Slashdot Mirror

← Back to Stories (view on slashdot.org)

GV, Formerly Known as Google Ventures, For Years Has Used an Algorithm That Effectively Permits or Prohibits Both New and Follow-on Investments (axios.com)

Posted by msmash on from the closer-look dept.

Dan Primack, reporting for Axios: When most venture capitalists want approval to make a new investment, they go to their partners. When venture capitalists at GV do it, they go to something called "The Machine." Axios has learned that the firm, formerly known as Google Ventures, for years has used an algorithm that effectively permits or prohibits both new and follow-on investments. Staffers plug in all sorts of deal details into "The Machine" -- which is programmed with all sorts of market data, and returns traffic signal-like outputs. Green means go. Red means stop. Yellow means proceed with caution, but sources say it's usually the practical equivalent of red. It was initially designed and used as a due diligence assistant that could be overruled but, according to three sources, it has evolved into a de facto investment committee.

40 of 74 comments (clear)

  1. Responsible for the chaos in Google's messaging? by bogaboga · · Score: 2

    When venture capitalists at GV do it, they go to something called "The Machine." Axios has learned that the firm, formerly known as Google Ventures, for years has used an algorithm that effectively permits or prohibits both new and follow-on investments

    You see, I confidently believe that you all agree that Google's messaging paradigm is in and remains in "chaos/confusion/mayhem/tumult land" right now.

  2. Re:Responsible for the chaos in Google's messaging by zlives · · Score: 1

    ding ding ding ding

  3. But does it work? by orev · · Score: 2

    It doesn't really matter unless there is data on whether the system works or not. Nothing in the article seems to mention this.

    1. Re:But does it work? by stephanruby · · Score: 1

      My brother did the same thing with Excel. Green, the project was good. Red, the project was a no-go. He designed the criteria himself, so there was nothing groundbreaking

      It was just a way to keep himself honest. Too many times, he got too emotionally invested in a transaction and lied to himself about the numbers. By deciding on the criteria beforehand and codifying the rules in a spreadsheet, it helped him see things more clearly and it made it easier for him to walk away.

    2. Re:But does it work? by dj245 · · Score: 3, Insightful

      It doesn't really matter unless there is data on whether the system works or not. Nothing in the article seems to mention this.

      This is actually 2 questions. Whether it worked in the beginning, and whether it works now.

      There is presumably no data available for initial or follow-on investments that were not made. The investment either wasn't made at all, or it was invested by someone else so Google probably doesn't have good or any data. Therefore, any adjustments to the algorithm based on greenlighted projects that then failed will probably overfit due to incomplete data on yellow or redlighted projects that could have succeeded. Unless Google deliberately invests in Yellow or Red projects to test the algorithm, the algorithm may get more and more conservative over time.

      --
      Even those who arrange and design shrubberies are under considerable economic stress at this period in history.
    3. Re:But does it work? by Aighearach · · Score: 1

      Well, we know it doesn't have any inputs about, "Would it piss our users off if we change course abruptly? Did we make any representations to users about if they should use this, or avoid it because it is temporary?"

      Or, to put it in terms bean-counters can understand: It doesn't even have an account to track Goodwill provisionally earned from past decisions!

    4. Re:But does it work? by 91degrees · · Score: 1

      yea, the data is there. and, yea. it's obviously working.

      Well, no. How does it compare with the red, or yellow ventures? How does it compare with a monkey picking randomly?

  4. In related news ... by fahrbot-bot · · Score: 2

    ... an algorithm that effectively permits or prohibits both new and follow-on investments.

    My parents, friends, wife, etc... sometimes say "yes" or "no" to things I want to do and/or do again.

    [ Is the problem here that's it's an "algorithm" (presumably written by people) or that it made decisions? ]

    --
    It must have been something you assimilated. . . .
    1. Re:In related news ... by suutar · · Score: 2

      from the summary, it sounds like the concern is that it's being treated like an infallible oracle. It may not have come up with a significant false positive yet, but there's no way to tell if it's had false negatives, and unlike human advisors it's probably not very good at explaining why it came to a given conclusion (unless you're willing to trace the code while it processes something).

    2. Re:In related news ... by sinij · · Score: 1

      The problem is that we replacing irrational but possible to understand human decision making with an opaque algorithm written by irrational humans.

      It is not as easy to detect and correct flaws in decision making.

    3. Re:In related news ... by mccalli · · Score: 1

      An infallible scapegoat. If it says no and you proceed, your judgement on the line. If it says yes and you proceed but it fails...well, you did all your diligence, right? Not your fault...

    4. Re:In related news ... by alvinrod · · Score: 1

      I'm not even sure human advisers are particularly good at explaining their rationale. Sure they can give you an explanation, but I'm not sure if it's actually any good. Silicon Valley angel investment has terribly low success rates, even among investors that tend to outperform the average. I suspect it has less to do with picking winners and more to do with having the winner you pick providing massive returns.

      If as you point out the only concern is just how good the system is, just consider every case where it disagrees with human advisers and see which is right more often. Maybe it's inline with human predictions to such a high degree that no one really notices the fringe cases.

