biological anchors
In plain wordsA way of forecasting when AI could transform the world, by estimating how much computing it would take to match the human brain and when that much computing would be affordable.
A forecasting framework that estimates when transformative AItransformative AI, n.: AI that changes the world as deeply as farming or the Industrial Revolution did. It is defined by its effects on society and the economy, not by how humanlike or clever it is. becomes feasible by anchoring the training computetraining compute, n.: The total number of arithmetic operations performed to train a model. It is the standard yardstick for the size of an AI training effort, and some laws use it to decide which models face extra rules. it would require to quantities from biology, then projecting when that compute will be affordable given trends in hardware price-performance, algorithmic efficiency and spending. The main anchors start from the brain's estimated computation (on the order of 10^15 operations per second), size a model accordingly and ask how much training it would need, with variants depending on how long each training episode must be (short, medium or long horizon); other anchors use the computation of a human lifetime of learning, the information in the genome, or all the computation performed by evolution in producing brains. A weighted mixture of these anchors yields a probability distribution over arrival dates. Supporters value it as an explicit, checkable model; critics, including Eliezer Yudkowsky, argue that biology says little about the compute needed by algorithms very different from the brain's.
Bio anchors put most of its weight on the neural-network anchors, not on recapitulating evolution.
Origin
Developed by Ajeya Cotra at Open Philanthropy in the draft report 'Forecasting TAI with biological anchors' (2020), which gave a median of around 2050 for transformative AI; in 2022 she moved her personal median to about 2040. Precursors include Hans Moravec's 1998 extrapolation of when hardware would match the brain.
Related
- AI timelinesAI timelines, n.: Forecasts of when AI will reach major milestones, such as matching humans at most work. People with short timelines expect this within years; others expect decades, or never.
- transformative AItransformative AI, n.: AI that changes the world as deeply as farming or the Industrial Revolution did. It is defined by its effects on society and the economy, not by how humanlike or clever it is.
- training computetraining compute, n.: The total number of arithmetic operations performed to train a model. It is the standard yardstick for the size of an AI training effort, and some laws use it to decide which models face extra rules.
- scaling lawsscaling laws, n.: Measured patterns showing that an AI model's error shrinks smoothly and predictably as the model is made bigger and given more data and computing power, so labs can forecast a giant model's performance from small, cheap experiments.
- effective computeeffective compute, n.: A way of measuring AI progress that counts better methods as if they were extra computing power: if a new technique lets a model match one trained with twice the computing, it counts as doubling effective compute.
- Hans MoravecHans Moravec, prop. n.: An Austrian-born roboticist who observed that skills humans find effortless, like seeing and moving, are the hardest for machines, and who forecast when computers would match the brain.
Sources
- Cotra, A. (2020). Forecasting TAI with biological anchors (draft report). Open Philanthropy.
- Moravec, H. (1998). When will computer hardware match the human brain? Journal of Evolution and Technology 1.
Mentioned in
AI timelinesAI timelines, n.: Forecasts of when AI will reach major milestones, such as matching humans at most work. People with short timelines expect this within years; others expect decades, or never., transformative AItransformative AI, n.: AI that changes the world as deeply as farming or the Industrial Revolution did. It is defined by its effects on society and the economy, not by how humanlike or clever it is., Ajeya CotraAjeya Cotra, prop. n.: An American AI researcher and forecaster whose 'biological anchors' report estimated when transformative AI might arrive, and who argued that standard training could produce deceptive AI., Hans MoravecHans Moravec, prop. n.: An Austrian-born roboticist who observed that skills humans find effortless, like seeing and moving, are the hardest for machines, and who forecast when computers would match the brain., Open PhilanthropyOpen Philanthropy, prop. n.: A large American charitable foundation, renamed Coefficient Giving in 2025, that has been the biggest private funder of research on the risks of advanced AI.