Bayesian Reasoning and Updating Priors

Rosie is a model XB-500 robot. She was hired by the George and Jane Jetson from U-Rent A Maid, and despite being an older model, she’s hardworking and dependable — everything you’d want in a robot maid. She might be a machine, but her artificial intelligence (A.I.) is advanced, so much so that she once went through a bout of depression over never having met her prototype. 

The Jetsons first aired in 1962 — we’ve been thinking about A.I. for a long time. Today, the world’s most renowned scientists, futurists, and mathematicians work effortlessly to make artificial intelligence more robust and valuable. You may not realize it, but you are using A.I. daily in the form of search engine recommended searches, digital assistants like Siri and Alexa, robot vacuums, parking assistants, fraud detection, and more. It’s not quite Rosie, but valuable nonetheless.

In 2011 we consumed five times more information than 25 years earlier. It’s safe to say that factor has increased markedly since then and will continue to climb as technology facilitates faster, denser, and wider-reaching information transmission. 

How are we to handle this information onslaught? 

It’s easy to blame technology for our current condition. Rarely does a week pass that I don’t hear an anti-technology diatribe that blames the world’s problems on the internet, social media, or another modern technology. It’s an easy target, but I don’t know too many people who would willingly transport to pre-industrial society, at least after thinking, even minimally, about it. 

In looking for solutions to technology accelerated problems, we might find the answer (in part) within technology itself.

One of the chief concepts behind machine learning is Bayesian reasoning. In a nutshell, Bayesian logic determines the probability of an event based on the outcomes of previous events. Machine learning requires the rapid extraction of information from large data sets, and Bayes’ theorem allows for predictions based on conditional relationships found in those data sets. 

What does this have to do with humans being bombarded with information? 

If A.I. models are to improve, their priors must be updated— this is how they get “smarter.” New data leads to better predictions, so accepting further information pushes artificial intelligence to new heights. 

The flurry of information from all sides requires increased flexibility of mind — an extra willingness to think, rethink, and question our existing beliefs: updating our priors.

It’s natural when being bombarded to erect impregnable defenses, dig unpassable trenches, and hunker down to avoid the stress of self-examination. 

It’s easy to ignore nuance, double down on priors, and edge closer to the nearest extreme where you’re sure to be comforted by comrades with matching views — shielding your psyche with a force field of confirmation bias. But ignorance isn’t bliss. “The unexamined life is not worth living.”

In his book, Think Again, Organizational Psychologist Adam Grant said:

“Calcified ideologies are tearing American culture apart. Even our great governing document, the U.S. Constitution, allows for amendments. What if we were quicker to make amendments to our elemental constitutions?”

We recognize the need for other people to change, institutions to adapt, and the importance of evolution in artificial intelligence. Still, we struggle to flex our minds enough to consider that we might be wrong. Grant blames the totalitarian ego for our reluctance to rethink. 

The totalitarian ego is an inner narcissistic dictator who controls the flow of information with an iron first, heading off inconvenient truths before they induce upheaval in our minds. Just like outside information can spark a revolution and topple a totalitarian regime, some info would shatter our sense of self. So our inner dictator puts the kibosh on this threatening information before it destroys our long developed identities. 

To learn, we must fight back against complacency and oust our inner dictator — we can do so by updating our priors. 

How to Update Your Priors:

Prioritize values over beliefs

You are not your beliefs. By entangling your identity and beliefs, you create a mental prison that prevents you from intellectually engaging beyond the existing walls. Anyone who disagrees with your views becomes an enemy, and you find yourself constantly mired in conflict and destructive arguments instead of constructive discussions. 

Values, like integrity and freedom, make up your foundation. Those are much less fluid than beliefs. Focus on adhering to your values, but revise your beliefs as you gain experience, knowledge, and wisdom. 

Embrace mental flexibility over consistency

We love consistency. We call people who change their minds flip-floppers and call out their lack of backbone. But intellect requires an openness to new ideas, seeing learning opportunities in disagreements, and remaining open-minded, even about long-held beliefs. Embrace intellectual flexibility — look for opportunities to become less wrong than you were before. 

Avoid echo chambers 

It’s impossible to learn from others if you only interact with those who share your views. While it might be more comfortable to have reaffirming conversations, innovators and thinkers are comfortable with people who broaden their perspectives and provide new and different ones.

Artificial Intelligence will play a more significant role in our world in the coming years and decades. As processing power increases, so too will the ability of machines to tackle human tasks. This wouldn’t be true if models were stagnant; they’d never learn — never improve. 

I don’t think we’ll see general artificial intelligence in my lifetime. We don’t understand the human brain well enough to replicate it, but we can learn from how we teach neural networks. In this instance, the machine student can be the teacher — update your priors.