Artificial Intelligence and the stock market

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
masonstone
Posts: 659
Joined: Thu Feb 01, 2018 3:01 pm

Artificial Intelligence and the stock market

Post by masonstone »

With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
Conch55
Posts: 507
Joined: Tue Mar 01, 2016 10:03 am

Re: Artificial Intelligence and the stock market

Post by Conch55 »

Isn't that what robo-advisors do currently?
User avatar
djpeteski
Posts: 1078
Joined: Fri Mar 31, 2017 9:07 am

Re: Artificial Intelligence and the stock market

Post by djpeteski »

IMO it already has. Heck a stock picking agent was part of my master's thesis that was written 13 years ago. That was one guy, working for about a year that did a decent job of picking stocks. Imagine what 20 guys, much smarter than me, could do working over 15 years could do.

While my data is only empirical, it seems to be the market has become far more "skittish" in the past 5 years or so though. Swings seem to be greater in one direction and continue throughout the day. Like we might have a down market three days in a row, then the forth day will start a rally that lasts for 3 days. These all seem to be in the 2-3% range which seems like a lot to me.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Artificial Intelligence and the stock market

Post by willthrill81 »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
I don't know and neither does anyone else. Such conjecture is, at best, pointless.
The Sensible Steward
User avatar
David Jay
Posts: 14586
Joined: Mon Mar 30, 2015 5:54 am
Location: Michigan

Re: Artificial Intelligence and the stock market

Post by David Jay »

It may separate the men from the boys, figuratively speaking. Only a few individuals with true talent will be valuable, most of the Wall Street herd could be replaced with AI.

The consequence is that there is a real potential Wall Street firms to reduce costs.
It's not an engineering problem - Hersh Shefrin | To get the "risk premium", you really do have to take the risk - nisiprius
User avatar
mhc
Posts: 5257
Joined: Mon Apr 04, 2011 10:18 pm
Location: NoCo

Re: Artificial Intelligence and the stock market

Post by mhc »

I read your question differently from how the others have responded.

If AI is an advancement, then I would think it would help the stock market. Businesses will become more efficient.
I know this is a simplistic view, but I think AI should be good for the market.
52% TSM, 23% TISM, 24.5% TBM, 0.5% cash
DonIce
Posts: 1453
Joined: Thu Feb 21, 2019 5:44 pm

Re: Artificial Intelligence and the stock market

Post by DonIce »

mhc wrote: Wed Oct 09, 2019 2:19 pm I read your question differently from how the others have responded.

If AI is an advancement, then I would think it would help the stock market. Businesses will become more efficient.
I know this is a simplistic view, but I think AI should be good for the market.
I agree but I think the expected pace of development in AI and the resultant benefit to business is already priced in. Current valuations, especially of companies involved in AI and other hot fields, suggest that the market expects a rapid pace of technological progress leading to substantial profit growth. If anything, the reality of AI will be hard pressed to keep up with the highly hyped expectations.
Stormbringer
Posts: 1207
Joined: Sun Jun 14, 2015 7:07 am

Re: Artificial Intelligence and the stock market

Post by Stormbringer »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
Volatility. Currently, 70% of trades are algorithmic. As machines increasingly trade with machines, the market can rapidly swing one way or another in response to data and news. Traders are locating their servers at or near the stock exchanges to shave milliseconds off their time to react. It wouldn't surprise me to see more mysterious events where the market drops 10% in a day due to how the algos respond.
“The greatest shortcoming of the human race is our inability to understand the exponential function.” - Albert Allen Bartlett
TomCat96
Posts: 1040
Joined: Sun Oct 18, 2015 12:18 pm

Re: Artificial Intelligence and the stock market

Post by TomCat96 »

I think one thing that must of us lack here is a "feel" for the power of AI.

Artificial Intelligence is some abstraction, but mentioning as the latest technological craze doesn't give one a good intuition of how much power there is.

This is a great video for those of you who play chess.
https://www.youtube.com/watch?v=lFXJWPhDsSY
I don't personally play chess but I do play Chinese chess, and I can tell you what ELO scores meant with respect to what Ive experienced. ELO is a rating system for how strong of a chess player you are.

A friend who has just learned how to play chess would probably have an ELO of around 1200. At 1400, you are a bad player with competency, but you are still strong enough to beat random friends who want to play you from time to time. At 1600, you should be able to beat your social group in Chess with ease unless they of course play chess online: your intelligent friend you debate stocks with, your friends uncle who is supposed to be "pretty good", the one guy who was in chess club in high school for a few years. At 2000 ELO, you're very strong. If you went to a university, you might be the strongest player on campus.

Apparently at 2300-2400 you are a master.
At 2500-2700 you are a grandmaster
If you are higher than 2700 you are a viable contender for the world championship.

So, an 800 point ELO difference would feel like, the strongest player in college against someone who just learned the rules of chess.
Magnus Carlen, world champion has a rating of ~2870. He crushes grandmasters with extreme ease online.

Stockfish 9 is a computer program. It is so good that it's elo cannot be measured by human players. In other words, stockfish 9 is so far ahead of human players we do not have the power to determine how good it is. We do so only by measuring against other computer programs.

