Why generative AI isn't going to change the world overnight
but stand by for substantial changes over the next decade and beyond
I’ve been traveling this week so had time to catch up on a few podcasts. One that stood out was this fairly lengthy chat on The Gradient with Arjun Ramani, global business and economics correspondent at The Economist and Zhengdong Wang, research engineer at Google Deepmind.
Back in June, Arjun and Zhengdong published an article with the then very against trend title, Why transformative artificial intelligence is really, really hard to achieve. The podcast dives into the article and the research that led to it so if you’re keen to know more, you can pick your preferred format. Below are a few of the points that stood out for me.
Historically, fast growth just isn’t that fast
Even in the decades following the previous technological revolutions - steam engines, internal combustion engines, electricity, the internet - the trend growth rate of GDP per capita in the US has never exceeded three percent per year.
Industrial revolution #1 (1750-1830) - steam and railroads
Industrial revolution #2 (1870-1900) - electricity, internal combustion engine, running water, indoor toilets, communications, entertainment, chemicals, petroleum
Industrial revolution #3 (1960-present) - computers, the web, mobile phones
IR #2 was more important than the others and was largely responsible for 80 years of relatively rapid productivity growth between 1890 and 1972. Once the spin-off inventions from IR #2 (airplanes, air conditioning, interstate highways) had run their course, productivity growth during 1972-96 was much slower than before.
In contrast, IR #3 created only a short-lived growth revival between 1996 and 2004.
Many of the original and spin-off inventions of IR #2 could happen only once – urbanization, transportation speed, the freedom of females from the drudgery of carrying tons of water per year, and the role of central heating and air conditioning in achieving a year-round constant temperature.
Robert J Gordon, National Bureau of Economic Research Working Paper 18315
Because … bottlenecks
Why? Because you hit bottlenecks that aren’t removed by the transformative technology of the day. So really rapid change needs an invention so transformative that it doesn’t hit ANY bottlenecks of significance.
I don’t know about you but to me that feels extraordinarily unlikely.
As Arjun and Zhengdong outline, it seems a lot more likely that …
The transformational potential of AI is constrained by its hardest problems
Despite rapid progress in some AI subfields, major technical hurdles remain
Even if technical AI progress continues, social and economic hurdles may limit its impact
Arjun Ramani and Zhengdong Wang, The Gradient
And this point gave me pause. The idea of bottlenecking is very far from new - the authors chart it back to the 1960s - it’s just getting very little airtime.
Perhaps next time I detect a whiff of underpants gnome thinking when it comes to Generative AI, I’ll start quoting 1960s economists.
Online still needs offline
While it feels like we spend a lot of our time online these days, we’re still corporeal beings and a lot of what we do online is either bolstered by or needs to change things in that offline world to be useful (like having my groceries delivered after a week away from home).
It seems highly unlikely to us that growth could greatly accelerate without progress in manipulating the physical world. Many current economic bottlenecks, from housing and healthcare to manufacturing and transportation all have a sizable physical-world component.
Arjun Ramani and Zhengdong Wang, The Gradient
And manipulation of the physical world, i.e through robots, while it is advancing, certainly hasn’t leapt forward along with generative AI. Nor is it obvious how Gen AI could do much to help robotics improve enormously.
Technical change is easier than organisational and social change
If you’ve ever tried to deliver a substantial ‘IT project’ inside a company of more than say 1000 people this one will resonate. The worst mistake you can make (or allow your client to make) is that you only need to change the tech. Adoption is critical and that requires people to change.
I’ve been pondering since January why it is that very hard nosed commercial people are being so enchanted by generative AI that they’re not asking about the price tag for all that magic (my current hunch is that our brains are being seduced by the fact that ‘it talks!’). Maybe that’s also why we’re all overlooking the pain and toil of organisational and societal change too? If people have to change - habits, routines, ways of interacting with each other and the world around us - that is going to take time.
So what?
This one hasn’t been changed by LLMs it would appear.
Accordingly, invest in the hardest problems across innovation and society. Pause before jumping to the most flashy recent development in AI. From technical research challenges currently not in vogue to the puzzles of human relations that have persisted for generations, broad swaths of society will require first-rate human ingenuity to realize the promise of AI.
Arjun Ramani and Zhengdong Wang, The Gradient
Wishing you all a peaceful and productive way in the physical and virtual world of your choice.