Genetic Modification

Reading things like this article make me understand why Taleb comes across as so abrasive and frustrated in his books. Anyone saying “we’ve had X for decades, DECADES, and nothing bad has happened, surely this is evidence that it is fine” has completely and absolutely missed the entire reason that heavy tailed distributions are problematic. Say it with me one more time –

Sampling from heavy tailed distributions looks just like sampling from thin tailed distributions until an event in the tail occurs.

We have had the internet for decades, and everyone thought it would usher in a new era of knowledge sharing and democracy. Few predicted that it would so effectively facilitate the death of quality journalism and the rise of populism and fake news.

We have had agriculture for thousands of years, but it is only in the last 50 or so that the dangers of monocultures have become obvious, following the Fusarium fungus wiping out the Gros Michel banana.

Reducing Correlation Reduces Systemic Fragility

I have heard pro-GM advocates rail against GM labelling on food packaging – “by labelling it, you are singling it out as something for consumers to avoid – there is nothing wrong with GM, so consumers avoiding it are irrational, and we shouldn’t enable them”. The counterargument is that by labelling it, you give people the choice, which reduces the systemic fragility somewhat. If you don’t give people the choice, you correlate all outcomes, which means that if there does turn out to be a problem, it affects everyone.

That is not to say that it wouldn’t affect everyone in some way – the collapse of Lehman Brothers negatively affected most people in the world in one way or another. However, with the choice of whether to invest your life savings in Lehman Brothers or not, the people that chose not to were far less negatively affected than the people that chose to.

Similarly, people had the choice of whether to create a Facebook account or not. Few predicted that it would be such a source of stress and anxiety for many people, and few predicted that it would be so damaging to public discourse. Through the way it has impacted our democracy, it has negatively impacted everyone, but those people that chose not to create a Facebook account at least happen to have insulated themselves from a source of stress, anxiety and fake news.

Taleb is not saying “do no GM ever”, just as he is not saying “ban Facebook” – he is simply saying “move slowly and don’t make massive systemic changes, especially if it uses new technology”. Gradual changes are safer, and older technologies have been more thoroughly tested. Since the Thalidomide disaster, new drugs are never approved for pregnancy – they are not even tested on small numbers of pregnant people. The way we get new drugs for pregnancy is through people accidentally using drugs when they don’t even realise themselves that they are pregnant, and hospitals recording the outcomes. This takes years to collect sufficient data to give us confidence that a drug is safe during pregnancy, but we get there because the data does accrue, slowly but surely. Maybe some miracle drug could save lives during pregnancy, and we aren’t even testing it – how inefficient! The Thalidomide disaster has taught us to be cautious however – progress takes time.

The Rule of Two Generations

Stepping away from systemic issues slightly, to look at technological ones – a fairly good rule of thumb for technological safety is “wait 2 generations” – if nothing bad happens to the children of the people that grew up with the technology, it is much more likely to be safe. Nothing is ever guaranteed of course, but this covers a wide range of ways in which subtle problems can manifest themselves. We are about at this point with the internet itself – people that cannot remember a time before the internet are having children. Social media is not at this point yet and is already showing problems. GM is realistically at a similar stage to social media – yes it has “been around” for decades, but the general population has not had access to it for long enough for any adults to not remember a time before GM products were available in supermarkets.

If we continue to do GM research, and some companies release clearly labelled GM products, eventually we will reach a stage where some people have eaten GM food all their lives, and their children are eating GM food too. If you can’t think of any issues that could manifest themselves in this time, you are not thinking hard enough:

