Advice, warnings and genetic algorithms

In the Knowledge Project podcast with Adam Robinson there is an interesting observation about warnings.

Human beings always ignore warnings.

It was stated in reference to a book that he had written for Warren Buffet. He shares a load of advice but some of them come with a warning. He states that most people when reading this would ignore the fact that there are warnings associated with them and take them as the truth.

He then goes on to make a very interesting observation:

All advice is a warning. All wisdom is a warning. If I say the best way to boil an egg is this, there’s an implicit warning. If you don’t boil it like that, the egg is not going to be as good as it could have been. If I say the best way to negotiate a contract is to always let them make the first offer. There’s a warning.

It is not something I’d thought about before. When we receive advice we have a choice to take it or not. If we don’t take it there is a risk that we are throwing away an opportunity to lean on someone else’s knowledge and experience. However we might also be missing an opportunity to learn and find out the best way ourselves.

It is similar to the problem of local maxima with genetic algorithms. A genetic algorithm generates a result and evaluates it compared to previous results. If it is the “best” result then that becomes the next starting point. However this can create a problem called local maxima. It does not always find the best result and can get stuck at a lower point.

If you imagine a map grid where the height of every point reflects how good the result is. If you pick a starting point you can then look at all the squares next to you and move to the highest one. Using this approach you can navigate the map and will eventually end up at the top of a hill. The problem comes in that you will not move from the top of that hill as no squares around are higher. However this may not be the highest hill on the map - this is called a local maxima.

The way they solve this with genetic algorithms is through mutations - in our example this would be deciding to head off in a new direction even though it might not be the highest point.

Coming back to the advice and warnings, if everyone always takes a piece of advice and heeds the implicit warning then we could potentially get stuck at a local maxima - we wouldn’t know if there is a better way to do things. If we always ignore the advice then we might find a better way or we may find a much worse way.

Sometimes we may want to ignore advice and just try things. We need to take into account the risks and the impact of being wrong. The advice to not touch a hot stove should probably be followed however the advice on the best way to learn to juggle can probably be safely ignored (although it may take you longer).

Links

Adam Robinson #168

Related Posts

Show seconds in the Windows 11 taskbar

I noticed the other day that my wife’s laptop was showing seconds in the time in the taskbar but my laptop was not; a very handy feature to have.


Read More

Antilibrary - should I finish the book I am reading?

I read a lot of books and once I start reading a book I normally finish it. This article proposes that it may be more valuable to have a bunch of books that are unread. Not sure I completely buy in to the proposition that “Read books are far less valuable than unread ones” however maybe I won’t feel so guilty buying a book and then leaving it unread.


Read More

Blackout by Marc Elsberg

This story documents one scenario that could happen when the nightmare happens of a massive power cut across Europe. It starts with a number of threads and weaves them well together. It follows an Italian as he realizes it is a terrorist attack on the smart power meters. This destabilizes the power grid which takes down most of the grid across Europe. A second attack targets the power plants. This leads to a power cut across Europe that lasts days and the storyline of how they track down the cause and the perpetrators.


Read More