The chess fanatics on Reddit estimated Stockfish's elo at about 3475. I'm not a chess player. I am inclined to agree. That would mean Stockfish 9 playing the world championship contenders, if I were to anthropomorphosize it, would be like your college chess champion playing a 6th grader who has no special chess skills.

In this video, Alpha Zero, crushes Stockfish so decisively, it doesn't even look like Chess anymore. Those of you who play will appreciate the sheer domination of this game.

https://www.youtube.com/watch?v=lFXJWPhDsSY

Alpha zero's record against stockfish is something like 28 wins, 72 draws, 0 losses.

Alpha zero's ELO is so high again, that I would say we don't really have the programs to gauge it's strength.
That's where we are in AI right now as an intuition.

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
User avatar
whodidntante
Posts: 13114
Joined: Thu Jan 21, 2016 10:11 pm
Location: outside the echo chamber

Re: Artificial Intelligence and the stock market

Post by whodidntante »

At this point, AI has probably affected the life of every American. I guess that'll do until the real explosion shows up.
rich126
Posts: 4475
Joined: Thu Mar 01, 2018 3:56 pm

Re: Artificial Intelligence and the stock market

Post by rich126 »

At least 20 years ago I know people were trying to use neural net software to predict the stock market. We were using the software in other ways.
----------------------------- | If you think something is important and it doesn't involve the health of someone, think again. Life goes too fast, enjoy it and be nice.
glorat
Posts: 1041
Joined: Thu Apr 18, 2019 2:17 am

Re: Artificial Intelligence and the stock market

Post by glorat »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
Coming? In the context of stock markets it arrived in a big way a long time ago. You could say it was and still is an early adopter of innovation
khh
Posts: 343
Joined: Sat Dec 27, 2008 9:31 pm

Re: Artificial Intelligence and the stock market

Post by khh »

TomCat96 wrote: Wed Oct 09, 2019 9:16 pm I think one thing that must of us lack here is a "feel" for the power of AI.

Artificial Intelligence is some abstraction, but mentioning as the latest technological craze doesn't give one a good intuition of how much power there is.

This is a great video for those of you who play chess.
https://www.youtube.com/watch?v=lFXJWPhDsSY
That was fascinating. I have a chess app that I loaded onto my Fire tablet. You can adjust the skill level from 1 to 25. I was wondering what the elo of a level 25 would be.
sid hartha
Posts: 442
Joined: Fri Dec 09, 2016 1:17 pm

Re: Artificial Intelligence and the stock market

Post by sid hartha »

The AI's will decide to go all passive index funds citing better things to do with their time than trade.
ScaledWheel
Posts: 283
Joined: Sat Jul 23, 2011 8:06 pm

Re: Artificial Intelligence and the stock market

Post by ScaledWheel »

Time for me to put on my "crotchety old man" hat (I'm in my early 30's). When I started grad school, what is now being called AI was starting to be called machine learning (ML), and before that was called applied statistics. There is nothing magical about AI, unless we are actually talking about general artificial intelligence, but I am much less bullish about general AI than a lot of people.

Hedge funds and investment shops are already hiring for "AI" researches (deep learning, neural networks, probabilistic graphical models, etc.). In my opinion, most of the benefits that you'll see from "AI" are going to be in increased efficiency and new product lines adopted by public companies, and many more new public companies with AI/ML as core to their business strategy. I don't think it will have a dramatic effect on trading or other financial functions of the stock market, as these techniques are already being used for trading and finding information in the market.
YogurtRunner
Posts: 12
Joined: Mon Sep 30, 2019 8:15 pm

Re: Artificial Intelligence and the stock market

Post by YogurtRunner »

I work for a major Blue Chip Dow component company. They have hired some of the best minds available to tinker with ML and AI for them. I attend the in-house tech conferences all the time. Thus far, I can assure you, the AI stuff here is so far 100% bs and I am embarrassed for the "great minds" who have so far come up with no actual working solutions for any business problem we have. I can't go into too much detail for risk of exposing my company and role, but let me be very clear: The tech they demo is all College and High School level toys like little self-driving autonomous vehicles and real-time video appearance changers (e.g. they had a video stream going live that made everything look like a Van Gogh painting. It was all very interesting, but they never seem to tie anything back to the business at hand. They are paying these folks big time money to work here, probably more than double what I make and they look like a graduate school department rather than a business unit tasked with adding value. The most compelling system I have seen was one that automatically approved certain kinds of expenses using an untrained neural net. The thing is, I could write a rules engines to accomplish the same thing--it didn't require some esoteric neural anything because a simple statistical approach would have been fine. That may fall under the broad category of ML, but it isn't that impressive, honestly.