    5. Re:In related news ... by fahrbot-bot · · Score: 1

      ... infallible oracle ...

      Words not often used in the same sentence - no matter how much Larry Ellison would like it.

      --
      It must have been something you assimilated. . . .
    6. Re:In related news ... by ShanghaiBill · · Score: 1

      there's no way to tell if it's had false negatives

      if GV declines to invest but other VCs invest successfully, then that is a clear false negative.

      it's probably not very good at explaining why it came to a given conclusion

      As long as it works, it isn't really necessary to get an explanation. Plenty of decisions are based on statistical analysis of outcomes, rather than understanding complicated internals.

      A human VC may give you a long winded rationale for an investment, but the real reason is that he is having an affair with the target's CEO, or that he made a side investment with his own money. The algorithm works because it bypasses all the emotion and ulterior motives.

    7. Re:In related news ... by rgmoore · · Score: 1

      False negatives are much less of a big deal in investing than false positives. One of the basic problems with venture capital is that there are way more investment opportunities than any firm has money to invest. That means you're always going to have missed opportunities no matter how smart your decision making process is. As long as you find enough good opportunities, it doesn't really matter than you missed others.

      --

      There's no point in questioning authority if you aren't going to listen to the answers.

    8. Re:In related news ... by datavirtue · · Score: 1

      It is gambling.

      --
      I object to power without constructive purpose. --Spock
    9. Re:In related news ... by datavirtue · · Score: 1

      I bet if you said "Oracle phallice" Larry would not be able to keep from smiling.

      --
      I object to power without constructive purpose. --Spock
    10. Re:In related news ... by datavirtue · · Score: 1

      One could are that false negatives are worse than false positives. Sure, a false positive means that your money goes somewhere else and that sucks for you but the fact remains that capital entered the market and took on life. While a false negative means that capital remains dormant and an opportunity for economic growth was lost--bad for everyone.

      --
      I object to power without constructive purpose. --Spock
    11. Re:In related news ... by Aighearach · · Score: 1

      Traditionally, an Oracle is a religious designation for a type of person who can listen to the words of the Gods, and translate them to human speech.

      You don't do what the Oracle tells you because the Oracle is infallible, for the Gods are themselves neither infallible, nor guaranteed to be honest. You do it because you traveled all the way to the temple, already made the required donation, and it will make the Gods angry if you do not then heed their words. That's true regardless of if they chose to give you a useful answer, or not.

      When the Oracle at Dephi was asked (by another person) if Socrates was the wisest man alive, and the Oracle said that Yes, he was, Socrates did not take this as proof of his wisdom; he instead sought out to disprove it! He felt that it was an affront to his dignity; the Gods were not speaking Truth through the Oracle, but merely making a joke of him. In the end he discovered the basis of science; that he was completely ignorant of everything other than his own ignorance; and that no purported wise men knew more! In fact, most were unaware even of their own ignorance. And so, by virtue of being the most aware of his own ignorance, the words of the Oracle might actually be true. And he was clearly a strong believer in the Gods; when his friend Xenophon asked his advice about joining a mercenary band, he advised to consult the same Oracle. And when Xenophon returned having asked merely which God to give sacrifice to in order to assure success, he chastised him for asking a leading question instead of asking if he should even go at all.

      They don't need to do what the algorithm says because they believe it is infallible; if they treat it as an Oracle they merely need to be fearful that lack of respect for the System might have negative personal consequences, or believe that human wisdom is insufficient to provide adequate guidance.

    12. Re:In related news ... by Aighearach · · Score: 1

      False negatives are much less of a big deal in investing than false positives.

      So you're saying that $1 of opportunity cost is less than $1 of operational losses?

      If you said the exact opposite, it would be more true, because $1 of losses is a write-off, and $1 of opportunity cost is not. Opportunity cost is market share you'll never have another chance at!

    13. Re:In related news ... by suutar · · Score: 1

      Fair point. Unjustified reverence for computer output has been common for long enough that I assumed that to be the case and did not consider that they might think it could be wrong but it's not worth the effort to disagree.

  5. Re:hmmm surprised... by DontBeAMoran · · Score: 1

    If 42 is an optional answer then something is very wrong with the Universe.

    --
    #DeleteFacebook
  6. If word in buzzWordDB by ark1 · · Score: 1

    color == green

  7. Juicero by Swistak · · Score: 3, Interesting

    Wasn't that the fund that put shittons of money into a Juicero?

    1. Re:Juicero by HornWumpus · · Score: 2

      From: https://www.bizjournals.com/sa...

      The San Francisco-based company is backed by plenty of Silicon Valley investors including Kleiner Perkins Caufield & Byers, GV and DBL Partners. Other investors include Thrive Capital, Campbell Soup Company, Acre Venture Partners, Two Sigma Ventures, and First Beverage Group.

      Nice job 'redistributing the wealth', Juicero. But that's giving Juicero too much credit. They were apparently not 'smart cynics', but were 'dumbass vegan true believers'. They apparently spent the money overbuilding the juicer, rather than just taking it and running as they should have.

      --
      John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
    2. Re:Juicero by HornWumpus · · Score: 1

      WTF? It's clearly possible. Just a waste of time and money.

