Making predictions and the challenge of expertise
New year, new predictions. How do I think about prediction and forecasting? The most important part is that you do it. Do it consistently. Not once a year when everyone is doing it, but on a regular basis. My specific thoughts:
- Always write it out. Writing your thinking out makes it clearer, stronger, better.
- After writing it out, share it. Test it. Refine it. List out the ways you could be wrong. Review it after the fact. And then learn. But the only way to do it is by doing the work.
- Most predictions you read (particularly year-end) are sales and marketing. They are people selling or marketing a position they often have already put their money, time, and/or career behind. View them through that lens.
- Predictions are only as valuable as what you are willing to do to act on them. How you place a bet to capture value or upside is the hardest part and the most vital to get right.
- Spend time deepening your knowledge but also examine the bias from your expertise and past experience. What do you do when you know a sector inside and out and you are looking for something that might change?
This point on expertise is where I want to go deeper.
A couple of weeks ago, I attended a deep tech conference, which was full of insight and learning. The final keynote was Peter Diamandis who founded the Singularity University. His presentation, which you can find here, maintained the thesis that everything is changing at an exponential rate, and you have to go all-in on Artificial Intelligence. He went through examples, e.g., there should be energy cheap and easy from the sun; everything will be connected; and, gene therapies will be approved and utilized in mass. Near the end, he proclaimed that technology converts scarcity into abundance. Which seems to make logical sense. Of course, the level deeper is scarcity for who? An abundance of what?
At this point, I'm buying the narrative. Eating it all up. And he keeps talking about all the things he's worked on and then...bam. He lands on something I know well.
On slide 77 out of 90, we get to the X-prize in 'Global Learning'. See below. It's, like, hold up.
What is going on here? I spent a good several years as the head of a venture fund that invested in emerging market education and EdTech companies. We had four portfolio companies in sub-Saharan Africa, and I had looked at over 100 'solve education in Africa' companies, including a few that had participated in the X-prize.
Let's break down what is happening on the slide. There was $15M of prize money. Organizations spent an estimated $300M of resources to address the challenge. The goal was to improve learning for 2,500 students in rural Tanzania.
Some basic math tells us the ROI was not great. $300M divided by 2,500 means that $120,000 was spent per student in Tanzania, where the average GDP per capita is $985. So there's that. You could have sent each kids to a US boarding school... for several years.
Second, the stipulation that "1 House of Tablet Use = Full-time School." Well, they don't give a lot of details about what that means, but I am 95% sure that it's mostly a reflection on the fact that children in school in rural Tanzania don't learn very much. So if you don't learn much in full-time school, it's not too hard to not learn much during 60 minutes on an iPad.
As much as people have tried to make 'if you give them all laptops and tablets with good content, they will learn' happen, it never comes to fruition.
It was a great reminder of the level of belief that I should give to anything else from this presentation about the areas where I knew much, much less.
Joel Greenblatt had a similar story at the beginning of the book:
When I was fifteen, the only gambling establishment that would let me sneak in was the Hollywood Dog Track. This was a great thing because, during my illicit visit, I discovered a sur-fire route to big greyhound riches. In the third race, there was a dog who had run each of his previous six races in only thirty-two seconds. The odds on this dog - Lucky - were 99-1. None of the dogs up against Lucky in the third race had managed a time better than forty-four seconds in any previous race.
Of course, I bet what passed for a small fortune at the time on Lucky to WIN. If all those fools who bet on the other dogs wanted to give me their money, so be it. However, as Lucky struggled down the home stretch in last place, my opinion of the other gamblers slowly began to change.
This was Lucky's first race at a longer distance. Apparently, as everyone else already knew, Lucky's spectacularly fast times in his previous races were achieved at much shorter distances. All the other dogs were experienced long-distance runners.
I learned a valuable lesson. Without a basic level of knowledge and understanding, you can't tell a great investment from a real dog.
The tough part is when you know a sector super well and you are looking for the things that will change the game, alter the outcome, make it different this time. Maybe it will be different with AI?