Oh, the horror. I have suffered through those too. Thank heaven they seem to be rather a past fad nowadays… but ‘consultants’ made a lot of money running them in the day.
I doubt if any useful ideas ever came out of them. As Kipling said: All work is one man’s work. Or at least a small group of bright people who bounce off each other…
We’re still talking about work, right?
I can’t find the quote, but years ago I read a quote by an economist that went something like
economics is the field of starting with raw materials and changing them until they are useful finished products
Something like that. Basically you start with the complete raw materials like minerals in the earth, and the end result if you have computers, cars, food, etc.
I guess productivity could be seen as ‘how effective are we at goal oriented behavior, problem solving, overcoming obstacles, combatting entropy, and converting useless raw materials into useful finished products, per unit of effort/capital invested’.
But again, this also ties back into what I mentioned earlier that we always go for the easiest problems first. once those are solved, the next layer of problems is more complex and the payoff is smaller. Then the next layer is even harder with a smaller payoff.
This quote about John Von Neumann is also relevant
He demonstrated theorems in 5 minutes, resolved master’s and doctoral problems in a conversation, and had more knowledge than specialized people on all kinds of subjects
Especially the part about resolving masters and doctoral problems in a conversation. The amount of effort it took a graduate student to solve those problems is far far higher than the amount of effort it took Von Neumann to solve them.
I don’t know if there is any objective way to measure how much effort it takes to solve a problem, esp considering that humans have different capabilities of problem solving.
Like if you measured cognitive productivity in hours of thought for example, Von Neumann could accomplish far more in 12 hours of thought than most people could accomplish in 120 hours of thought.
With AI we have the cost of tokens. Like a million input tokens may cost $0.10 and a million output tokens may cost $0.40. But as AI advances it’ll be better at problem solving, and the cost of tokens will continue to go down. In a couple years AI will be better at problem solving and it may only cost $0.01 per million input tokens and $0.05 per million output tokens.
I have been listening to his podcasts for some time and just read his book Slow Productivity. excellent info.
I strongly recommend Newport’s Deep Work if you are interested in his stuff. He’s a very good writer.
Lately his podcast is split between productivity content and “AI reality check” content. He describes his work as living in the intersection between technology and productivity. He’s no Luddite, but he’s been highly critical of the fearmongering in the media and tech circles.
He’d definitely have some interesting things to say on this topic.
I listened to one of the his youtube videos on AI. I don’t know if it’s confirmation bias but he seemed to be saying pretty much what I’ve been experiencing with respect to the churn of AI content being produced in the workplace. Most of it superficial and in need of editing. Summarizing it with AI over and over doesn’t make it any better. I end up ignoring pretty much every email that I recognize as being AI generated. If a person can’t be arsed to write an email in their own words, I can’t be arsed to read it, and they aren’t likely expecting a response anyway.
My old advisor (a theoretical physicist) said that the twin curse of being a theorist is that, first, you can never stop working, because whenever you’re thinking, you’re working; and second, that nobody can ever tell when you’re working.
How “productive” was the guy who invented fire? What about the guy who invented the Pet Rock? Follow up question: Which one got paid more for their contributions to humanity?
Frankly I don’t think “productive” is a useful question when discussing this topic. Needless to say, measuring it is even more remote from reality. Good luck explaining that to the bean counters, though.
You Idiot, said the Elders. Do you have any idea how dangerous that is?
They put him to death. But it was too late: a bright lad watching from the shadows had memorised the procedure…
Mod note: sure, I realise this is quite off topic. Couldn’t resist though: a new thread perhaps?
To return to this - the counter is that if such was a true general principle then the reality of “productivity” per capita over time would be a gradual decrease. Meanwhile what we see is closer to an exponential curve.
I suspect what actually happens is the point of diminishing returns hits and incentivizes the disruptive idea that changes the effort return equation again. Overall productivity being a relatively smooth exponential curve but at a granular level consisting of subcurves that are more staircase like, with more frequent steeper steps?
Human knowledge is a multidimensional hypersphere. As it expands the surface area increases.
It is not true that all the low hanging fruit was back in Neanderthal days. New low hanging fruit in new areas is constantly uncovered as the sphere expands.
Now and again, yes, one particular area gets stymied, say Quantum + Relativity unification. Lotta effort, no results. Like mining blindly through a mountain, you’re gonna hit harder and softer rock and won’t know any place’s hardness until first you uncover it, then you whack it and see.
Meanwhile, while that one area is stymied, others are constantly being opened up.
The argument is that the reason it feels like an exponential curve is that the amount of financial capital and human capital devoted to solving problems has grown dramatically.
The paper I linked earlier doesn’t just talk about agricultural yields. It also talks about things like moores law. You need more and more researchers each decade to keep the pace of Moore’s law going.
What you’re describing sounds like S curves, which claims a new technology takes of slow, then dramatically improves, then levels off. If you have a bunch of S curves all layered on top of each other it looks like exponential growth.
Slogged through the paper as best I could.
Comments:
Probably most important relevant to the subject of the thread - I appreciate the difficulty expressed about how to measure cognitive productivity. They reject just counting the number of ideas and instead look to outcomes. Crop yields, life expectancy, so on. And I get that but ideas can be novel and important, even transformative for understanding, and have application lag far behind and in disparate fields. Einsten’s relatively few ideas, for example … Which they point out: the impact is still exponential because ideas are “non rival” -
ideas are different from all other goods in that they can be used simultaneously by any number of people. Exponential growth in research leads to exponential growth in At . And because of nonrivalry, this leads to exponential growth in per capita income.
And I appreciate the point that societally the increase continues to large degree because the pool of educated members of the world population has increased so much.
It also has to made clear that they are specifically asking whether exponential growth of cognitive productivity can be maintained:
The evidence presented in this paper concerns the extent to which a constant level of research effort can generate constant exponential growth, either in the economy as a whole or within relatively narrow categories, such as a firm or a seed type or a health condition.
But the biggest point is raised in a follow up paper.
Vis-à-vis per capita research productivity we may hitting a new exponential phase:
In the model, AI automates a growing share of cognitive work, raising productivity …
The continuation of that sentence is displacing cognitive workers, but maybe better thought as potentially multiplying their impacts? Instead of throwing exponentially more educated people at idea production, AI develops to where the same number maintains exponential idea growth output?
AI is a transformative idea (okay set of ideas), even if how it transforms is an open and potentially existential question. Ideas have been like that. Gradual improvement, then the click as the long slow work hits some critical point. Thinking about life expectancies as one of their examples: the gradual increase topping off, receding a bit even before recovery, may just be preamble to a massive change as various lines of research in longevity hit an applicability phase in the next decades (very much a real possibility). Then jump up to a new norm.