For the first time in known history an algorithm has been given a seat on a corporate board of directors. I always suspected that algorithms would take over the world, but this is an unexpected development.
Deep Knowledge Ventures, a Hong Kong-based venture capital firm, appointed the algorithm, named Vital, to its board last week, according to Business Insider. Deep Knowledge is a firm that focuses on age-related disease drugs and regenerative medicine projects, and it wants Vital to help the board make investment decisions. Vital, (Validating Investment Tool for Advancing Life Sciences), was developed by Aging Analytics UK.
Is this a good idea, or a PR stunt that can encounter unexpected consequences? I am not suggesting that Vital will try and take over the world, or even the firm, but there are many examples of algos-gone-wrong - lessons from which its fellow board members may want to learn.
A fictional example is SkyNet of Terminator fame. In the movie, SkyNet was a computer system developed for the U.S. military as a global defense network, which had command over all computerized military hardware.
But SkyNet went wrong. When it began to 'learn' it eventually gained self-awareness, along with the paranoia and anxiety that plagues much of human kind, and decided that people wanted to destroy it. The consequences were not pretty, millions were killed, and SkyNet became the poster boy for algorithms-gone-wrong.
In April 2011 Michael Eisen, an evolutionary biologist at UC Berkeley, and one of his postdocs observed that a book on Amazon - Peter Lawrence's The Making of a Fly - a classic work in developmental biology - was priced at $1,730,045.91 (+$3.99 shipping).
The next day they checked again and found the book was priced at $2.8 million. Before anyone at Amazon noticed, the price rose to $23,698,655.93 (plus $3.99 shipping). Eisen and his postdoc realized an interaction of pricing algorithms from two competing booksellers caused the issue.
In August, 2012 one of broker Knight Capital's ETF trading algorithms lost control, spinning away to losses of $440 million in less than an hour, and nearly bankrupting the firm (it was subsequently sold).
But algos are not all bad, many can be used for good as well. Algorithms are used for online social activities; they can help you find friends on Facebook or to decide which films you might enjoy on Netflix. They can set you up with people that you may enjoy dating, or even marrying, like on Match.com. They can assist you in everyday Google searches, ensuring that you are directed to the most likely sources of information.
Still, it seems that algos can "break bad" just like people do, only with perhaps exponentially more potential for financial damage. Like SkyNet, Vital has been deployed to make decisions that will impact the fate of human beings. Deep Knowledge says that Vital will make investment recommendations about life sciences firms by poring over large amounts of data. Then Vital gets to vote on whether the firm makes an investment in a specific company or not.
So is Vital a good algo or a baddie? Most algos only break bad because of lapses in human judgment when designing them. Self-learning algos, which draw conclusions from incoming data, have even more potential for going rogue. Only careful planning and assessment of possible unintended consequences can prevent an algo from making expensive mistakes.
In Vital's case, he is only being asked to make decisions by scanning data on prospective companies - looking at their financing, clinical trials, intellectual property and previous funding rounds - then decide if they are a good investment. And he gets to vote on it, along with the human members of the board.
It could well be that Vital is a better board member than his colleagues. It's not as if human board members did a great job at managing their companies pre-2008 (or even post-). I saw an unofficial survey that said a potted plant would make a better board member than a human, and algos came second after the plant (to protect the innocent, we will not name the plant).
Vital could be the precursor for many more algos on boards around the world. The only caution is that, before the algo is promoted, the company needs to be sure what it is asking for.
Deep Knowledge's Vital reminds me of another powerful computer, Deep Thought of Douglas Adam's Hitchhiker's Guide to the Galaxy. Deep Thought was asked to provide the answer to Life, the Universe and Everything.
After the great computer program had run for "a very quick seven and a half million years" it answered: "42."
Which suggests that you really have to know what you are asking for.