My last five years of work?

Something different for today. Are we prepared for the exponential growth of AI in the next 3 to 5 years?

Avital Balwit, who works at an AI company, believes that within the next 3-5 years artificial intelligence may advance to the point of making most human knowledge work obsolete. While work is not the only way to participate in society, it has proven to be the main way to transfer wealth and resources, provide social connection and meaning, and offer political stability.

With each iteration, AI language models are becoming more capable at a wide range of cognitive tasks like content generation, text summarization and analysis. While they still struggle with some higher-level tasks, economically what matters is whether they can outperform the human who would otherwise do a given task, not the very best humans. The author expects AI to excel first at online knowledge work like copywriting, tax prep, customer service, software development, and legal work. Physical and social/care work will likely be automated later.

Studies on the mental health impacts of unemployment have had mixed findings, ranging from 10-100% increases in mortality to negligible effects. Challenges include disentangling causality (since mental/physical health issues can lead to job loss) and isolating psychological effects from financial stress. However, the shame and blame associated with job loss seems to play a key role. Studies on workers laid off due to the pandemic or plant/industry closures, where job loss was normalized and widespread, found lower distress levels, at least initially. This suggests the psychological impact may be less severe with mass AI-induced unemployment.

The amount of work deemed necessary has declined over the past 150 years as countries became wealthier, so current norms around a 40-hour workweek may not be optimal. Research indicates that moderate amounts of discretionary time are best for wellbeing, and that increased leisure time is not detrimental if it is spent on social and productive activities. Many are already occasionally happy without working, such as on weekends and vacations. Some studies find improved mental health after retirement.

Women's entry into the workforce in the 20th century did not seem to increase their self-reported happiness. Technological advances have reduced time needed for housework. While AI may automate more domestic tasks, a human preference will likely persist for relationship-based work like childcare and teaching.

Historically, the aristocracy led lives of relative leisure, filling their time with social events, hobbies, art and literature. This may be a relevant analog for a post-work society where peers are similarly unemployed. Science fiction, such as Iain Banks' Culture series, has depicted post-scarcity civilizations grappling with citizens' need to feel useful and find meaning.

The author engages in some suboptimal activities, like ballet, for the hedonic and self-improvement value. An AI researcher mentioned taking up hobbies he's not naturally skilled at to prepare for a post-AGI world. Ultimately, the author believes that if we develop AGI capable of replacing human workers, it should also be able to help us find solutions to the existential challenges and loss of meaning that technological unemployment may create.

In summary, the article provides a nuanced look at the complex psychological and social implications of potential mass job displacement due to AI, drawing on economic and sociological research as well as historical and fictional analogues. It considers challenges but also identifies some reasons for optimism about human wellbeing and meaning in a possible post-work future.