With the development of “intelligent” algorithms and, more generally, smartness at all scales—from smart energy grids to smart buildings and smartphones—the dominant ideology in our societies rather praises the benefits of optimization in all sectors. This trend may have started with the invention of farming tools and may have been fueled by Taylorism and the standardization of products, processes, and tasks in assembly lines. Today, while our algorithmic world may let us believe that our contemporary technosphere becomes more personalized and provides customized services to all, it does not really bifurcate from a trajectory towards more optimization. Instead, new niches are found where products can be sold more efficiently to new consumers, at a lower financial, and sometimes environmental, cost. Optimization just got even more optimized.

Here we propose to question this paradigm. To do so, we take some inspiration from the biosphere. Biology offers multiple examples of systems that are apparently extremely efficient. For instance, the motor that set bacterial flagella in motion has been compared to a human-engineered motor, with a rotor operating at approximately 10,000 rpm, allowing a remarkable motion of up to sixty cell lengths per second! 11R. M. Macnab, “How Bacteria Assemble Flagella,” Annual Review of Microbiology, vol. 57 (2003): pp. 77–100; G. Lowe et al., “Rapid Rotation of Flagellar Bundles in Swimming Bacteria,” Nature, vol. 325, no. 6105 (1987): pp. 637–40; M. Silverman et al., “Flagellar Rotation and the Mechanism of Bacterial Motility,” Nature, vol. 249, no. 452 (1974): pp. 73–4; H. C. Berg and R. A. Anderson, “Bacteria Swim by Rotating their Flagellar Filaments,” Nature, vol. 245, no. 5425 (1973): pp. 380–2 However, this engineering view of biology may be too one-sided. In fact, it masks a larger, deeper, trend: most biological systems are suboptimal. Let’s illustrate this statement with a few examples.

The theory of evolution by Charles Darwin is in essence based on heterogeneity, i.e. on differences between individuals. Selecting the most adapted individuals to a given environment implies that not everyone is optimally adapted. The source of such suboptimality resides in part in the presence of random mutations in the populations, preventing homogeneity and providing a medium for natural selection. This also applies to selective breeding: only a diverse population can provide new individuals with advantages and increased fitness to their environment (or apparent benefits in agriculture). In short, from an evolutionary point of view, an element of randomness, and thus inefficiency at the population scale, is the price to pay to generate adaptation, and thus resilience, to variable environments.

Similarly, in agriculture, it is well known that clones—i.e. genetically identical individuals—are very sensitive to pathogen attacks at the population level, simply because the absence of diversity hinders the resistance of the population in the long term. Beyond the fact that homogeneous populations would die homogeneously when attacked, an element of heterogeneity also provides new interactions, and potentially new forms of resistance. This in fact is one of the driving forces for the promotion of agroecology: mixing species in a field provides synergistic resistances to pathogens. Conversely, in contrast to monocultures in fields, human populations remain genetically diverse and large pandemics always spare individuals lucky enough to have the “right” genetic background amenable to sustain viral attacks. Although the price to pay for genetic diversity at the population scale is the possibility of getting kids with a genetic disease or handicap, this suboptimality turns out to be an effective way to maintain a human population, even after the terrible black death pandemics in the fourteenth century or Spanish flu in the twentieth century.

A classic case for such suboptimality in human health is that of sickle cell anemia. In some areas of the world where malaria is frequent, 10 to 40 percent of the population also carries a mutation that triggers sickle cell anemia. This apparent injustice has now been explained: sickle hemoglobin (which causes the sickle cell anemia syndrome) promotes the expression of a protein called heme oxygenase-1, which produces more carbon monoxide, a molecule that protects against malaria.22A. Ferreira et al., “Sickle Hemoglobin Confers Tolerance to Plasmodium Infection,” Cell, vol. 145, no. 3 (2011): pp. 398–409. Here again, an apparent disadvantage also provides a positive gain, in a suboptimal scenario.

Fever provides another example of suboptimality that everyone can experience. Most of our human enzymes exhibit their highest activity near 38°C, not 37°C. This behavior may have been selected during evolution to allow that extra boost that is much needed when the environment, in the form of a pathogen attack, becomes less favorable. In other words, being suboptimal most of the time may be seen as a disadvantage when one is healthy, but reveals its benefit when sick.

