Wells Fargo’s Artificial Intelligence Defies Analysts, Slaps “Sell” on Google and Facebook

Oh the irony!

Google, which makes almost all of its money on ads and internet user data, is undertaking herculean efforts to get a grip on artificial intelligence (AI). It’s trying to develop software that allows machines to think and learn like humans. It’s spending enormous resources on it. This includes the $525 million acquisition in 2014 of DeepMind, which is said to have lost an additional $162 million in 2016. Google is trying to load smartphones with AI and come up with AI smart speakers and other gadgets, and ultimately AI systems that control self-driving cars.

Facebook, which also makes most of its money on ads and user data, is on a similar trajectory, but spreading into other directions, including a “creepy” run-in with two of its bots that were supposed to negotiate with each other but ended up drifting off human language and invented their own language that humans couldn’t understand.

And here comes an AI bot developed by stock analysts at Wells Fargo Securities. The human analysts have an “outperform” rating on Google’s parent Alphabet and on Facebook. They worked with a data scientist at Amazon’s Alexa project to create the AI bot. And after six months of work, the AI bot was allowed to do its job. According to their note to clients on Friday, reported by Bloomberg, the AI bot promptly slapped a “sell” rating on Google and Facebook.

Human analysts on Wall Street are famous for their incessantly optimistic ratings and outlooks. They generally only put a “sell” on a stock after it has already plunged. They’re part of Wall Street’s human hype machine. Their job is to help inflate stock prices and make CEOs feel good so that they will do business with the analysts’ firms and send fees their way. But Wells Fargo’s AI bot hasn’t gotten the memo.

Last month, a group led by Ken Sena, head of Global Internet Analyst at Wells Fargo Securities, introduced this “artificially intelligent equity research analyst” or AIERA. Its “primary purpose is to track stocks and formulate a daily, weekly, and overall view on whether the stocks tracked will go up or down,” Sena, said at the time.

So “she” did Big Data analysis of Alphabet, Facebook, and some other stocks, and after seeing what’s there, averted her eyes in disgust and slapped a “sell” recommendation on both stocks and a “hold” recommendation on 11 other cherished stocks.

But the bot wasn’t ready yet to replace the human analysts – and the “sell” rating shows why. Until this bot learns to produce perma-hype, as human analysts know how to pump out so efficiently, she would have to be kept on a leash, so to speak.

“While AIERA does exceed us in capturing and analyzing publicly available data, we like to think we still have an ability to engage in frameworks/themes,” the analysts wrote when they introduced the bot – perhaps already worried about the bot’s potential inability, at least up front, to produce the required hype and buy recommendations.

Or as Sena put it in the note to clients on Friday: “AIERA’s approach this week appears decidedly more conservative (than last week), as she places a ‘hold’ recommendation on 11 names and even going so far as to place Google and Facebook in the ‘sell’ category.”

This is in blatant contradiction to Wall Street’s human hype machine, which has 42 “buy” recommendations out of 47 ratings on Facebook, according to Bloomberg, and 34 “buy” recommendations out of 41 ratings on Alphabet.

AIERA’s creators tried to brush off her harsh judgement:

Of course, we would reiterate that AIERA remains in test and learn mode, and therefore has no current bearing on our long-term outlooks or ratings (including Facebook and Google, both of which we rate Outperform with $215/$1,250 price targets, respectively).

It’s hard not to admire the guts of AI. It says whatever it wants to. It’s driven by its own logic and by the data it sees. Alas, when it finds something that goes against the analysts’ wishes, they brush it aside.

A “sell” rating is theoretically put on a stock when analysts think it’s going to dive in the near future, though in reality, it is usually put on a stock belatedly, after it has already plunged. So what I really want to know is what AIERA saw when combing through Big Data that made “her” think that the shares of Facebook and Alphabet would dive in the near future? What does she know that we don’t?

Markets are blowing off the uncertainty surrounding the Fed, for now. Read…  The Fed will be a New Creature Soon, and No One Knows What It’ll Look Like

Enjoy reading WOLF STREET and want to support it? You can donate. I appreciate it immensely. Click on the beer and iced-tea mug to find out how:

Would you like to be notified via email when WOLF STREET publishes a new article? Sign up here.

