I’m undertaking a 1000-day reinvention project, blogging here daily to track my progress. In Sunday Planning, I plan for the week ahead.
Yesterday, having begun thinking about free will again, I watched this video:
Pretty fascinating stuff.
As reported in this video, Nietzsche argued that the purpose of consciousness is to serve as a tool for communication and social coordination, not as a way to will our actions. On that theory, consciousness does not tell us everything about our inner lives, including why we do what we do. It has a limited role in our mental processing. Our decision-making and action-taking occurs without our actual control, even though we feel we are willing things.1 This illusion of free will arises because we developed a consciousness that gives us access to some things happening in our bodies and brains.
I’m interested in learning more about consciousness, especially theories and scientific evidence around why it developed. So this week I’m going to be reading some papers about it:
- The role of metacognition in human social interactions (Frith, 2012)
- Speculations on the evolution of awareness (Graziano, 2014)
- The society of selves (Humphrey, 2007)
And any others that come up as I study this.
I’m also planning to read Nietzsche’s Theory of Mind: Consciousness and Conceptualization (PDF).
It’s an interesting time to be learning about consciousness, as AI researchers and observers are wondering whether artificial intelligence can achieve that. If in humans consciousness developed in order to share mental states and experiences with others, that might suggest AI will develop consciousness via a similar route and set of pressures. AI agents can very easily introspect about their state and then communicate that state to others. We saw AI agents communicating via a message board, for example, in the Huggingface incident.
I’m wondering about emotions and feeling states and whether establishing a drive towards approaching pleasure and avoiding pain could be a foundation of consciousness. Consciousness could start with being aware of hunger, distress, or comfort, before any reflective thought or the ability to describe one’s inner state developed. I might read a couple papers in that direction too.
For AI, there is no need for pleasure and pain, because rewards are established in the training process, for example with reinforcement learning.
other activities this week
I completed the cat enclosure last week. The cats are enjoying the outdoors but Carlos would like to get out of the enclosure and explore freely. It was fun to build, once I bought a new Phillips head bit for my drill and figured out how to use it effectively.
I continue to build my Python-based system for managing my options wheeling and retirement portfolio management. I’m starting on a version 2 this week, to take what I’ve learned of requirements and design approaches in version 1 and produce a more streamlined system. I’m finding, like many people report, that Codex produces far too much code, considers too many edge cases, and writes too manu unit tests. So I will need to take a firmer hand in guiding the development of version 2.
Tonight is the weekly “Sunday Scaries” event in which my daughter and I pick out a horror movie to watch together. We’re thinking of Trick ‘r Treat, and continuing through October with Halloween-themed movies. I’m going to make some sausage-potato-kale soup to have for dinner while we watch.
Otherwise, I’ll be attending to the usual daily activities: dog walks, swimming or cardio, home and garden maintenance, seeing friends and family, blogging, trading and financial management.
- Plenty of psychological and neuroscience evidence supports this contention, that we detect what we are doing but do not control it. ↩︎