1. Direct issues with GM food being somehow toxic or damaging – this one is unlikely, but still not impossible. What if some by-products of the DNA editing process stick around in the organism’s cells and are able to be replicated? What if these are able to make random edits to people’s DNA once ingested? Again – unlikely, but then prion diseases were completely unheard of 30 years ago.
2. New opportunities for disease – what if viruses or bacteria are able to somehow piggyback on the process, gaining antibiotic resistance or other dangerous features?
3. Extreme monocultures – what if the companies growing these crops fail to learn the lessons from history, and produce even more monocultural plantations, crowding out more diverse but “less efficient” cultivars? We could legislate against this, but would the laws be effective?
4. Intellectual property abuse – would farmers end up trapped, having to buy seeds from a GM supplier, unable to grow crops from their own seeds? This was one of the reasons why Haitian farmers burned the seeds given to them by Monsanto after the earthquake in 2010 – these seeds would have grown into seedless cultivars, requiring them to buy more seeds the following year, rather than being able to harvest a small portion of their crop as seeds, allowing them to be self-sufficient.

As I said, Taleb is not saying “do no GM ever”, he is saying “move slowly and don’t make massive systemic changes, especially if it uses new technology”. This is not a Pascal’s wager like scenario in which we are using a small risk of something arbitrarily bad happening to forever ban a particular action. We may end up using GM a great deal when we are more confident in the technology and the society that has grown around it. The argument is not “don’t do this, as it has a risk of infinite downside”, it is more like “if you want to do this, make sure you do it slowly and carefully on a small scale, as doing things quickly or on a big scale has a much higher risk of extremely large downsides”.

Good Intentions are Rarely Sufficient

If we allow ourselves to consider the possible downsides for large systemic changes such as encouraging impoverished regions of the world to switch to the GM produce, the number of things that could go wrong multiplies up enormously – not only do all of the possible problems above become much more serious when the size of the roll-out is increased, but we have to consider other more societal issues:

1. Would the proposed solution actually increase these countries dependence on the developed world, rather than decreasing it?
2. Would such a large change destabilise countries economies, leading to political instability?
3. Does this allow us to ignore other problems that we would be forced to solve otherwise, such as women’s access to education and contraception, the developed world’s wasteful consumption of imported food and the exploitation of developing countries in general?
4. Related to the last point – would it even solve the problem in the long term? These countries are currently in a Malthusian trap, where they cannot grow enough food to feed their population, but having children is the retirement plan, so paradoxically, fewer children results in more likelihood of starvation. Increasing the food supply without solving any other issues (women’s education and access to contraception being key ones) could result in people continuing to have many children, and within a couple of generations the population would be sufficiently large that it would be back at the level of subsistence again. Nothing would have changed except that rather than 1 billion people living in poverty and starvation, you would have 10 billion.

Slow and Steady Wins the Race

Although we have a nice shiny tool that looks like it can solve a really serious problem, we just need to be careful and not get carried away. If it still looks shiny in 34 years (60 years or 2 generations since the first genetically modified food – the Flavr Savr tomato in 1994), we can look at piloting a couple of small programs in developing countries – at least then we can’t credibly be accused of “experimenting on the third world”. 30 years after that, a whole country might decide to implement some sort of program, based on the success of the pilots. This is a systemic change for the country, so quite fragilizing for them, but less so for the world or humanity in general. If this doesn’t result in war, societal collapse, Malthusian population growth or some other calamity within 30 years or thereabouts, other countries might be well advised to get in on the action. At any point in this however, problems could be noted, and programs could be rolled back. Pro-GM people get what they want, and they only had to be patient for 94 years or so!

Fundamentally, heavy tails are not about variance – we understand variance very well and it doesn’t surprise us. As Taleb has written several times, it is often in trying to reduce variance within a system that we end up with a system featuring heavy tails – the system is much more predictable with lower variance, with the unfortunate downside that a completely catastrophic event will happen eventually. Heavy tails are also much more common in complex interlinked systems, and I can think of few things more complex and interlinked than genetic modification in the food chain. It features DNA, which we still don’t fully understand, geopolitics and macroeconomics, which we don’t understand nearly as well as we think, people’s behaviour, which is often bizarre, and food, which is absolutely critical for our day to day survival.

Seriously – chill out pro-GM people. Being cautious about genetic modification is not the same as being an antivaxxer.