I can't fully speak to why someone's AI stock picking tool did well or how all these algorithmic traders succeed. I'm highly interested in this subject and I visit algo trading forums and other places were such matters are discussed but 99% of the people there are neophytes. I think books like "Fooled by Randomness" are useful for debunking some of what we observe. Anyone can make money in a bull market, you can be completely stupid and make money when everything is headed to the moon. I'm keeping an open mind that someone has cracked the code, but I wonder whether insider knowledge weighs heavily in creating the "buy" or "sell" signal for successful algos.
ScaledWheel
Posts: 283
Joined: Sat Jul 23, 2011 8:06 pm

Re: Artificial Intelligence and the stock market

Post by ScaledWheel »

YogurtRunner wrote: Thu Oct 10, 2019 8:04 am I work for a major Blue Chip Dow component company. They have hired some of the best minds available to tinker with ML and AI for them. I attend the in-house tech conferences all the time. Thus far, I can assure you, the AI stuff here is so far 100% bs and I am embarrassed for the "great minds" who have so far come up with no actual working solutions for any business problem we have.
A big problem here is upper management are so scared about "missing out" on ML/AI so they hire these people out of grad school without thinking to ask whether or not they have a business problem that can be solved with the technology. Most "deep learning" research has been based on image problems or manipulations; if your company doesn't have computer vision problems to solve, it's unlikely to be of great help. There is a lot of new work being done on natural language processing which has greater potential for being applied to general business problems.
User avatar
JoMoney
Posts: 16260
Joined: Tue Jul 23, 2013 5:31 am

Re: Artificial Intelligence and the stock market

Post by JoMoney »

The key word in "Artificial Intelligence" is "ARTIFICIAL". It's an algorithm made by people, it may be adaptive, but it's still just a formulaic trading system that's trying to beat other formulaic trading systems. Trading is still a zero sum game. The big change in recent years is that computers and networks are much faster now then they were in the past. For people who expect to garner something from there trading (as opposed to buying assets for their intrinsic investment value), speed is of the essence, and a human being isn't going to get any speed advantage over a computer (at least not while the power is on and comm lines are up :) . )
The stock crash back in 1987 was blamed on "program traders", maybe we'll have more incidents that hit circuit-breakers in one form or another.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Artificial Intelligence and the stock market

Post by randomguy »

TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
beehivehave
Posts: 466
Joined: Thu Aug 08, 2019 1:21 pm

Re: Artificial Intelligence and the stock market

Post by beehivehave »

Unlike chess and other closed system games, the stocks depend on many exogenous factors in the real world that are unpredictable - management, consumer sentiment, politics, wars, climate, etc. Worse, their longer term effects are also unpredictable. While AI may be useful in limited circumstances for the short term, it must be able to account for those factors as well to "solve" the market. Granted, under ordinary circumstances it may be able to give a small edge and that is all you need to beat the market.
User avatar
greg24
Posts: 4511
Joined: Tue Feb 20, 2007 9:34 am

Re: Artificial Intelligence and the stock market

Post by greg24 »

masonstone wrote: Wed Oct 09, 2019 11:22 amWith the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
As crazy as it sounds, I think it will increase Keynes' "animal spirits".
rj342
Posts: 726
Joined: Tue Apr 30, 2019 12:21 pm

Re: Artificial Intelligence and the stock market

Post by rj342 »

randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
adam1712
Posts: 566
Joined: Fri Jun 01, 2007 5:21 pm

Re: Artificial Intelligence and the stock market

Post by adam1712 »

Impossible to know.

I think it does have large potential for the stock market and the general economy. The effects in the small domain of trading I expect to be minimal. If I had to guess, AI and automation will follow a similar trajectory to the internet. A potential bubble like 2000 but slowly and surely it realizes its potential reducing expenses, making workers more productive, and reducing labor demand at a lot of companies. Over the last 30 years, I'd say it's been long enough to say there appears to be a fundamental rise in valuations like P/E compared to the 50 years before that which I think can partly be explained by the internet/tech. The same could happen with AI/automation. It's one reason I never plan to base my investments on valuations. There's a not insignificant chance that even though valuations right now may seem high, there may never be a better time to be broadly buying all the companies that are about to be more valuable because of their AI and robots. Could be completely wrong too but I don't want to take the chance of missing out.
shorvath
Posts: 75
Joined: Tue Aug 13, 2013 3:20 pm

Re: Artificial Intelligence and the stock market

Post by shorvath »

I think a key issue with the stock market is that, compared to other fields using ML/AI there is actually very little data available. There is no capacity for replication. We can analyze the 3D structure of a building from thousands of pictures each taken by a different camera, we can study genetics through hundreds of experiments on cloned mice or flies, but a publicly traded company has a single price at a given time.

There are ~3000 companies in the US market. In ML, 3000 cases is an okay (not great) amount of training data, but not when each case can be fundamentally different from the others.
Drunken Roboticist
garlandwhizzer
Posts: 3565
Joined: Fri Aug 06, 2010 3:42 pm

Re: Artificial Intelligence and the stock market

Post by garlandwhizzer »

AI, like all innovations from tech, will be a powerful engine for creative destruction. Hard to know exactly what financially is going to get destroyed and what created. My guess is that a lot of high paying jobs in the financial industry which has already started losing to index funds will be on the destruction side, speeding its decline. On the other hand those who create and refine AI programs will benefit on the creation side. Automation in the auto manufacturing industry drove jobs from production lines to high tech engineering cubicles. I suspect Wall Street will take a hit from AI and programmers in Silicon Valley will get yet another boost. IMO this change is already underway and will likely happen more drastically and more quickly than many expect. I believe the investment market will end up dominated by only two categories, AI driven funds, factor and otherwise, and cap-weighted index funds which instantly reflect AI's buy and sell insights and will have the advantage of lower costs as markets that get ever more efficient from AI. More efficiency means less available alpha which may become an endangered species with AI. Like all predictions this one can be wrong, and even if it turns out to be correct I don't believe it's actionable at this time.

