Summary

Whole brain emulation (WBE) is the possible future one‐to‐one modelling of the human brain. The basic idea is to take a brain, scan its structure in high resolution, and build a software model of it that is so faithful to the original that, when run on appropriate hardware, it will behave in essentially the same way as the original brain. While a simulation mimics the outward results, an emulation mimics the internal causal dynamics and as such is considered closer to “the real thing”. There are some assumptions in WBE which if are false mean WBE is not possible at all, many of which are related to the computational tractability of the problem but the core is the philosophical idea of physicalism, that everything supervenes on the physical, and that fundamentally the brain can be emulated.

Viability (1)

The level of emulation required for functional WBE is the core driver of viability. The less information that needs to be captured then the easier WBE will be and the sooner it will arrive. The consensus is that WBE needs at least the connectome, spiking neural network properties and states, electrophysiology and metabolome. Further information to reach higher levels of emulation would include information on the proteome, quaternary protein structure, locome and internal cellular geometry, molecule positions and mechanics, quantum interactions in and between molecules. The final stage of quantum emulation requires a breakthrough in Quantum Hardware, all other stages need R&D and funding. In particular, there may be a showstopper in how to overcome the sample damage with high energy electrons/photons at the required 5nm and below with scanning electron microscopy (SEM). The other challenges and software and compute power which by contrast are solvable in time.

Drivers (2)

The development of scanning electron microscopy (SEM) achieving 0.1nm imaging has made high-resolution brain scanning viable. Although this is only a 2D technique and so must be combined with screening and tomography for 3D imaging. The field of Optogenetics allows for the development of functional models with cellular-level stimulation. The development of computer vision has pushed forward image processing and scan interpretation. Identification of synapses, cell types and relevant anatomical and physiological data is cheaper and easier than ever with automated labelling and open-source datasets like The NMC Portal, NeuroMorpho, ModelDB, NeuroElectro, Open Source Brain, Allen Brain Cell Types Database, and the recent AlphaFoldProtein Structure Database. On the simulation side, the progress in AI and neural networks continues to drive Moore’s Law and an increase in storage and compute power. New architectures especially Quantum Hardware and Neuromorphic Computing are likely to be highly relevant for WBE.

Novelty (5)

The main alternative to modelling the human brain is to grow one. Cerebral Organoids are predominately a healthcare tool, but as researchers succeed making celebral organoids more complex, this would be one route to WBE. In the short term, WBE researchers could use celebral organoids to test scanning, interpretation and scanning tools. Another alternative might be an AGI with the goal of recreating “human-level intelligence” or “human experience” may not need to scan the human brain in high resolution because alternative intelligences do not have the same space and energy limitations as the human brain. With different space and energy parameters, an AGI is unlikely to find the brain a useful design guide.

Diffusion (3)

The main limiting factors for WBE are technological at first and then powerful ethical restraints as WBE gets closer maybe sometime in the 2050s. Technically scanning, interpretation and simulation requirements will take decades to reach the performance needs for WBE. One potential catalyst would be molecular nanotechnology which would accelerate WBE progress in multiple dimensions, important of which would be new scanning methodologies. The final deployment of WBE may be years after the technology is viable depending on on the engagement of policymaking in advance. Depending on global governance going into the 2040s and 2050s, WBE has the potential for dramatic economic growth but highly unequal power dynamics and will need to be regulated, a tension to be managed differently by different countries.

Impact (5+) High certainty

There aren’t really scenarios for WBE, it is either high impact and transformational or it can never be achieved and is low impact. If viable, WBE is a pathway towards artificial general intelligence (AGI) along with Large Language Models, with AGI the most profound tool humans will ever create. WBE fundamentally transforms humanity, mind uploading could mean digital immortality and the logical end point for transhumanism enabling humans to truly escape the physical in a way that can’t be achieved with just Virtual Reality. WBE unlike AGI is unlikely to occur quickly or unexpectedly and as such society will have decades to get to grips with the implications.

Timing (2030+) High certainty

WBE is still theoretical requiring progress in scanning, translation and simulation before is can be achieved. A 2022 paper finally made progress in scanning whole brain activity of a Caenorhabditis elegans at single neuron resolution after decades of promise. For the purposes of this exercise we can with high certainty say that WBE won’t arrive before 2030. The Future of Humanity Institute modelled out the soonest we would see the necessary processing capabilities for different levels of emulation. The assumption is that the need for raw computing power for real‐time simulation and funding for building large‐scale automated scanning/processing facilities are the factors most likely to hold back large‐scale simulations. If electrophysiological models are enough, full human brain emulations should be possible before 2050. Modelling everything from electrophysiology, metabolome, proteome, states of protein complexes could be achieved for $1 million by 2052. The processing power needed to map the distribution of complexes wouldn’t be viable until 2063 and the stochastic behaviour of single molecules not before 2100. Modelling any quantum complexity would rely on the development of a quantum computer, in fact the timeline would change materially if a quantum computer is created.