Now, one may think that all this relates only to populations and thus that suboptimality in biology is rather restricted to global behaviors. Is biology also suboptimal at the microscale? Recent discoveries in the field of developmental biology rather support that hypothesis. For instance, it is well established that shape changes during embryogenesis rely on asymmetries, in the form of the outgrowth of groups of cells (e.g. during the formation of new limbs), the death of selected cells (e.g. during the formation of fingers), or cell contraction and tissue folding (e.g. during the formation of the neural tube). There is now evidence that such events are not the consequence of fully stereotyped processes, but instead are primed by existing heterogeneities between cells.33See, for example, D. K. Singh et al., “Patterns of Basal Signaling Heterogeneity Can Distinguish Cellular Populations with Different Drug Sensitivities,” Molecular Systems Biology, vol. 6, no. 1 (2010): p. 369; M. F. Wernet et al., “Stochastic Spineless Expression Creates the Retinal Mosaic for Colour Vision,” Nature, vol. 440, no. 7081 (2006): pp. 174–80; P. Laslo et al., “Multilineage Transcriptional Priming and Determination of Alternate Hematopoietic Cell Fates,” Cell, vol. 126, no. 4 (2006): pp. 755–66; P. B. Gupta et al., “Stochastic State Transitions Give Rise to Phenotypic Equilibrium in Populations of Cancer Cells,” Cell, vol. 146, no. 4 (2011): pp. 633–44. In other words, heterogeneity is pre-existing, and, as in Darwin’s theory, cells are selected to amplify existing heterogeneities, leading to apparently consistent and well-choreographed processes in the end. Interestingly, this is observed both in animals44See, for example, A. C. Martin et al., “Pulsed Contractions of an Actin-Myosin Network Drive Apical Constriction,” Nature, vol. 457, no. 7228 (2009): pp. 495–9. and in plants,55See, for example, M. Uyttewaal et al., “Mechanical Stress Acts Via Katanin to Amplify Differences in Growth Rate Between Adjacent Cells in Arabidopsis,” Cell, vol. 149, no. 2 (2012): pp. 439–51. further supporting the generality of this finding.

It then appears that biology simply never fuels homogeneity, despite the apparent reproducibility of shapes in nature. In fact, even genetically identical organisms in stable environments vary: a degree of randomness in gene expression is sufficient to generate some level of diversity. This sometimes refers to “incomplete penetrance”: individuals with the exact same mutation may have a different appearance or behavior because of the presence of natural fluctuations in biological processes, such as protein synthesis. Conversely, these fluctuations are used in biological processes to generate variable shapes and behaviors, which also means that modern biology increasingly embeds an element of probability.66See, for example, H. H. Chang et al., “Transcriptome-wide noise controls lineage choice in mammalian progenitor cells,” Nature, vol. 453, no. 7194 (2008): pp. 544–7; D. A. Hume, “Probability in Transcriptional Regulation and its Implications for Leukocyte Differentiation and Inducible Gene Expression,” Blood, vol. 96, no. 7 (2000): pp. 2323–8. This can have very practical consequences, for instance in biomedical science: the presence of noise in gene circuits (e.g. fluctuations in protein synthesis) can enable subpopulations of cells to become transiently resistant to antibiotics, thus allowing their survival.77See, for example, E. Rotem et al., “Regulation of Phenotypic Variability by a Threshold-based Mechanism Underlies Bacterial Persistence,” Proceedings of the National Academy of Sciences of the United States of America (PNAS), vol. 107, no. 28 (2010): pp. 12541–6.

To be complete, one must underline that what may appear as “white noise” in biology—i.e. fully random fluctuations—is sometimes in fact “blue noise,” as a pattern can be recognized in certain circumstances, or as techniques become more resolutive (typically seemingly random fluctuations may sometimes reflect cryptic circadian rhythms). There are, however, many situations where noise really is white, questioning the very origin of such fluctuations. The answer to that question is in fact surprisingly simple: to generate white noise at the microscale, one needs to have a low number of molecules. To take an example: if a cell contains a thousand copies of a protein that acidifies the cell content, small fluctuations in the number of proteins would have little effect on the cell acidity because the large number of proteins buffers such fluctuations; white noise does not appear. If instead, a cell contains only two or three copies of that protein, any changes in protein content would generate huge fluctuations in acidity. With the development of single molecule tracking, such fluctuations can now be investigated in the lab, revealing that many proteins are indeed in low copy numbers, and thus in principle are amenable to generate such noise.88See, for example, W. J. Blake et al., “Noise in Eukaryotic Gene Expression,” Nature, vol. 422, no. 6932 (2003): pp. 633–7; J. M. Raser and E. K. O’Shea, “Control of Stochasticity in Eukaryotic Gene Expression,” Science, vol. 304, no. 5678 (2004): pp. 1811–4; J. R. S. Newman et al., “Single-cell Proteomic Analysis of S. cerevisiae Reveals the Architecture of Biological Noise,” Nature, vol. 441, no. 7095 (2006): pp. 840–6.

This again may appear as very inefficient and clearly not optimized (would you accept having only one or two buses in your city?), but the fluctuations this system generates are key to the adaptability of organisms to variations in their environment. In other words, suboptimality at the microscale can generate resilience at the macroscale.

To conclude, these examples illustrate how biology is fundamentally suboptimal, at all scales. As the technosphere instead takes the route of increased optimization, one may question whether optimization is in essence carrying its own end or instead whether our optimized societies will generate other ways to become resilient. More generally, suboptimality goes beyond biology and can easily be explored in our daily lives. Assuming that a system is suboptimal when it retains some element of randomness or apparent inefficiency, then we may experience such suboptimality or absence of suboptimality around us. By stressing it, we may open a window on the tensions between the technosphere and the biosphere, and provide some element of reflection on how the efficiency we (humans) project onto natural processes in turn feeds back on our own societies and behaviors. This may also underline how much we can learn from the biosphere and, in that sense, offer some opting-out options within the contemporary technosphere.