Bayesian inference

You can do things to get kind of geeky on it.

Like you can build Bayesian models, so they auto update.

So you say, well, I think these are the things that will happen.

These were the actual outcomes.

You're illiterate, and so are your followers, so you can show them that the paper says what I'm saying it does, and just pretend that it doesn't, and they'll believe you because they're stupid.

One potentially attractive avenue is to use Bayesian modeling to estimate a nominated date or event.

For this analysis, the completion date of construction was deemed the most suitable parameter.

Unlike the beginning or midpoints, this boundary was more likely to be static and unaffected by inbuilt age.

In real terms, it should represent the point at which short-lived, non-stockpiled or recycled material was included on the last day of construction.

What they did was use their Bayesian modeling and inserted as prior information the completion date of these edifices, the completion date, which is the latest possible date.

So already they're skewing it as far possible as they can away from those older dates.

Because Bayesian methodology, you're talking about a freaking mathematical equation.

And I think...

What we need to bring back a bit more, there's a concept called Bayesian reasoning from Thomas Bayes a few hundred years ago, which is actually what informs a lot of the scientific discoveries that have happened over the last couple of hundred years.

And it's the idea that you walk into a situation with a prior.

So this is what I believe to be true now.

And, you know, and the problem that you kept running into in your career, you know, with this, with the imaging data, the ClinQuery, the rules is still the data.

Do we have the right data? Do we have enough data? And is the data the right shape? The second theme that I heard was that every decade seemed to have its own version of what AI is, right? So we had Bayesian rules in 75, and we had in the 80s, we had expert system type rules that you were working in.

And today we have large language models and neuro-symbolic approaches, right? I'm sure we'll have something else in the 2030s.

I think it's not a reason to give up.

It's fun to be able to collaborate with people that have the same type of values.

Yeah, the open-mindedness is a really key characteristic because we like to talk about things within our investment community, having a, or I like to say, a Bayesian process, and the saying is, strong beliefs weekly held.

Like with our investment philosophy, we want to stick to our investment philosophy, but as new information comes in, new data comes in, we want to be able to adapt that to see are our beliefs still solid or do we need to modify those to handle new information.

And our clients that operate like that from a financial planning perspective as well, they're curious, they're adaptable, they're willing to explore new ideas.