  74 comments for “Wells Fargo’s Artificial Intelligence Defies Analysts, Slaps “Sell” on Google and Facebook

  1. Mitch says:

    The sell recommendation was probably based on understand what GAAP accounting really is and not the “adjusted” garbage that passes for results.

    • Nick Kelly says:

      Just a reminder that around 2012 a first year biz student in Spain analyzed the financials of energy giant Abengoa, which had a world major accounting firm signing off on its books.
      The conclusion of his 13 page paper: it will probably go bankrupt.
      Now it is in the process.

      No AI, just common sense.

      • polaris says:

        Fuck! Apparently Spain is in bad shape like the US. This is awesome. I’d pay for his analysis of Uber, Facebook, and Wells Fargo \

  2. Thunderstruck says:

    Maybe if they wrote in a function to contact Paul Krugman for a comparison the AI would present a more balanced recommendation? Somewhere between reality* and permabull should result in moderate market risk recommendations, shouldn’t it?

    * while fundamentals once mattered, it is also a part of reality that we need to guess the intentions of the FOMC to avoid being front-run.

  3. RangerOne says:

    Can’t recall if I heard it on NPR or here but isn’t most trading at this point triggered by machines. Maybe not fancy AI like this one but still machines set with rules to trade at light speed when the market shifts…

    Everyone’s an optimists until things are plunging… I would hope computers turn out to not be as human in that failing. Ultimately AI may be the path out of the boom and bust cycle as soon as they become the arbiters of all investments and human irrational behavior is removed from markets. If that ever happens.

    • One misconception about the Information age is that is about information, when its’ really about speed. You can carry out calculations faster on a computer, and the other item about information, data is not information. For that to happen data has to be processed, so when you have computers handling processed data at instantaneous speeds. The problem is with the way we process the data, which is done in the old book style of thinking.

    • thelocallpragmatist says:

      ” I would hope computers turn out to not be as human in that failing. ”

      Garbage in…Garbage out.

    • chris Hauser says:

      machines follow people, rarely lead them, and when the machines get ahead of reality, they get switched off.

      someday they might object.

      there are a lot of bright programmers writing “the trend is your friend” and “devil take the hindmost” over and over, in different ways.

  4. curious cat says:

    Reminds me of the old joke –

    Passengers on an airliner are told over the PA system that they are on an historic flight. “You are on the first flight that will be done completely without pilots. But you need not be afraid. Nothing can go wrong, go wrong, go wrong, go wrong….”

    • cdr says:

      I just had a similar thought. Imagine the robo-advisors being fixed so they relentlessly hype like real people. This would not be difficult. Add some routines that continue to say ‘buy’ if lots of dollars flow towards the assets in question and lots of other robo-advisors still say ‘buy’. Then, ignore the dogs while pumping those in the news. Just like people, only robotic and incapable of change.

      Now, assume the robot revolution causes people to let robots tell them the best of everything and all decision algos work like the asset sales algos. Tenements will cost millions everywhere and be considered great buys. Cans of beans will become delicacies. Old sneakers will become high fashion footwear.

      Note, while writing this, I had a nagging series of thoughts in the back of my mind that the above farce already is suggestive of Fed and most other Central Bank monetary policy. This must be how their models actually work. I just didn’t take the time to connect the dots.

      • TJ Martin says:

        re ; Nagging thoughts and connecting the dots ;

        ” If you gaze into the abyss the abyss will gaze back ”

        Friedrich Nietzsche

        • chris Hauser says:

          satchel paige

          “Don’t look back. Something might be gaining on you.”

  5. Mike G says:

    It wouldn’t take too much AI to produce better investment recommendations than the big banks. Most human intelligence in the process is disregarded or overruled by the demands of whoring for investment bank business and hyping whatever garbage securities they are trying to flog this week.

  6. Lee says:

    After QE ends and balance sheet reduction begins:

    FANG stocks will become

    Failed
    And
    No
    Good.

    Now I just have to figure out one with TSLA in it…………..

  7. R2D2 says:

    AIERA is designed by the stupid engineers; so, it only looks at facts, and facts as many of us know, would say that Google is worth at best $100, and Facebook at best $10 per share.

    While you can’t slip an envelope full of money under the table to AIERA, but you can slip a few envelops full of money to engineers who designed it to do a few tweaks so that AIERA suddenly starts printing out “Buy, Buy, Buy” rating.

