UNU - The artificial hive mind

[tombgrt]

Hi all.

 

 

Anyone heard of UNU before? Found out about the project through reddit.

 

Also see: http://unanimous.ai/what-is-si/

 

This is how it works. One user asks a group of (knowledgeable) people a question on a certain subject. This group can then in real time vote their answer. The difference with a simple poll is that this group of people can see what others are going to vote. So as a member of the group you are able to change your answer on the question in real time depending on what others are thinking. Other than that it is completely anonymous so no input for biases there.

 

Apparently it is able to consistently outperform the individuals that make up the groups (often experts in their field). This immediately reminded me of how the stock market as a whole works where experts are basically making real time decisions based on the buy and sell decisions of other experts. Often also increasing the accuracy of the predictions as the group of experts gets bigger (small vs large caps). It directly shows why - in general - it is silly to try to outperform the market as such markets actually seems to transcend the pure knowledge of the group on a superficial level. It is extremely rare to go beyond that level of understanding of complex situations as an individual.

 

Pretty interesting. Imagine what could be done if they could finetune this process further so that we can organize groups of thousands of experts in specific fields that can answer questions on hot issues in today's society. Would we see much improved accuracy and results? You could basically hand out knowledge, let a group study it and then let them come back together online to do their magic. I'm probably seeing this way too simplistic but still...

 

[tombgrt]

On second thought I was thinking that I might be overstating the potential here. It sounds nice in theory but aren't fund managers after all performing barely below the market after considering fees, taxes and costs? And the stock market is as big as it gets if you are looking for a group of experts.

But then it hit me that these fund managers are buying stocks that are already valued by the market. The market has already given it's stamp of (dis)approval so it's pretty normal for fund managers to do ok as long as they hold enough positions. (I can hear some say "Duuh"..  :D)

 

So basically what they should do to see what the markets alpha over knowledgeable fund managers is, is giving fund managers a blind test. Make them value companies without looking at the current stock price and see how they perform over time (don't ask me the specifics). Wouldn't that give us an indication of what a market is capable of? In any field? Or is it not transmittable to other fields?

 

 

I know the above is likely obvious for many here but my thoughts were slow to catch on tonight, it's almost 3:00 AM here after all.

[Jurgis]

How is this different from prediction markets or from Tetlock's Superforecasting and http://goodjudgment.com/index.html ?

 

BTW, there is investing UNU: http://go.unu.ai/unums/601

Most of the questions are crap. Some answers might be interesting. Some are contradictory (across multiple questions related to the same company): Apple is losing cool, won't outperform, don't buy, but don't sell either...

[tombgrt]

How is this different from prediction markets or from Tetlock's Superforecasting and http://goodjudgment.com/index.html ?

 

BTW, there is investing UNU: http://go.unu.ai/unums/601

Most of the questions are crap. Some answers might be interesting. Some are contradictory (across multiple questions related to the same company): Apple is losing cool, won't outperform, don't buy, but don't sell either...

 

Obviously the public that populates the website isn't generally going to be made up out of experts, especially when you just did an AMA on Reddit. A public website where anyone with an internet connection can claim to know something about a subject isn't going to yield great results. Especially if the questions aren't properly formulated and thought out. I think it has potential no matter how poor these public projects might seem now.

 

I don't know how it's different from prediction markets, Tetlock or Good Judgement as I haven't heard from them yet. :) Just thought the subject was interesting in and of itself, regardless of what the used platform is.