The wisdom of crowds - true or false?

[Jon]

There is a book called "The Wisdom of Crowds." I was tempted to get it, but my learning list reaches for the sky so I have no chance of getting to it.

Are crowds wise or full of folly? I mean you could get some AI that is an outlier with superior perception that can out-think anybody. But the crowd may have more people but with less individual wit. You can replace the AI super-intelligence with a super-intelligent wet-ware human brain if you like. I don't really differentiate between wet-ware and hardware. Just a personal thing!

So which is wiser the crowd or the AI bot?

 

[The_Doc_Man]

The AI bot has no wisdom that it didn't get from another source. For instance, IBM's "Watson" project is an AI that does some truly amazing things. If you give it medical info, it makes diagnoses to the same level of accuracy as a live, experienced physician. In particular, I hear it is good with cancer detection. But that is because they distilled 500+ specialists into it.

There was a TV show on that topic last year, a melodrama about someone who used crowd feedback to solve crimes. In essence, the wisdom of the crowd is that everyone has different experiences and if you can somehow distill that down, you can determine probabilities. When looking for leads, you have a wider group of potential witnesses. But as we know, if there is something that is counter-intuitive, crowds don't necessarily do any better than an intelligent, observant, and logical person at problem solving. And even if someone saw something, they might fear getting involved if the thing they saw seemed exceptionally violent.

https://www.imdb.com/title/tt6522758/

 

[Jon]

We have no wisdom that didn't come from another source either, hence we go to school, learn by experience etc. You could argue that the collective wisdom, or the whispers of ten thousand generations in your DNA is your own source, but that DNA came from your parents!

I know the police have started to use AI bots to guestimate where the next crimes will take place, making them a wise sage. Their source is probably a collection of dimensions that are data crunched to provide correlations. It reminds me a bit of that Tom Cruise movie where they preemptively arrest the person before they commit the crime. Perhaps that is where we are heading! The cost to society is too great to wait for the crime to be committed, if the bot says there is a 0.77 correlation that a crime is about to be committed. A bit like the Iraq war and the preemptive action there.

 

[The_Doc_Man]

The progress IBM has made with Watson is a matter of collecting dimensions and generating statistics to develop correlation relationships. They interviewed oncologists who made a diagnosis to determine WHY they made a diagnosis, to try to winnow out the source of that "gut feeling" that drives so many decisions. And it does so mostly because we internally develop our own correlations so that they operate almost at an unconscious level. Their research into "digitizing the gut feelings" simply required the specialists to quantify stuff at a deeper level than they had ever done before. But the results were indeed interesting. Unfortunately, the results aren't as good as I thought.

I have to retract the "same level of accuracy" statement from my post #2. Turns out I had read some advertising about the research. But advertising is not unbiased reporting. The link below shows the REAL status of the Watson Oncology project.

https://www.darkdaily.com/ibms-wats...tient-outcomes-or-accurately-diagnose-cancer/