Skip to content
 

“There is this magic that our DNA enables”

I was just at a talk where a computer scientist was making dramatic claims for the value of human decision making. This is typically a crowd-pleasing position to take—after all, we in the audience are humans and we want to hear how great we are.

What really riled me was when the speaker said, “There is this magic that our DNA enables . . .” as a way of saying why humans are so important.

As I wrote a few years ago, it used to be that humans were defined as the rational animal, and now we’re defined as the irrational computer.

It used to be that what we, humans, had to offer in the world was our rationality; now it’s our irrationality that’s valued.

I think the crowd-pleasing position taken by today’s speaker wass just B.S. Don’t get me wrong here, I think human decision making is important; I don’t think we can or should let computers make all our decisions for us.

But let’s not get mystical about it.

“There is this magic that our DNA enables . . .”: This seems to me to be a bizarre sort of techno-mysticism.

The funny thing is, when the speaker was talking about specific research and development, everything said was reasonable. My problem was only in the generalities.

19 Comments

  1. Yes, “DNA” has become a buzzword, and as you point out, “the magic of our DNA” seems like a kind of mysticism. Chesterton would have savaged the phrase with glee.

    I am glad that the substance of the speech made sense. In similar situations, I have had to remind myself that people who refer to “our DNA” (when *not* talking about genetics) were not necessarily talking nonsense; they had just fallen for an unfortunate fad on the margins of their thought.

    But a fad it is. I imagine if one were to tally the TED talks–not about genetics–that refer to “our DNA,” the count would be high.

  2. Eliot J says:

    Agreed, the messianic BS coming out of the CS/ML/NN/AI camp is way overblown. Melanie Mitchell’s just published book goes a long way towards deflating much of the hype, Artificial Intelligence: A Guide for Thinking Humans. She describes it as Sober AI…just wish there was an AA-type program for these evangelists to join and get deprogrammed.

  3. There is the magic of fairness.

    In the ultimatum game, person A gets to say how an amount is split up: x for A, 1-x for B. B gets to say whether the deal is acceptable. If not no one gets anything, if yes the money is split as A proposed. https://en.wikipedia.org/wiki/Ultimatum_game

    It would be rational for B to accept anything. Getting something is better than getting nothing. So it would be rational for A to offer as little as possible.

    Back in the real world, we have the magic of fairness and if A offers less than 50%, A has to seriously worry getting nothing.

    So humans do make different decisions than computers programmed to do the optimal thing.

    • Andrew says:

      Victor:

      You can program a computer to have a norm of fairness.

    • Victor, what’s rational in a single game is not necessarily rational in an iterated game. If you are going to play games like this all day long, penalizing a person for being unfair will eventually lead to that person being closer to fair, which leaves person B in a better long run iterated average.

      I took math classes on “artificial life” from Daniel Ashlock who was at Iowa State but is now in Guelph, he did a bunch of artificial life evolution of strategies for playing various games. You’d be surprised at how effective various odd strategies are… Like populations would evolve that learned a “secret handshake” if they played the game they’d always do some pattern of accept/defect for say 10 rounds, and if the other strategy they were playing against did the same thing they did, then they could assume the other strategy was the same as them, and start playing an optimal shared strategy. It all evolved through tournaments and breeding and selection.

      Eventually, you’d find some parasite that learned the secret handshake and then played the optimal personal strategy that hurt the other strategy, and would begin to kill off the secret handshake club… etc etc

      • The ultimatum game is a wonderful tool to explore the altruistic nature of humans. Yes, it matters how often the game it played. If you play it multiple times, in the beginning higher offers will be rejected than at the end or in a one-off game.

        There is a similar common goods-game where iterations are important. Everyone can put something into a common pot. This amount is then doubled and distributed equally over all players. People tend to start with putting a large part of their money in the common pot, but as soon as someone thinks they are smart (rational) and give less and still gets about the same, the collaboration collapses. If you give people the option to punish free riders at their own expense, the collaboration is maintained. People are willing to altruistically punish people who free ride on the others; in this game people at typically able to forgo getting 1 dollar to punish the free rider by one 1 dollar or even 50 cents. https://en.wikipedia.org/wiki/Public_goods_game

        The ultimatum game has been played in many cultures. In poor countries it is possible to play the game for relatively large amounts (a month of income); also in that situation people reject unfair offers, although the percentage that is accepted does drop when the amount is very higher. There are cultures where you exert your dominance over others by gifting them; here people regularly make offers above 50% and these offers are regularly rejected.

        • jim says:

          These games are interesting but they’re conducted in a sterile environment that’s not relevant to the real world.

          First, what’s rational with money you’ve just been handed as a gift isn’t necessarily what’s rational with money that you busted your ass to earn. Free money is ez come ez go. Founder billionaires mind their money, but the children, who never had to earn it, often don’t.

          The collective goes to hell in unions, where older members advocate for themselves, frequently agreeing to additional production mechanization in exchange for higher wages, which happens at the expense of entry level workers and creates a decline in union membership, which ultimately reduces the bargaining power of the union.

          Also personal relationships will long outlive any one round of a “game” so in the real world people may play for personal loyalties that have a higher benefit over the long term rather than try to maximize the benefit of a single transaction.

  4. jim says:

    +1 Andrew.

    Actually humans are pretty bad decision makers and frequently ignore or decide against an immense weight of data that’s staring them right in the face.

    Did you ever read the “13.7” Blog at NPR? One of the writers was Stuart Kaufmann, and evolutionary biologist and theorist. This guy is freakin’ brilliant, OMG. But he was also very passionate about certain things and he’d blow me away by developing some long complex argument and come to a conclusion that was the exact opposite of where his data pointed! Wow, even the most brilliant people can be blinded by their passions.

    You know what Neil Simon said, right? :)

    • jim says:

      I think the “humans are great decision makers” argument gets trotted out a lot by people who don’t like the results of a dispassionate analysis.

      • Curious says:

        Data does not tell us anything without interpretation. Causal analysis requires human intervention in the form of critical reasoning. The notion that one can engage in dispassionate analysis without human intervention is simply not true. Framing an algorithm as dispassionate is a popular notion, but it does not pass critical scrutiny.

        • jim says:

          “The notion that one can engage in dispassionate analysis without human intervention is simply not true. “

          This is absolutely and unequivocally false. Humans must generate the method, but once the method is created, tested and verified against new data, no further human input is required. In other words: the data can be used and the analysis completed with no human intervention except to gather the data.

          • Martha (Smith) says:

            Jim said,
            “Humans must generate the method, but once the method is created, tested and verified against new data, no further human input is required. “

            No — human intervention is also needed to decide whether or not the new data are appropriate for the previously created method.

  5. Mikhail Shubin says:

    > It used to be that what we, humans, had to offer in the world was our rationality; now it’s our irrationality that’s valued.

    There is a quote I heard long time ago what goes like this: “Statistics wants to make humans think like machines and AI wants to make machines think like humans”.

Leave a Reply to Mikhail Shubin