7,500 Faceless Coders Paid in Bitcoin Built a Hedge Fund’s Brain

[berkshire101]

Ran across this fund while listening to a podcast.  It's an interesting take on investing and leveraging crowd sourcing with machine learning.  How can a value investor compete?!  :o

 

https://www.wired.com/2016/12/7500-faceless-coders-paid-bitcoin-built-hedge-funds-brain/

 

https://techcrunch.com/2016/12/12/numer-ai-is-a-crowdsourced-hedge-fund-for-machine-learning-experts/

 

https://numer.ai/about

[Jurgis]

It's likely not there yet in terms of long term predictions/investing. Most of these models now do short term predictions that may or may not work.

 

I think they will get to long term predictions within 10 years or so. Although there's a lot of short-term/technical/price-based prediction mentality in financial AI/ML. Not many (any?) people are doing fundamentals-based longer term predictions.

[wachtwoord]

Interesting!

[rkbabang]

It's likely not there yet in terms of long term predictions/investing. Most of these models now do short term predictions that may or may not work.

 

I think they will get to long term predictions within 10 years or so. Although there's a lot of short-term/technical/price-based prediction mentality in financial AI/ML. Not many (any?) people are doing fundamentals-based longer term predictions.

 

 

+1.  The problem with machine learning is that the machine will observe the market and learn how the market works right now, but the market isn't static and these hedge funds are themselves a part of the market and the moment they act on the new information they have an effect on it.  It is in a way like quantum mechanics, by making the observation you are effecting the system which you are observing.  As more people use this type of algorithm development methodology the algorithms will stop working and new algorithms will need to be developed which will themselves  soon stop working, and so on ....  It is a race without end.  What if your algorithms have never experienced a market crash and have no way to predict it or at least have never experienced a crash which is caused by whatever it is which causes the next crash?  You may be left vulnerable to a major loss.

 

[WneverLOSE]

I clicked into the post, read a few words and dismissed it as oh another case of smart guys over complicating themselves.

Went back to the forum main page and entered another post, thought to myself, maybe I should give it a read and see what that's about and I'm so happy I did it.

Absolutely fascinating, thanks for posting.

[SharperDingaan]

The great thing about machine learning is the substitution of brute force for brains, ultimately driving the brains (& the ability to react to change) out of the algo altogether. We put people & technology together - for a reason.

 

Imagine I float some great ideas - the algo picks them up & pays me for each of them.

I do this a few times - & enjoy the proceeds via a two week vacation in the islands somewhere.

The beach is boring .. so I build a cloud of variables, that I am pretty sure the algo will pick up (it used my ideas)

I am also pretty sure I know how the algo will react to these variables; so set up accordingly, & release the cloud into the market.

To the algo the cloud is toxic - it goes nuts; and pushes my set-ups deep into the money. Sell.

 

I have just done a mugging, & have gotten away with it ...  simply because I was bored, & the algo had lax human control.

.... and I am one of the more ethical ones!

 

SD