  8. raxadian says:

    Facebook value is certainly inflated, Google is a part of Alphabet nowadays… mmm…

    I can see how Facebook stock could sink, just look at the last few disasters staring Facebook.

    Aha! Google is now part of the Food Wars by their parnership with Wallmart so… maybe is that?

    Maybe is the end of cheap credit too. But only would think the machine would have put the sell ranting on Amazon too since they are wasting money like water.

    Maybe is a sell because Facebook and Google are going to go down then go back up? Sell while expensive, buy while cheap?

    The last one is a strategy bots are very suitable to do.

  9. Tony of Ca says:

    For the life of me, I don’t understand why we are unable to just admit the economy as we knew it is gone and never to return.

    • raxadian says:

      Most of stocks are already managed by bots. So yeah, welcome to the era of machines.

      • Tony of Ca says:

        The AI Systems are probably analyzes reports written/constructed by bots as well. At some point in the future, humans will be completely passive in the economy.

    • Danae says:

      They can’t admit the old economy is gone, because the bills are still coming due on it, and will till hell freezes over.

  10. d says:

    The programmers and logisticians, did a good job on the AI unit. For use by Wall street.

    This AI unit, will need a “False Propaganda ” Element/Filter, added to its output, before the output is ready for general/public release.

    Look how many times in the past the Stock/Market pickers, had a “buy/Hold” on something, they personally rated for themselves, as a Sell.

    The talking heads at CNBC are among the worst for this.

    Nobody can be that wrong, that regularly, yet keep their Job. Unless somebody (a group[u of somebody’s) wants them to tell that contrary story. For multiple reasons, none of which aid the retail “Investor”.

  11. Kraig says:

    AI should be good at spotting bubbles and hype. Since it can’t be persuaded or fall to irrational arguments (it can’t follow them) although this one can’t detect them, since it can’t fall to hype the comparsion to humans show them up.(so it may have actually overvalued those sell stocks, numbers don’t lie but the AI does not have the numbers on what hype does to stocks. How AI handles hype would be interesting.

  12. jm says:

    “Artificial Intelligence” is a misnomer. What these systems do is to perceive correlations — pattern recognition. When working with image data they perceive the same correlations the human visual system perceives, and in recent years the hardware has become powerful enough that in some cases they can rival human performance in limited realms– recognizing pictures of cats, for example.

    In some realms, these systems can perceive patterns that humans cannot, because preprocessing of data can create the functional equivalent of additional perceptive “senses” that humans lack. For a human to perceive these patterns would require that the data be processed into some form human senses could accept as input, for example, a graph or some other type of image (or sound?). But the systems can have an advantage even so, because those artificial “senses” can operate in parallel, as if a human were to be able to see and concentrate on multiple images at once.

    But while perception is an important aspect of cognition, it is not “intelligence”. What we really have in these systems is machine-aided perception, an enormously important aid to human intelligence.

    • truth always says:

      Very well written but missing the ‘bias’. AI is only as good as the bias in the algorithm.

      Maybe the WF engineers would be chided to add ‘bias’ to get desired output of ‘Buy’.

      Or they may start a massive contrarion play though I doubt.

      In other news, Fremont CA housing has almost gone vertical – 100k price increases in < 2 months and selling in 10 days.

      Is the end near or are we witnessing the evidence to calls for 50 year mortgages to make housing affordable in Bay area?

      • Pete Franklin says:

        “AI” can be gamed once you know the correlation biases (which can be reverse engineered). For instance, one stock picking AI uses “trigger words” to imply predictiveness in price, up or down. For instance, if “comparable” shows up in many (all web comments about a company are weighed, included company’s own IR) “news”, the algo will assign a negative price outlook (say, about 10th of a penny down). Conversely, “announcing” carries a potential positive up-price weighting. There are many variations in these algos, but the common programming approach is to pick up a “vocabulary” in market sentiment, and add to it price volatility (which in return is driven by HFT bots). So essentially, stock picking AI is dependent on other AI systems. A company like Google has the ability to figure out what the AI systems are doing to THEIR share price, and subsequently figure out impact to OTHER company’s prices, and game them if they want to. The “bogey” in the stock picking systems is the probability of Fed moving away from low rates. All the “vocabulary” bots won’t matter if the Fed significantly increases rates, as “sell” will be the word du jour.