Garland Whizzer
TN_Boy
Posts: 4134
Joined: Sat Jan 17, 2009 11:51 am

Re: Artificial Intelligence and the stock market

Post by TN_Boy »

TomCat96 wrote: Wed Oct 09, 2019 9:16 pm I think one thing that must of us lack here is a "feel" for the power of AI.

Artificial Intelligence is some abstraction, but mentioning as the latest technological craze doesn't give one a good intuition of how much power there is.

This is a great video for those of you who play chess.
https://www.youtube.com/watch?v=lFXJWPhDsSY
I don't personally play chess but I do play Chinese chess, and I can tell you what ELO scores meant with respect to what Ive experienced. ELO is a rating system for how strong of a chess player you are.

A friend who has just learned how to play chess would probably have an ELO of around 1200. At 1400, you are a bad player with competency, but you are still strong enough to beat random friends who want to play you from time to time. At 1600, you should be able to beat your social group in Chess with ease unless they of course play chess online: your intelligent friend you debate stocks with, your friends uncle who is supposed to be "pretty good", the one guy who was in chess club in high school for a few years. At 2000 ELO, you're very strong. If you went to a university, you might be the strongest player on campus.

Apparently at 2300-2400 you are a master.
At 2500-2700 you are a grandmaster
If you are higher than 2700 you are a viable contender for the world championship.

So, an 800 point ELO difference would feel like, the strongest player in college against someone who just learned the rules of chess.
Magnus Carlen, world champion has a rating of ~2870. He crushes grandmasters with extreme ease online.

Stockfish 9 is a computer program. It is so good that it's elo cannot be measured by human players. In other words, stockfish 9 is so far ahead of human players we do not have the power to determine how good it is. We do so only by measuring against other computer programs.

The chess fanatics on Reddit estimated Stockfish's elo at about 3475. I'm not a chess player. I am inclined to agree. That would mean Stockfish 9 playing the world championship contenders, if I were to anthropomorphosize it, would be like your college chess champion playing a 6th grader who has no special chess skills.

In this video, Alpha Zero, crushes Stockfish so decisively, it doesn't even look like Chess anymore. Those of you who play will appreciate the sheer domination of this game.

https://www.youtube.com/watch?v=lFXJWPhDsSY

Alpha zero's record against stockfish is something like 28 wins, 72 draws, 0 losses.

Alpha zero's ELO is so high again, that I would say we don't really have the programs to gauge it's strength.
That's where we are in AI right now as an intuition.

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
What's interesting about Alpha Zero is how it got so strong. Computers have been good at chess for some time, and even before they beat a world champion (in 1997), it was only a matter of time before faster processors gave existing techniques the the ability to beat the best humans.

But those chess programs, while impressive, were very narrow in scope. At a high level, they basically enumerated all the possible positions, pruning out clearly bad branches as quickly as possible to limit search time. As evaluation functions got better (is position A better than position B) and deeper searches became possible, the outcome versus humans was preordained. Add to that you could give the computer as big an opening catalog as you wanted (common chess openings have been extensively analyzed many moves deep, and the computer won't have any trouble remembering any of that) and it became no contest.

Alpha Zero uses machine learning techniques to teach itself how to play games; its predecessor taught itself how to play go better than any human in an astonishingly short amount of time. To that point, people had not created go programs that could beat the best humans.

But, impressive as these accomplishments are, they occur in tightly constrained domains with a fixed set of rules. Compare the complexity of a chess board with real world domains.

AI is going to keep impressing. But it will continue, I think, to shine in situations where the "rules" and "the players" can be easily enumerated.
TomCat96
Posts: 1040
Joined: Sun Oct 18, 2015 12:18 pm

Re: Artificial Intelligence and the stock market

Post by TomCat96 »

rj342 wrote: Thu Oct 10, 2019 10:30 am
randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
That's what I thought too. It's certainly true and I won't deny it. In fact it was for that reason that for many years it was believed that AI would not be able to beat a human player in Go.

I believe the number of possible positions in Go is something like 2^398, significantly more than the number of atoms in the universe. Beating a human player from a simple search in the Go gamespace is completely intractable. Yet here were are, the top Go players in the world have been thoroughly beaten by AI. How did that happen?

Long story short, there have been incredible improvements. Kasparov lost in 1996 I believe, and we're talking about 20 years later.

