The case for substantially increasing adult intelligence with known biotechnologies remains unclear, writes Zvi Benson-Tilsen
The case for substantially increasing adult intelligence with known biotechnologies remains unclear, writes Zvi Benson-Tilsen
On November 18, LessWrong published an essay in dialogue form about using biotechnology to increase adult intelligence. The author works in reprogenetics (technologies that allow parents to influence their future child's genome) and proposes developing a roadmap before making large investments.
“Unfortunately, there is currently no clear path,” the author says: there have been very few deliberate attempts to increase intelligence in healthy adults. By a roadmap, he means bringing together knowledge about the brain, genes, neural interfaces, and drugs, and identifying the engineering questions that follow from the goal itself.
The author identifies four approaches for large projects: gene editing in the brain, brain implants, drug selection, and transplantation of cells that develop into neurons. Gene editing comes first on the list, with hundreds of edits to genes associated with intelligence. It ranks first because it draws on the only concrete data on factors affecting intelligence that might also act in adults: biobanks that link genotypes to IQ or educational attainment. Delivery is an obstacle: editing tools would need to reach many neurons in the cortex repeatedly, while the blood-brain barrier allows only certain substances to pass through.
Many applications require delivery to the brain, and gene therapy developers are already working on it. However, gene therapy needs only one round of editing. The author suspects that its viral carriers (adeno-associated viruses) are too small to accommodate hundreds of edits and would trigger an immune response with repeated administration. In his view, the delivery approach worth accelerating is repeated delivery, which nobody is working on. Examining the goal itself helps identify that need.
The deeper problem, in the author's view, is the lack of a technological principle: an idea for putting a natural phenomenon to use. In reprogenetics, the principle is clear: make small changes to an embryo's genome, and normal brain development will build the resulting abilities. In adults, that development has already taken place, and drawing on it would require additional engineering. Other phenomena remain available, such as neurons adapting to cognitive tasks or long-range connections between cortical regions. These are only vaguely understood, however, leaving it unclear which genes to edit and in how many neurons.
The author also points to a difficulty specific to biology: unlike mathematics and programming, it is a field where hard-won progress can easily be lost. Edited stem cells in a Petri dish can die or accumulate cancer-causing mutations. He has sketched out a billion-dollar project for gene editing in the brain and expects it would most likely fail. On spending ten billion dollars on the same approach, he writes: “I doubt that ten billion would let us move three times as fast as one billion, let alone ten times as fast, though I am not certain.”
From these two difficulties, the author concludes that committing to a single approach is currently a poor bet. Spreading funding across dozens of approaches dilutes the effort: there is too little capacity to examine each one thoroughly. The likely result is a simple strategy such as “fund everything in the field that already looks clearly worthwhile,” yet most of that work is already funded.
The author sees grounds for hope in the range of possible interventions and their potential to complement one another. In an imagined scenario, gene editing makes thinking faster and clearer, while implants that add long-range connections make it more creative. Such combinations offer little guidance on where to invest: “We still need to develop roadmaps.” For now, he says, useful steps include developing expertise in each approach and creating ability tests whose results are unaffected by practice.