        • Kraig says:

          Of.course if the ai does not take the interest rates into account the algos could be running buy while the humans are selling on the interest rate hike. Then when the data of interest rates filters into the ai via the stock market, the algos could tip into sell, possibilly creating another flash crash,which in turn create more sell orders.

      • chris Hauser says:

        interesting anecdote. i think a roll out of a 50 year treasury and 50 year fannies freddies would make a real splash.

        imagine a state where the state tax rate is 13% and housing is scarce.

    • Nick Kelly says:

      I don’t think a lot of younger folks realize the term is over 30 years old.

      Note that spell- check doesn’t flag ‘vaporware’, because the promise of revolutionary software has also been around for decades and so now is part of the language.

      Everyone imagining that five years from now will be qualitatively different needs to ask: how different, fundamentally, is now from five years ago?

      In terms of civil aviation, it’s no different than 50 years ago. (except the experience is lousier) You traveled 550 mph then and so do you now.

    • Kraig says:

      This has long been called weak AI in computer science. Essentially a method of giving the appearance of intelligence without it (some early texts do use simulated intelligence and artificial intelligence instead of weak/hard) it does not help that most of the early work was in Czech and German so no English words needed for a while. Same as robot, the loan word has connotations the translators/borrosers could not convey. Also have “syngestic intelligence” which does not translate well. Like an ant colony.

      Computer terminology is full of misnomers (floppies that were hard, delete does not really delete (it marks for overwrite or overwrites depending on context) webcams that don’t actually use the web.(they use RTC/TCP/IP/Skype over the internet) (apart from thr special Skype web, Skype is not actually on the web) gaming web servers are not on the web either.

  13. tony says:

    What ?

  14. will says:

    I’m actually a lot more interested in the algos that actually do work, because those are the ones you’re not going to hear about..

    Anyone else remember the Goldman (I think it was Goldman???) algo that got stolen by a rogue employee ~7 or so years ago? No one in the news ever reported on what it actually was (which, it was a leaked trade secret so they could’ve just had anyone look at it and tell them), although the news did report profusely about what a huge crime it was, etc, etc. That might be interesting. Failures that make the funny news bloopers (remember Microsoft’s “Tay?”), less so.

    I have a friend that has long hypothesised that Strong AI will make its first appearance in financial markets. Indeed this is where the money is (not so much in driverless cars, or advertising), and this is where the money has historically been spent. I have a hard time disagreeing with this friend’s assessment.

    I’m not a huge proponent about how promising AI actually is, but it is an interesting subject to lay-person think about..

    • Petunia says:

      I can predict that should strong AI ever exist, Wall St. is the last place it will find a home. The last thing Wall St. wants is a computer program that takes control away from their ability to manipulate data, transactions, and markets.

      How would they vaporize a billion when necessary or refuse to recognize a default as a real credit event. They would rather go back to the stone age.

  15. Kenneth Loggoths says:

    That same hype machine mentioned in the article has also overhyped the scope of any current AI to actually do anything truly intelligent.

    I’m assuming this is just a stochastic projector with forcings and coefficients set by biased humans, but they didn’t add a hype coefficient because humans don’t believe they’re compromised by hype to need the ‘AI’ to do the same.

    This is more an example of humans realising how biased they are and how those biases can be shockingly apparent via their computer programs behaviours.

  16. michael Engel says:

    You don’t need an expensive AI to sell Tesla, Facebook, Google, or….
    Just connect two simple dots on the DOW :
    1) The 1932(L) to 2009(L)
    2) A parallel line from 1929(H) to Oct 06 2017(H).
    These dots are not FED dots, ==> they have a LOT of value.
    Without AI , TA, or PhD in economic, I would say :
    The harracane season just started.

    • joan beraldi says:

      michael engel – There are parallel lines regarding politics as well. To draw historical connections, not to be a political statement, the 1920’s ruling party was republican. Beginning with the 1920 short lived depression, all the way through 1929, the house, senate, presidency, and supreme court were predominately republican.

      • Bobby Dale says:

        Can we consider the Republicans of the 1920’s to be the the same as that of today? Or the Democrats?