Modern artificial intelligence is a combination of Calculus, Linear Algebra, Statistics, and Probability. Without delving into the mathematics of it, which is incredibly fascinating, I will say that thanks to statistical algorithms, modern AI is able to discern abstract aggregations of data (really correlated variables) and organize them into "concepts" which can then be manipulated, much like the way chess pieces are.

This is a huge incredible step forward, because it means that AI can meaningfully manipulate concepts to be drawn from data rather than simplistic entire state space searches.

What that means is that thanks to the mathematics of machine learning, theres been a major step forward towards addressing the difficulty of the "open ended" problem.
roboskier
Posts: 2
Joined: Thu Oct 10, 2019 2:41 pm

Re: Artificial Intelligence and the stock market

Post by roboskier »

ScaledWheel wrote: Thu Oct 10, 2019 7:52 am Time for me to put on my "crotchety old man" hat (I'm in my early 30's). When I started grad school, what is now being called AI was starting to be called machine learning (ML), and before that was called applied statistics. There is nothing magical about AI, unless we are actually talking about general artificial intelligence, but I am much less bullish about general AI than a lot of people.

Hedge funds and investment shops are already hiring for "AI" researches (deep learning, neural networks, probabilistic graphical models, etc.). In my opinion, most of the benefits that you'll see from "AI" are going to be in increased efficiency and new product lines adopted by public companies, and many more new public companies with AI/ML as core to their business strategy. I don't think it will have a dramatic effect on trading or other financial functions of the stock market, as these techniques are already being used for trading and finding information in the market.
I work for a company that does structural engineering software and the big buzz now is also AI. I think ScaledWheel nailed it, there is nothing magical about AI - at the end of the day it boils down to a big curve fit which is still only as good as the input data. Some of the algorithms are incredible at solving certain problems - like chess described above - where a massive amount of data can be accessed or rapidly generated. For other problems though, there just isn't that much useful data available. And it seems like nowadays "AI" is thrown around as a catch all to solve these hard problems.
jibantik
Posts: 568
Joined: Fri Nov 24, 2017 12:05 pm

Re: Artificial Intelligence and the stock market

Post by jibantik »

My prediction: the market will continue to return market returns, therefore I will continue to achieve market returns.

AI or not, people will always think they can beat the market, but they will fail, simple mathematical fact. The croupiers, the men in the middle, will continue to smile cause up, down, or sideways, as long as people think they can beat the market, they will keep making money and stealing part of the market returns.
randomguy
Posts: 11295
Joined: Wed Sep 17, 2014 9:00 am

Re: Artificial Intelligence and the stock market

Post by randomguy »

TomCat96 wrote: Thu Oct 10, 2019 1:43 pm This is a huge incredible step forward, because it means that AI can meaningfully manipulate concepts to be drawn from data rather than simplistic entire state space searches.

What that means is that thanks to the mathematics of machine learning, theres been a major step forward towards addressing the difficulty of the "open ended" problem.
We are definitely a couple steps beyond where we were in the early 90s where the way to make your chess program better was to build a faster computer so you could search ahead by 7 moves instead of 5. The question is even with all those advancements, how close are we to revolutionary changes. Self driving cars are the current poster child for AI. The advances we have had over the past 15 years are incredible. But I am not willing to guess if we are 5 years away from the revolution or 25 years. Heck the revolution might not come in our lifetimes. We could get stuck at level 3 and 4 for general use and Level 5 being confined to very controlled situations.

Right now we have tons of stock market AI systems and most of the ones I am aware of tend to be short term (<5 mins) trading machines that seek to exploit inefficiencies in the market. What the systems can't do is Warren Buffet style investing where you say company X is undervalued and we should buy and hold for a long term.
regularguy455
Posts: 487
Joined: Tue Jul 18, 2017 2:08 pm

Re: Artificial Intelligence and the stock market

Post by regularguy455 »

Automation combined with AI is a potent combination. Companies who adopt this will drive down the costs of goods and services. This will create winners and losers but owning the basket ensures you’re along for the ride.
gougou
Posts: 1317
Joined: Thu Sep 28, 2017 7:42 pm

Re: Artificial Intelligence and the stock market

Post by gougou »

AI and ML could be next bubble to pop. I work in ML in a big tech company and I stay away from startups that only do AI/ML.

And people are afraid AI could take over the world in our lifetime. What a joke :D
The sillier the market’s behavior, the greater the opportunity for the business like investor.
User avatar
arcticpineapplecorp.
Posts: 15080
Joined: Tue Mar 06, 2012 8:22 pm

Re: Artificial Intelligence and the stock market

Post by arcticpineapplecorp. »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
it will be priced in accordingly, like everything else.
It's hard to accept the truth when the lies were exactly what you wanted to hear. Investing is simple, but not easy. Buy, hold & rebalance low cost index funds & manage taxable events. Asking Portfolio Questions | Wiki
2pedals
Posts: 1988
Joined: Wed Dec 31, 2014 11:31 am

Re: Artificial Intelligence and the stock market

Post by 2pedals »

AI will continue to advance but will they be "better than us", like in Netflix series? I think probably not in my lifetime but I do expect technology make things less expensive over time but possibly less humane with more complex decisions.
gougou
Posts: 1317
Joined: Thu Sep 28, 2017 7:42 pm

