#karlFriston and gang are at it again. They are formalising the notion of intelligence as "as the capacity to accumulate evidence for a generative model of one’s sensed world". In a #reinforcementlearning postulation the reward function is formulated as the need to maximize the evidence for
the agent's own generative model aka it's continued existence
maximize the evidence for
its generative model.

https://arxiv.org/abs/2212.01354

#freeenergyprinciple #artificialintelligence

I wonder what people here on @sigmoid.social think about this?

https://arxiv.org/abs/2212.01354

"... a milestone paper outlining a vision for the future of #AI."

🤔

Source of quote: https://twitter.com/neojipc/status/1599834270570549249?t=alMiRL5uXpSiVuRSZqcQfg&s=19
#FutureOfAI #scalinghypothesis
#freeenergy #freeenergyprinciple

Theories of systems and agents focus on mathematical frameworks providing formal #definitions of #systems and #agents that can take into account different features of #biological, #cognitive and #artificial systems, Some examples include: #freeenergyprinciple, #IIT and recent ideas formalised using #categorytheory.

2/

#Introduction time.

TL;dr: maths to formally study #agents, #worldmodels and #representations

I'm currently a Chief Researcher at Araya, a Tokyo based startup with the goal to understand (artificial) #consciousness.

My work focuses on general principles for the #origins of #agency, #life and #cognition. I previously used the #freeenergyprinciple (+ #activeinference) as my main framework, but I'm now looking into other directions, mostly using #categorytheory applied to #systemstheory.