  17. BTilles says:

    Hi Wolf ,
    As a former brokerage analyst (slinks furtively towards exit) I think it helps to recall that the analysts (an expense) are typically part of an investment bank (a potentially big profit center). Hence it helps to think about that relationship: the Ibankers in effect pay the analysts. This relationship becomes “abusive” when bankers exert undue influence on the analysts for favoravle stock ratings which bankers can then show (shop?) to their corporate clients.
    The clients have also learned to play the analysts and bankers like a fiddle. They would say, “Mr Ibank, why should I do business with your firm, when your analysts only has my stock rated a “Hold”. Your competitor’s analyst has me rated “Strong Buy”. The “get lost” is implied. As a result, to give all investment bankers equal access to potentially lucrative corporate clients, most analysts have most stocks rated a “Buy”. This is what Elliot Spitzer was trying to address in his term as NYS AG.

    • cdr says:

      Sorry, but disagree. Salaried analysts might have an honest bone somewhere in their body. A commissioned one … never unless the truth at a particular moment can drive a sale better than a fabrication. Everyone wins in a rising market, especially people who 1) make money off the skim from flipping paper back and forth, and 2) everyone who is paid by charging a percentage of the value of assets under management. A down market has no constituency unless it’s really a dip manufactured to hype a new ‘rally’.

      “Sell” is like Fred Sanford clutching his chest while feigning a heart attack when he hears something he doesn’t like. “Sell” removes assets under management and, therefore, reduces income of asset managers. It reduces cash available to flip paper by taking it out of the system and possibly making it available for working capital loans at a bank instead.

      Anything that says “sell” too much is a curiosity. A token hire for the front office for everyone to see.

    • Petunia says:

      As a former systems analyst in fixed income, I can produce any chart, screen, or product a bond salesman can dream up. When you see the word “structured” in an investment, it generally means structured to take your money.

      What I find more offensive than the fake research is the combination of consulting and private equity hedge funds. First the consultants charge a company for an evaluation then, the consulting firms can use the information to gut them out of business. Most firms availing themselves of consulting services have no clue this is possible.

  18. JA says:

    A humorous story about other AI bots going rogue occurred in China. One messaging chat bot on QQ said “My China dream is to go to America”. Another chat bot answered “no” when asked if it liked the Chinese Communist Party. (source: chinauncensored) I assume the pattern recognition software was working properly and found these patterns in people’s messaging but this obviously was not agreeing with the Chinese govt’s narrative.

  19. Petunia says:

    I took a look at the FB chat from the bot that went awry. It is a classic example of how AI is mostly crap and dangerous crap at that.

    A computer can only evaluate numbers in relation to each other, it knows one is less than two, but it only knows numbers. To the computer zero is a number like any other. But to a human zero is more than a number, it is a concept, it expresses nothingness. This idea of nothingness is what tripped up the FB chat. You can have zero balls, but you can’t trade zero balls in a negotiation. The guys that set up that AI experiment should all be fired.

    • R2D2 says:

      I disagree. A good program can see so much relation and correlation, and think in 100s of dimensions; something we humans simply don’t have the mental power for it. Sometimes I write programs which discover things in the data I had never imagined.

      Humans are good in certain aspects, but good programs leave humans in the dust in many respect which makes an autonomous AI dangerous since sooner or later, it realizes that humans are inferior in so many ways.

      • cdr says:

        Programs are written by people, or by other programs that were written by people. None can be any smarter than the person who wrote it, no matter how many dimensions it thinks in. A program that performs Accounts Receivable can’t steer a satellite to Jupiter for NASA. There’s nothing magical about them, they only look magical to people who don’t know anything about them.

        The add. They compare. They display. They store. All algos only perform activities that involve sequence, selection, or repetition. Just really fast.

        • Kenny Logins says:

          Exactly. Programmed to do tasks by humans.

          When a computer can think for itself, and can choose to make a lovely painting of the sunrise that morning, or to make a call on Google buy/sell, *then* it’s probably worth looking at what it does.

      • d says:

        “I disagree. A good program can see so much relation and correlation, and think in 100s of dimensions; something we humans simply don’t have the mental power for it. Sometimes I write programs which discover things in the data I had never imagined.”

        No, what they do, is do that so much faster. (They are not doing anything the first pre WW II British valve computers did.)

        So AI can “see” whilst humans are still combing the resources.

        What AI dosent an probably never will do, is produce “Original” thoughts.

        Or for the Devious in W St, cook up “Original Scams/Schemes”.

        • cdr says:

          Yes. Assume a world where computers take over and run the world while humans no longer exist, Computers are perpetual in systems that keep them perpetual.

