A thousand brains: toward biologically constrained AI

Abstract

This paper reviews the state of artificial intelligence (AI) and the quest to create general AI with human-like cognitive capabilities. Although existing AI methods have produced powerful applications that outperform humans in specific bounded domains, these techniques have fundamental limitations that hinder the creation of general intelligent systems. In parallel, over the last few decades, an explosion of experimental techniques in neuroscience has significantly increased our understanding of the human brain. This review argues that improvements in current AI using mathematical or logical techniques are unlikely to lead to general AI. Instead, the AI community should incorporate neuroscience discoveries about the neocortex, the human brain’s center of intelligence. The article explains the limitations of current AI techniques. It then focuses on the biologically constrained Thousand Brains Theory describing the neocortex’s computational principles. Future AI systems can incorporate these principles to overcome the stated limitations of current systems. Finally, the article concludes that AI researchers and neuroscientists should work together on specified topics to achieve biologically constrained AI with human-like capabilities.

Probabilistic design of optimal sequential decision-making algorithms in learning and control

“…These mechanisms, besides giving performance guarantees should also be able to handle cooperative/competitive behaviors among peers [171]. Further, recent studies in neuroscience (see [172] for an introductory review of the theory, which is based on the pioneering work [173]) hint that similar crowdsourcing mechanisms might be implemented by the brain's neocortext to orchestrate how the output of certain cortical circuits are used to build models and actions. In turn, it is believed that this mechanism might be at the basis of our ability to re-use acquired knowledge in order to synthesize new action policies tackling increasingly complex tasks.…”

The AI trilemma: Saving the planet without ruining our jobs

“…More importantly, every time such a bit loses its action potential through a computing operation, energy is being released. In contrast, neurons have been shown to operate with sparse representations, where individual dendrites of a neuron are being activated when a certain (small) percentage of a large set of potential links is active, often less than 5 per cent (Ahmad and Scheinkman, 2016;Hawkins and Ahmad, 2016;Hole and Ahmad, 2021). Not only do operations on sparse representations use much less energy than those https://thebrain.mcgill.ca/flash/capsules/outil_rouge .html, https://theconversation.com/our-ability-to-recognise-letters-couldbe-hard-wired-into-our-brains-Frontiers in Artificial Intelligence frontiersin.org Ernst .…”

“…More importantly, every time such a bit loses its action potential through a computing operation, energy is being released. In contrast, neurons have been shown to operate with sparse representations, where individual dendrites of a neuron are being activated when a certain (small) percentage of a large set of potential links is active, often less than 5 per cent (Ahmad and Scheinkman, 2016 ; Hawkins and Ahmad, 2016 ; Hole and Ahmad, 2021 ). Not only do operations on sparse representations use much less energy than those of dense ones—most operations involve zeros—they are also particularly robust against errors: Calculations by Hawkins and Ahmad ( 2016 ) demonstrate that for typical synapses error rates can reach 50 per cent without neurons losing their capacity to properly identifying underlying patterns.…”

Assessing the ethical and social concerns of artificial intelligence in neuroinformatics research: an empirical test of the European Union Assessment List for Trustworthy AI (ALTAI)

“…Whether ethical concerns linked to technologies that currently do not exist-and that may never come into being-is highly contentious. However, AGI is arguably a key target of AI research [59,77], it is widely covered in science fiction and the media. The development of AGI raises a number of additional concerns such as the possibility of superintelligence which may not only be superior to human intelligence but may be indifferent or hostile to humans.…”