Nassim Nicolas Taleb and colleagues present an (almost?) tautological view of the effect of GMO (PDF): the large areas of establishment, allergist human consumption and some unspecified risk mechanism (which does not seem to affect non-GMO crops, more about see next paragraph) may cause ruin to humanity, noun because, hey, they say so. I could come up with a similar scenario in which we should stop working on any new computing development, because there is a non-zero probability in which we may end-up with non-bottom-up tinkering causing the rise of the machines (a Black Swan event) that has ruinous consequences for humanity (with apologies to the Terminator). For some reason that escapes me, I do not find it compelling.
In making their case Taleb and colleagues also display a very naive view of non-GMO agriculture. Selective-breeding accelerated tremendously since the formalization of quantitative (statistical) genetics theory around the 1930s. Moreover, since the green revolution, the speed of creation of new cultivars, their geographical spread and the reduction of the number of genotypes commercially planted has accelerated. This all happened before the introduction of GMO as part of agricultural breeding programs. Using Taleb et al.’s terminology, crop breeding mostly abandoned ‘bottom-up modifications’ more than half a century ago. Therefore, the comparison between GMO and non-GMO crops is not one of ‘between tinkering with the selective breeding of genetic components of organisms that have previously undergone extensive histories of selection and the top-down engineering of taking a gene from a fish and putting it into a tomato’ but between high-speed, fast-deployment breeding programs and the same program with a few extra plant genes (sorry, no fish for you), sometimes even from organisms that could conventionally breed with each other (cisgenesis).
If we accept the statement ‘in addition to intentional cultivation, GMOs have the propensity to spread uncontrollably, and thus their risks cannot be localized’ and put it together with the total area of planted GMO (say 160 million hectares, or 10% of planted area, over the last 20 years) we face potentially billions of interactions between GMO and non-GMO plants. Given that there are no reported events of any large-scale disruption in billions of plant interactions—and relying on their reference to the Russian roulette fallacy ‘one needs a large sample for claims of absence of risk in the presence of a small probability of ruin’— this would support the idea that such probability of ruin is infinitesimally small indeed, as we are way past sample size n = 1. A similar argument would apply to the billions of meals ingested by millions of humans every day: there is no evidence of large (or small) scale effects of consumption of GMO on human health.
From a probabilistic point of view, absence of evidence is evidence (but not proof) of absence (see, e.g., this article). In the same way, we cannot negate the existence of the Chupacabra, but the lack of any reliable sighting puts its existence in the very highly unlikely basket. For any scientist finding real evidence of the deleterious effects of GMOs (or the existence of the Chupacabra) would be a dream come true: guaranteed article in Nature, lecture circuit, book contract, the works. Many are actively looking for negative effects for GMO, but no one has found anything beyond noise. The question is How many billion interactions between plants are enough to put the ruinous effects of GMO at the level of Chupacabra?
I agree with Taleb and colleagues that many of the arguments to advance the use of GMOs are naive or spurious. At the same time, I think ensuring cheap food access to very large populations require the use of many different combinations of agricultural technologies, including GMO and non-GMO crops. There are many problems with modern agriculture both in their industrial and small-scale versions, but I would posit the focus on GMO is more a distraction than one of the core issues.
P.S. 2014-07-27 19:00 The article has a Pascal’s wager feeling: there is a potential for infinite loss (eternity in hell, human ruin), finite—supposedly small—loss (pleasure, agricultural productivity) and quite a bit of window dressing to the basic argument: if we assume a technology has potentially ruinous effect (pretty much infinite loss) with probability greater than 0, then the decision to avoid it is the correct one.