My co-founder at Growcreate often teases me for thinking the Greeks invented everything. Whilst inventing pizza is perhaps far-fetched, ancient Greeks certainly gave us pan metron ariston, the practice of moderation (metron) in both attitude and consumption.
But the Greeks knew that moderation does not make for good TV, so they invented tragedy as well. In a typical storyline, the hero forgoes moderation, commits hybris and meets their inevitable nemesis.
Nothing in the modern world is less moderate than AI debates. Total utopia or total dystopia. Superintelligence or just words on a screen. Mass poverty or universal abundance.
With Anthropic heading for a listing, OpenAI hot on their heels, and American midterms in November, the drama is building up. As the various debates on AI heat up, we should remember that metron always wins the day.
Key ideas in this post
The AI debate is conducted at the extremes, and neither side is right. It is transformative, but a “normal” technology that will diffuse slowly. Use it selectively, to save brain space rather than to outsource thinking. Manage the transition rather than betting on abundance. Most people are already there: they adopt the technology but mistrust the pitch.
“AI is the final tool” versus “AI is just a tool”
One camp holds that solving intelligence means everything is solved. The other reckons that AI is a fancy autocomplete. Arvind Narayanan and Sayash Kapoor offer a better framing of AI as a normal technology.
“Normal” does not mean small: electricity and the internet were also normal, but transformed the world. At the crux of the argument is that technology, applications and adoption move at different speeds. Electric dynamos had no impact on productivity statistics for 40 years until factory owners redesigned the floor around them.
AI is transformative, but transformation will take a long time. Diffusion is the constraint, itself constrained by things like governance and work habits. AI will change the world, but the “revolution” will not be televised.
“Use AI for everything” versus “never use AI”
Work slop is everywhere. Stanford and BetterUp found that 40% of American employees received AI-generated non-content, resulting in nearly 2 hours of rework each time. At the other end, excitement among Gen Z has fallen by 14 points over the past year, while anger has climbed to 31%. Is there really nothing between “always” and “never”?
Ethan Mollick’s guide remains the best account of the moderate argument. Use AI where the volume overwhelms and where you can test the output. Avoid it where accuracy is non-negotiable and where the friction is the point. If you use AI to create more stuff, then it’s time to log off. Freed time is only worth having if you spend it on things that matter.
“AI will kill all jobs” versus “AI will create abundance”
Few debates around AI are as emotive as those about its impact on jobs. Molly Kinder’s “Messy Middle” frames today as Reality One, characterised by alarming headlines but scant evidence. The promised post-AGI world is Reality Three, a world of abundance and universal incomes. Whether that materialises or not, how do we manage the transition without scarring the generations in the middle?
Deindustrialisation benefited many and hurt many more, laying the foundation for our extreme politics twenty years later. The knowledge-work equivalent will be wider and sharper. The moderate approach is to protect jobs now rather than waiting for the data to confirm the problem.
Nothing in excess
It is surprising that in an era of polarisation and rage-bait, most people appear to take a moderate attitude towards AI.
We clearly love using it: AI reached mass personal use faster than any consumer technology before it. Gallup nonetheless finds that 52% of Americans think it does equal amounts of harm and good, and that only 27% trust business to use it responsibly.
Adoption and scepticism are not altogether surprising bedfellows. People like the tech and mistrust the hype, a moderate position that no orchestrated campaign could have created.
Drama is interesting to watch, but moderation gets things done. The ancient Greeks gave us myths and epics, but they accomplished more by arguing the toss in the agora.
Reading list
AI as normal technology — Knight First Amendment Institute. The middle position on capability, and why diffusion sets the pace.
15 times to use AI, and 5 not to — One Useful Thing. Ethan Mollick on selective use as a form of judgement.
The messy middle — Kinder Futures. Molly Kinder on the long transition that neither pole wants to plan for.
AI-generated “workslop” is destroying productivity — Harvard Business Review. What using AI for everything costs your colleagues.
3 reasons kids hate AI — Fortune. The refusal at the other pole, and why it is aimed at the sellers.
Americans cool toward AI — Gallup. Adoption and mistrust, measured side by side.
Why AI companies are hiring philosophers — NPR. Benjamin Sutherland on reasoning as a growth industry.