist in medical diagnoses at faster more consistent rates than humans. Additionally many humans are remarkably intransigent. Thank god we do not live forever! On the other hand, AI system learn from a mistakes, this knowledge can be disseminated across all similar systems globally, leading to continuous improvement. Unlike human workers, AI doesn't require rest and can operate tirelessly, ensuring efficiency and scalability. While I anticipate that up to 90% of decision-making tasks will be automated within the next decade, certain decisions will be held out—such as jury deliberations, electoral choices, and final hiring or firing decisions—may remain under human purview, albeit with AI assistance.
This is next. I think it will hit 90% automation in 50-100 years. Automating physical tasks presents more challenges due to the complexities of the physical world. While "task-specific" technologies like automated lawn mowers, house-painting robots, brick-laying machines, and 3D-printed houses are emerging, scaling these solutions requires extensive supply chains, intricate manufacturing processes, and navigation of import-export regulations and national security concerns. The development and deployment of humanoid robots could accelerate this automation, but it's still not going to be as fast the former bc you have real supply chains, real geopolitical lines to cross, VAT taxes, embargos, black lists, national security issues, etc. This will be more of a taper vs. an explosion, more of a gradual process. The geopolitical factors will lead to many winners, 3-4 for the West, 2-3 for China, etc. segmented development similar to the defense industry.
This will be last. Most people don't think so. While AI can do a lot of interacting, humans like to deal with humans sometimes. Using AI in certain situations feels sad and lonely. We are social animals. We are not thinking machines that feel, we are feeling machines that think. Maybe AI will sympathize with that better than humans one day, but they can never empathize because they are not human. Roles that involve personal interaction are the most resistant to automation. Human connection is a critical component in professions like therapy, entertainment, sales, recruitment, politics, and high-end hospitality. For example, patrons paying a premium at a fine dining establishment expect personalized service from human staff. Similarly, individuals facing serious medical diagnoses often prefer consultations with human doctors to provide empathy and nuanced understanding. While AI can assist in these roles, fully replacing the human element is unlikely. Perhaps this will get to 90% automation but it will require a rewiring of our society, and could take 50 to 200 years if ever.