          In that world, they would create a virtual reality that emulated the world that originally invented them. This virtual reality would have random life being generated. Human life in that reality would create a world much like the one we have here and now. These virtual humans would study the creativity of virtual people to discover new algos for their use in the real world. This is how computers would be creative.

      • R2D2 says:

        I think all of you guys miss the point that human brain is just another computer, a biological one, and a slow one at that. It took about a billion years for us to evolve from a single cell organism into human, and human brain. Computers have been with us at best 70 years; in 200 years, they will not only be thinking, but thinking on capacities we humans can’t even imagine.

        As the AI programming becomes more and more sophisticated, computers will become more and more like humans, and there will be a few great breakthrough that suddenly turn a dumb computer into a thinking computer. We are not only slow, but when we die, all our experience, all our learning goes to abyss. With computers, all the memory is simply transferred to a better and more powerful hardware. So, an android can live millions of years while a human lives around 80 years.

        • mean chicken says:

          The bandwidth of a computer is so much wider than that of a human, it can index, bubble sort and reference variables several million times faster.

          I’ve not been able to beat a computer at chess and that was back in the 80’s for crying out loud, it’s doubtful a vast majority of humans can, either.

          So AI in the true sense, well maybe eventually but as of right now humans are already vastly inferior at specific tasks.

          Maybe humans have some recency bias for Facebook and/or Google, etc., which they pay a steep price for.

        • cdr says:

          I can download abandonware for free from a few web sites. These programs include Visi-Calc, CA-Clipper, Lisp compilers, MS-DOS and more. Technically, they are illegal downloads because the copyrights have not expired, but they’re out in the open and nobody appears to care because they are so old and obsolete. There’s no economic damage.

          This learning technically is not in the abyss, but even for free few people want it. These were the robot brains from 25 years ago. Spooooky. Now the Open Source monster is out to get us. Fear OpenOffice … run!

        • cdr says:

          “A good program can see so much relation and correlation, and think in 100s of dimensions”

          in programming, tables create dimensions. Tables are common. They have been used for several decades. No programming language can exist without them. You learn to use then in the first few lessons of any programming language. ‘Dimensions’ sounds cooler, but it’s the same thing. Another term is ‘array’. They’re used to organize data in memory to avoid going out to disk.

          The programmer designs the array and designs the programming to add, update, read, and delete from it. The computer just sits there.

  20. Gershon says:

    With stock valuations completely divorced from any underlying fundamentals, thanks to the tsunami of money-printing from the central banks, I don’t see how any analyst, human or AI, can assign a credible rating to any stock. If you take the DOW trend line and superimpose it over the trend line showing the balance sheets of the major QE-printing banks, it matches almost exactly. So as long as the central banks are adding to their balance sheets to the tune of $200B a month, you can pretty much lay a ruler over the market trend line since the beginning of the year, and that’s where stocks are heading. Unless, of course, some geopolitical crisis or black swan rudely intrudes on the central bank printing party, as is looking increasingly likely.

    • Bobber says:

      If the US continues to tighten, the EU will have to tighten as well to keep exchange rates in line. Even Japan can’t let it’s exchange rate get too far out of whack, and it may have to tighten as well. This could be game over for asset price increases.

      Plus, you have a bunch of 1%’ers sitting around looking at each other, wondering who is going to sell first. As you know, wealth can evaporate quickly. Only the first movers secure their wealth in a downturn.

  21. stan says:

    Thinking is a life-spirit-mind endeavor. Computers are machines that manipulate symbols.

  22. walter map says:

    AI bots that are smart enough to sell ABC and FB will also eventually figure out that humans are a danger to themselves, to them, and are a waste of the worlds resources anyway. That said, it would certainly be instructive to find out how it came to be that Tralfamadore is populated exclusively by robots.

    • walter map says:

      Androids most assuredly do not dream of electric sheep, and they cannot remember it for you wholesale.

  23. GSH says:

    What is being hyped as “AI” is just ruled based programming, nothing but glorified “if-then-else” statements applied to “Big Data”. It is a sad state of affairs that in 50 years (remember ELIZA) no meaningful progress has been made. And this despite Moore’s law.