Re: Artificial Intelligence and the stock market

Post by gougou »

rj342 wrote: Thu Oct 10, 2019 10:30 am
randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
Agree. For example, AI's can't play poker as well as human, and AI's can't play some computer games (DotA, League of Legends) as well as human. AI's need to have complete information to be good at games and when some information are hidden they become very dumb. This makes me believe the general intelligence is very very far away.
The sillier the market’s behavior, the greater the opportunity for the business like investor.
jibantik
Posts: 568
Joined: Fri Nov 24, 2017 12:05 pm

Re: Artificial Intelligence and the stock market

Post by jibantik »

gougou wrote: Thu Oct 10, 2019 8:51 pm
rj342 wrote: Thu Oct 10, 2019 10:30 am
randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
Agree. For example, AI's can't play poker as well as human, and AI's can't play some computer games (DotA, League of Legends) as well as human. AI's need to have complete information to be good at games and when some information are hidden they become very dumb. This makes me believe the general intelligence is very very far away.
Yes, this is dead on. AI can seem very impressive when the solution space is large and the problem is simple, e.g. Go.

However, watching DOTA AI, which had years of research and supercomputers for training, was remarkably unimpressive. And it was even playing a constrained version of the game. It seemed good, but the AI was superficial. As soon as people realized it's strategy, it became easy to beat and the AI could not adapt at all.
Topic Author
masonstone
Posts: 659
Joined: Thu Feb 01, 2018 3:01 pm

Re: Artificial Intelligence and the stock market

Post by masonstone »

mhc wrote: Wed Oct 09, 2019 2:19 pm I read your question differently from how the others have responded.

If AI is an advancement, then I would think it would help the stock market. Businesses will become more efficient.
I know this is a simplistic view, but I think AI should be good for the market.
This was my question, how AI will affect business (and their stocks) and not how it will affect Wall Street.
StandingRock
Posts: 344
Joined: Sat Feb 02, 2019 5:54 pm

Re: Artificial Intelligence and the stock market

Post by StandingRock »

Could someone let me know when this explosion is about to happen so that I can put up some blast shields around my home.
Call_Me_Op
Posts: 9881
Joined: Mon Sep 07, 2009 2:57 pm
Location: Milky Way

Re: Artificial Intelligence and the stock market

Post by Call_Me_Op »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
In a way that cannot be predicted accurately.
Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein
alex_686
Posts: 13320
Joined: Mon Feb 09, 2015 1:39 pm

Re: Artificial Intelligence and the stock market

Post by alex_686 »

So, I have seen lots of applications on the trading and stock analysis side. 2 specific examples come to mind. One firm bought up 10% of all credit card transactions in the US. They then analysed them to determine same-store revenue and growth for Starbucks. They also analyzed what portion was being spent by people in the loyalty program. They then scraped job sites to determine the number of open positions for Starbucks, trying to figure out where and how they were expanding. They also used it to estimate employee turnover. Low turnover implies happy employees which implies productive employees.

For Bogleheads, I suspect that AI will be used to create direct indexing products for us. A very simple example is where the program would take the S&P 500 allocate the bits with the highest tax drag to tax advantaged accounts and the lowest tax drag to taxable accounts.
Former brokerage operations & mutual fund accountant. I hate risk, which is why I study and embrace it.
User avatar
tadamsmar
Posts: 9972
Joined: Mon May 07, 2007 12:33 pm

Re: Artificial Intelligence and the stock market

Post by tadamsmar »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
Two questions:

How will it affect active trading?

It should make trading more rational by exploiting behavioral anomalies. But that assumes there are behavioral anomalies that still can be exploited reliably for profit. The first hedge fund developed by Ed Thorpe was based on advanced computing as a factor.

There is an AI-powered EFT that has not done well in its short lifespan:
https://www.fool.com/investing/2019/01/ ... g-the.aspx

How will it affect publicly traded companies?

The most extreme speculation about impact is called the singularity: https://en.wikipedia.org/wiki/Technological_singularity

The idea is that intelligent machines will start building more intelligent machines and there will be a huge increase in the automated production of every product. It seems obvious that the owners of such technology would have high revenues. But not sure about the profit margins in non-monopoly scenarios. It none of the owners was a publicly-traded company than that could be bad for stock investors I guess.

I don't think that the impact of AI will invalidate the principles of Boglehead investing unless it happens in an unpredictable manner. That would probably mean that it happens fast. The singularity seems like wild speculation but not impossible, I think.

Historically, there was a singularity in the explosive power of bombs: the creation of the A-bomb and H-bomb. There is apparently no limit on the explosive power of the H-bomb, the US and Russia just stopped building bigger H-bombs in 1961. This singularity has not been all that profound in its effects, but we are still muddling through it. Of course, an AI singularity could have more far-reaching effects.
Last edited by tadamsmar on Fri Oct 11, 2019 11:47 am, edited 1 time in total.
User avatar
tadamsmar
Posts: 9972
Joined: Mon May 07, 2007 12:33 pm

Re: Artificial Intelligence and the stock market

Post by tadamsmar »

Warren Buffet reported that AI was an unanticipated threat to Berkshire Hathaway:

https://fortune.com/2017/05/06/warren-b ... iving-car/
User avatar
tadamsmar
Posts: 9972
Joined: Mon May 07, 2007 12:33 pm

Re: Artificial Intelligence and the stock market

Post by tadamsmar »

I think we tend to focus on intelligence and ignore experience.

