Machine learning is innately conservative and wants you to either act like everyone else, or never change

Originally published at:


“Can you sit on my face” doesn’t sound very conservative to me!


Is anyone surprised by this? After all, matrix multiplication was invented by a white conservative man, (given his support of the previous King, Charles X).


Machine learning / AI is no different than any other program: garbage in, garbage out. If you create an algorithm to predict predict the future based on the past, it’s going to echo the past. That behavior isn’t limited to machines. If you want an algorithm that doesn’t echo the past, those are even easier to make.

Turns out that a method I came up with to design medical products is considered AI by the FDA. I feed it data from testing and use it to generate new designs that should perform better. I’m not asking my algorithm to predict the future per se. It’s still conservative in the sense it will only make predictions in the range of n-dimensions I let it. And that’s a good thing. But it’s not conservative in that it never echos what was fed into it.


Maybe it is Conservative with a large “C”. I guess we could ask David William Donald “Pigfucker” Cameron.

David Cameron


Everyone is conservative about what he knows best.

Conquest’s First Laws of Politics

1 Like

His son also supported Charles ÷ and extended his dad’s work to invent matrix division.


Conservationism is also an indicator of simulated intelligence in humans.

1 Like

Can’t tell if serious or auto-correct…



1 Like

I write machine learning models to help with the diagnostics of eye disease. I guess this is conservative in the sense that I’d quite like to conserve people’s vision, if at all possible.

Mea culpa I guess.


Machine learning is going to work well for some problem domains and not so well for others. Diagnosis is (hopefully!) one of the areas where it will work well. Whether or not it works well for most or all types of diagnosis is another issue, but I could see it being a big help with your field and others (identification of skin cancer comes to mind) right off the bat.

I am shall we say less enthused about machine learning in other areas. For instance, recently Microsoft sent me an invitation for MyAnalytics, which analyzes my email and calendar usage to allegedly improve my productivity. I turned that shit off immediately.

Youtube and Amazon also presumably use machine learning for recommendations, but they’re both terrible at it (at least in my experience). Amazon search in particular has gone to hell, although the proliferation of third party Amazon sellers could also be a factor here.

1 Like

Machine learning provides the opposite of discovery. You’d think they’d learn. Sometimes you actually do want to do the same thing again. Sometimes you want something new. There are algorithmic ways to do this. An OKCupid study suggested looking for bimodal rankings, but it should be obvious that if you have just bought a bunch of shirts, odds are the last thing you want to buy is another bunch of shirts.


Uninstalled Google keyboard on my phone once it was trying to get me select stuff like “Pepe.” Etc. At the time I thought this was hilarious but eventually I just felt insulted.

1 Like

It discerns interest, but does not take into account completion. It cannot tell if you are browsing shirts, have shirts in your cart or have actually purchased shirts recently. That could be an oppo for some engineer to build a better feedback widget that indicates recent purchases are to be demoted in the display queue.

Since IOS introduced a swipe keyboard, I’m finding it even harder to use profanity in my texts.


Yep. Machine Learning seems to be all about finding the rut for you to get stuck in.

This topic was automatically closed after 5 days. New replies are no longer allowed.