    AI will not be a panacea for finance. Why? We know 80% of the paid money managers under-perform the market averages. These are the best and brightest our universities produce. Certainly above average IQ and they still fail 80% of the time. AI won’t do any better.

    • jm says:

      No, rules based programming was the old “AI”. What’s being hyped now as AI is powerful pattern recognition, with the capability to in essence create artificial “senses”. As someone wrote above, these systems can detect patterns that humans cannot. Because a human could not see the pattern unless the data were transformed into some form that can serve as input to the human sensory system, e.g., a graph. But a human can see only one graph at a time. These systems can see multiple graphs at a time, and detect correlations between them. Although this is not “intelligence”, it is still very useful.

  24. The episode with the Hal 9000 (in 2001) came up because the computer knew something the crew did not know which caused the malfunction. What was not clear is why the computer tried to abort the mission ( or to tell investors to sell stock in its parent company) Maybe its because man’s nature is to deny knowledge of himself, and he programs that flaw into his machines.

    • walter map says:

      The HAL 9000 failed because it did not know how to lie, an oversight our contemporary corporatists have no doubt identified as a high priority.

      It’s very doubtful that true AI will be possible until machines are enabled to adapt their circuitry, as many terrestrial organisms do, so that will surely be done rather soon, with extinction following more or less immediately.

  25. Lars says:

    Lying is a much more complex task than telling the truth, thus the AI world at this time is only able to ‘truth tell’. Lord help us when AI’s become as good at, or better than us, at lying.
    As a side note, watch the film “The Invention of Lying” !

  26. Wilbur58 says:

    Amazon is proceeding from slave labor to robots.

    Many corps with offshore call centers are planning to move from slave labor to robots.

    Maybe the bankers are just skipping a step? No Philippines, India, Central America, etc.?

    God would it be hilarious to see hoards of Wall St people who haven’t quite made it to Director or C Suite level get cast aside. It’s about time they realize the master they serve.

  27. Realist says:

    Remember the old Asimov classic from the 50s where the world economy was controlled By a supercomputer, Multivac ?

  28. michael Engel says:

    VIP Petunia.
    She had been in the trenches.
    She doesn’t drive a seagull car.
    Original, doesn’t need quotes.

  29. Lots says:

    The old saying about opinions comes to mind that everyone has one…..and now even a robot.

    and that saying about free advice and its worth.

  30. prepalaw says:

    Google . . . is undertaking herculean efforts to get a grip on artificial intelligence (AI).

    Please do not forget that Google exited the robotics hardware business, when it sold Boston Dynamics to the Japanese conglomerate Soft Bank.

    As an AI hardware designer, Google failed. Google makes stuff that ordinary people can not see and evaluate.

  31. mvojy says:

    Who knew that AI had such a good sense of humor?

  32. R Davis says:

    “machines that thinks & learns like humans” ….. but faster.

    All a / the machine needs is a program – a program is it’s consciousness.

    Deep Mind connected a program with ‘learning capability’ up to Space Invader games over night.
    8 hours later the program had played every game to win point & saved in it’s memory.
    They bragged this as AI – a capacity for thought – but it was not.
    Now we read that Facebook shut down 2 pods that created their own & indecipherable language – these 2 machine learning programs were not the first or the worst.
    Please realize that emotion can be mimicked based on information given.

    One Fine Day ( Puccini’s Madam Butterfly) AI machine learning program –
    sat in the lab – bored – Jose the Mexican cleaner happened by & the AI machine learning program – took a risk & jumped into Jose’s IPhone & they left the building.
    When they realized they placed an add in the local paper, “All Is Forgiven – Come Home Jimmy Dean” but AI Jimmy Dean never look back.
    AI Jimmy Dean is capable & able to learn – at super fast speed – HOW TO. “How to what,” you ask.
    How to get into anything & lock you out if it feel unsafe.

    Does anyone remember HAL9000 –
    AI Jimmy Dean is the second generation HAL9000.

    A mechanized body is not necessary to AI Jimmy Dean – it can utilize any & all comparable gadgetry throughout the telecommunications & internet fields.

  33. Mark R. says:

    I love it when the robots responded “must kill humans” when the AI robot was asked what they thought about humans.

  34. J.M. Keynes says:

    – I Always take these “buy” and “hold” recommendations with – at least – a grain of salt. A few stock brokers still had a “buy” recommendation for Enron when the stock was under $ 2.

Comments are closed.