      Nobody wants home squeezed $8/half cup juice.

      Watch the Ave teardown. They really did spend _stupid_ money building the juicer. Hardened steel gears! Morons.

      --
      John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
    3. Re:Juicero by datavirtue · · Score: 1

      The juicer comes to life and decides it no longer wants to be a juicer. Appealing to your ten your old daughter for freedom she lets the cat out of the bag before you have a chance to intervene. In the resulting drama somehow the family dog ends up "juiced." I think we all know how this ends. Eventually the last of the humans escape and battle their way through the galaxy to seed another planet called Earth. ...and here we are again.

      --
      I object to power without constructive purpose. --Spock
    4. Re:Juicero by Aighearach · · Score: 1

      I thought the purpose was to redistribute high quality machine parts of a size useful to the maker community to the salvage market?

    5. Re:Juicero by Aighearach · · Score: 1

      A few years ago I spent $329 on a horizontal masticating juicer, from an existing brand that has been on the market for decades and has a late-generation product with a 15 year warranty.

      If you want one that is a "cold press" instead of "masticating" juicer, you'll probably spend around $1500 or more. But the warranty will be good, and those are made by companies with long presence in the commercial kitchen machine sector.

      Juicero, OTOH, was trying to sell packets of vegetables, with a model like the single-serve coffee pods. But to get it to actually work as quickly and conveniently as they wanted, they had to prep the vegetables before putting them in the pack; and then you don't even need the fancy machine. And yet, the machine was engineered to be able to really do the job; so it is expensive.

      The advantage over a horizontal masticating juicer is simply that you save the 2 minutes of cleaning that is required immediately after using the machine. (If you don't take a masticating juicer apart and clean all the parts right away, it gets hard to clean it at all; that's the whole reason for there being a potential niche here) The problem is, the juice packets are them more expensive than just buying bottles of premium organic vegetable juice. Most of the value in spending $300+ on a masticating juicer is that you can get premium-quality juice out of any sort of vegetable including greens; the cheap juicers are "centrifugal" and only work well for certain things like carrots and apples, but can't really get juice out of greens. If you drink vegetable juice even once a week, you're saving money by buying the juicer. 12 ounces of carrot juice might cost $3 or more, but 10 lb of carrots

      A masticating juicer can also make almond milk from almonds, peanut butter from peanuts, tahini from sesame seeds, etc. Even sorbet from frozen fruit. Tahini alone, it costs like $7 for a small jar, the same weight in sesame seeds costs $2. Add a few cents of olive oil, and that's close to $5 saved. It pays for itself if you don't mind a mandatory 2 minute cleanup routine that has to be completed before the residue dries.

      So obviously, if all it does is make premium juice then the consumables have to beat the price of premium bottled juice, since that is the only product it is really competing with. Just like coffee pods are not really competing with brewed coffee, but with coffee shops. Even for businesses providing coffee in customer waiting rooms; the choices that customers appreciate are either the pods or a coffee service that delivers branded coffee to carafes daily. If they instead just have an employee brew a pot of coffee and leave it sitting on a burner, very few customers drink it, and none are impressed or feel like you provided any luxury or convenience. Most likely it is still half full when somebody turns off the pot and throws away the remnants because it smells so burnt.

    6. Re:Juicero by Aighearach · · Score: 1

      If you adapt it into a new Stargate spinoff it might be good.

      I'd definitely rent the DVD from the library in 5 years, or just download it.

      --NoTVGuy

    7. Re:Juicero by Aighearach · · Score: 1

      Sorry, it was supposed to be: "but 10 lb of carrots... only costs $4."

    8. Re:Juicero by HornWumpus · · Score: 1

      Just one point. As the Onion says. It's no longer 'Almond Milk', it's now 'nut sweat'. Please adjust your usage.

      --
      John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
    9. Re:Juicero by Aighearach · · Score: 1

      If you steam the liquid out, surely it is true, all the more reason to use a masticating or cold press juicer.

  8. Meh by TheCastro1689 · · Score: 2

    People are often afraid to make their own decisions. I'm not surprised by this though considering how often banks rely on a computer program to approve or deny each of their products. People make almost zero decisions at many levels.

    1. Re:Meh by datavirtue · · Score: 1

      Consult the machine motherfucker!!

      --
      I object to power without constructive purpose. --Spock
    2. Re:Meh by datavirtue · · Score: 1

      Actually, my credit union uses human decision making. I know because they often come out to talk with me and make sure I'm white.

      --
      I object to power without constructive purpose. --Spock
    3. Re:Meh by Aighearach · · Score: 1

      They probably only check the box, and then let the computer "decide."

  9. Re:I do the same thing... by datavirtue · · Score: 1

    I was going to say, perhaps it should always return yellow. What an amazing invention...they should patent it.

    --
    I object to power without constructive purpose. --Spock
  10. Re:Should I stay or should I go? by datavirtue · · Score: 1

    No...that was irrational dumb Marxists calling the shots.

    --
    I object to power without constructive purpose. --Spock
  11. What are the chances.. by trawg · · Score: 1

    ... that this is just a spreadsheet?