Suppose I gave you fully licensed ownership of a computer that is was as smart and capable of newborn Einstein. You would probably go bankrupt. It took newborn Einstein over 20 years to produce any product of real value.

In chess and go, it has been easy to produce lots of high-fidelity artificial experience for artificial intelligence. In self-driving cars, gaining experience has been a slower more expensive and riskier business.

I think that companies like Google and Facebook can work on one big brain that gains tons of experience quickly from interacting with billions of people.

As far as I know, self-driving cars don't have one big brain. Technology transfer of experience can he hard between our many individual human brains. Not sure how easy that is in artificial intelligence domains.
Topic Author
masonstone
Posts: 659
Joined: Thu Feb 01, 2018 3:01 pm

Re: Artificial Intelligence and the stock market

Post by masonstone »

tadamsmar wrote: Fri Oct 11, 2019 12:33 pm I think we tend to focus on intelligence and ignore experience.

Suppose I gave you fully licensed ownership of a computer that is was as smart and capable of newborn Einstein. You would probably go bankrupt. It took newborn Einstein over 20 years to produce any product of real value.

In chess and go, it has been easy to produce lots of high-fidelity artificial experience for artificial intelligence. In self-driving cars, gaining experience has been a slower more expensive and riskier business.

I think that companies like Google and Facebook can work on one big brain that gains tons of experience quickly from interacting with billions of people.

As far as I know, self-driving cars don't have one big brain. Technology transfer of experience can he hard between our many individual human brains. Not sure how easy that is in artificial intelligence domains.
It would be instantaneous for AI
finite_difference
Posts: 3633
Joined: Thu Jul 09, 2015 7:00 pm

Re: Artificial Intelligence and the stock market

Post by finite_difference »

Arguably, “Mr. Market” is a form of artificial intelligence.

As the market grows and traders continue to become smarter, faster, and more knowledgeable — perhaps aided by new ML tools and even emergent AIs — presumably Mr. Market will become more “intelligent” as well. But as Graham warned us, Mr. Market can be irrational at times. Since humans are irrational at times, perhaps that makes sense.
The most precious gift we can offer anyone is our attention. - Thich Nhat Hanh
finite_difference
Posts: 3633
Joined: Thu Jul 09, 2015 7:00 pm

Re: Artificial Intelligence and the stock market

Post by finite_difference »

jibantik wrote: Thu Oct 10, 2019 9:28 pm
gougou wrote: Thu Oct 10, 2019 8:51 pm
rj342 wrote: Thu Oct 10, 2019 10:30 am
randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
Agree. For example, AI's can't play poker as well as human, and AI's can't play some computer games (DotA, League of Legends) as well as human. AI's need to have complete information to be good at games and when some information are hidden they become very dumb. This makes me believe the general intelligence is very very far away.
Yes, this is dead on. AI can seem very impressive when the solution space is large and the problem is simple, e.g. Go.

However, watching DOTA AI, which had years of research and supercomputers for training, was remarkably unimpressive. And it was even playing a constrained version of the game. It seemed good, but the AI was superficial. As soon as people realized it's strategy, it became easy to beat and the AI could not adapt at all.
Do either of you play Go? It is a sincere question, because I think you are vastly underestimating the complexity of the game. I thought it looked simple too until I started playing.

Has Google really put the same amount of effort it put into defeating Go grandmaster Ke Jie, toward defeating one of the top DOTA or LoL players, and come up empty handed? If not, I don’t think it’s a fair comparison.
beehivehave
Posts: 466
Joined: Thu Aug 08, 2019 1:21 pm

Re: Artificial Intelligence and the stock market

Post by beehivehave »

gougou wrote: Thu Oct 10, 2019 8:51 pm
rj342 wrote: Thu Oct 10, 2019 10:30 am
randomguy wrote: Thu Oct 10, 2019 9:32 am
TomCat96 wrote: Wed Oct 09, 2019 9:16 pm

Imagine someone looking above your head as if you were a 6th grader and they were a college kid.
Now imagine someone looking at that college kid the same way, and then someone looking at that guy like he was as sixth grader.

That's about the strength of AI right now with respect to chess. I anticipate that trend to follow in other fields.
The best computer chess player has been at the level of the best human for ~30 years. So far there has been very limited transfer to other fields. Things like chess are pretty trivial (complete information, very limited number of options) compare to a most problems . It is always important to remember what we find hard isn't necessary what a computer will find hard. On one of the self driving car discussions, some posted a picture of a tractor trailer doing some complex parking mover. It was pointed out how this is hard for a human (you need to be accurate with in 6" not to hit things) but trivial for a computer (i.e. your sensor net has that level of accuracy) to plot out. On the other hand pulling up to an intersection and deciding if a person standing there is an authority figure whose hand waves are instructions to be followed or a crazy person is something that any human can do but is tough for a computer. How many problems can we get AI to solve (i.e. will your health care providers visits in 20 years be a trained professional inputting things into a computer and the computer spitting out a treatment plan?) over the time is very hard to predict.
In the more general universe of problems, Chess, Go , etc are *simple* problems in a sense. They are susceptible to the modern "machine learning" correlation model of problem solving because the problems are nicely bounded even though the solution space is large. The boards are simple enough to define precisely, the legal moves, and the victory/defeat conditions. Completely different from any sort of truly open ended, creative problem solving amidst ambiguity or apparently conflicting data.
Agree. For example, AI's can't play poker as well as human, and AI's can't play some computer games (DotA, League of Legends) as well as human. AI's need to have complete information to be good at games and when some information are hidden they become very dumb. This makes me believe the general intelligence is very very far away.
While general intelligence as opposed to specialized intelligence may be very far away, there is AI that has indeed beaten DotA and poker with regularity.
Workable Goblin
Posts: 101
Joined: Fri Mar 01, 2019 7:37 pm

Re: Artificial Intelligence and the stock market

Post by Workable Goblin »

tadamsmar wrote: Fri Oct 11, 2019 12:33 pm As far as I know, self-driving cars don't have one big brain. Technology transfer of experience can he hard between our many individual human brains. Not sure how easy that is in artificial intelligence domains.
In this case, it's pretty easy, relatively speaking (the operative term is "online learning"). Cars within a company will share experience, so that all of the cars will learn from the experience of any individual car. This may be less possible if and when self-driving cars are deployed commercially (essentially, customers may not consent to sending back data to the company providing the algorithm), but for the present there are systems to ensure that all cars are "synchronized" in terms of how capable they are.

A more pressing issue, however, is that (as recent research has shown) some machine learning systems are vulnerable to failure with very small, even undetectable variations between their training and operational data sets. Depending on how easy this is to solve and how applicable it is to the stock market, this could significantly change how much machine learning techniques impact the stock market.
shess
Posts: 2164
Joined: Wed May 17, 2017 12:02 am

Re: Artificial Intelligence and the stock market

Post by shess »

masonstone wrote: Wed Oct 09, 2019 11:22 am With the explosion of AI that is coming in the next decade how do you think it'll affect the stock market?
A lot of the responses here seem to concentrate on AI doing what humans do, which is to say make predictions about companies and invest accordingly. I think that that's not what's happening, instead AI is being used to ferret out things that humans cannot move fast enough to figure out. Basically, high-frequency trading, squared. The algorithms will dynamically analyze the stream and find patterns to take advantage of, and then take advantage of them to make tiny bits of money in short bits of time which add up to large amounts.

I think how it will affect the stock market is that it will make things more sketchy, and it will also push things towards being less transparent. The sketchy side of things is basically the AI version of fads and momentum, which can happen hundreds of times faster because AIs don't need to read articles in Forbes to make bad decisions. So you'll have hundreds of AIs moving at once, which will mostly cancel out, but periodically some will do the wrong things at the wrong time and get a perfect storm and the market will drop a bunch for no apparent reason. Heaven help us if things cascade from there.

The less-transparent side of things is that for the most part the companies running these AIs are parasites on the market. Things work best for them when there are a lot of routes for trades to get sourced and sunk to without much oversight. Free trades make me really nervous on this front - brokerages aren't really going to do this as a free service for you, they're going to sell their order flow to the highest bidder, and that bidder does not have your interests at heart.
User avatar
tadamsmar
Posts: 9972
Joined: Mon May 07, 2007 12:33 pm

Re: Artificial Intelligence and the stock market

Post by tadamsmar »

Workable Goblin wrote: Sat Oct 12, 2019 12:44 am
tadamsmar wrote: Fri Oct 11, 2019 12:33 pm As far as I know, self-driving cars don't have one big brain. Technology transfer of experience can he hard between our many individual human brains. Not sure how easy that is in artificial intelligence domains.
In this case, it's pretty easy, relatively speaking (the operative term is "online learning"). Cars within a company will share experience, so that all of the cars will learn from the experience of any individual car. This may be less possible if and when self-driving cars are deployed commercially (essentially, customers may not consent to sending back data to the company providing the algorithm), but for the present there are systems to ensure that all cars are "synchronized" in terms of how capable they are.
It's true that all the vehicles could record their inputs and the inputs from all vehicles could be used to train a single neural network. The inputs would be the experience of the vehicle to the extent that the inputs capture the experience. Unlike Chess and Go, the sensors may not capture all relevant experience, of course.

Another way that separate neural networks can be "merged" is by having them vote or collaborate in some manner. Humans do stuff like that to combine their knowledge. Market pricing is a (possibly imperfect) mechanism that combines the knowledge and experience of many humans, and these days it probably also combines the knowledge and experience of AI entities. Also, there might be multiple (not fully integrated) neural networks collaborating in our own brains.
Post Reply