On Neuralink

This post is a reflection on Neuralink and the Brain’s Magical Future, by Tim Urban (2017) that attempts to answer the following question: In what fundamental way could a neural implant redefine what future humans will be? Is there some line beyond which we would no longer consider a person to be human? If so, how could/would/should we specify that line?

To answer the posed question, I want to start with a clear definition of a human. A quick Google search shows sentences such as “Humans (Homo sapiens) are highly intelligent primates that have become the dominant species on Earth”. These do not satisfy my question of what a human is.

Allow me to offer my view on humans that is based on the definitions that high intelligence is what separates us from other species.

Continual learning, adaptation, and reasoning about the consequences of our actions are three components that distinguish a human. A process when the agent has to change what it knows, rather than knowing whether something is a fact, is defined as learning by Selfridge (1993) and is considered the most important part of intelligence (Woodrow 1946). This learning leads to adaptation. In my recent work (Kudashkina et al. 2020), I talk about intelligent assistants and their ability to evaluate the consequences of actions. I believe this ability is what makes them closer to humans. To turn it around, my belief is that this ability is what distinguishes us as humans.

How do humans demonstrate continual learning, adaptation, and reasoning? I suggest that writing is one of the forms of activity where these qualities are demonstrated. Writing is what allows as to take our thoughts, work with them, structure them, and then massage them to the point of them being crips and clear to communicate further. Learning, adaptation, and reasoning happen during this writing process.

Let’s go back to Neuralink now and see if the proposed future would take away these abilities or enhance them. Let’s make it clear for a reader that we are not engaging in a discussion whether the BMI’s future is possible, is a good one, is well or poorly defined, is overlooking important potential issues, is raising right governance, control, and society questions, is giving us more choices or takes them away … This list could go on. Instead, here we simply take an assumption that given what’s described by Tim Urban is possible and will happen, what would that mean for being a human in the limited definition of a human offered above.

Humanity could go two ways in the Neuralink scenario. If in human-AI integration we maintain our ability to work with our thoughts before spitting out some garbage unstructured thinking into someone else’s brain or another machine, then we will preserve our ability to think about the consequences of our actions, evaluation solutions, structure out thinking.

In contrast, if human-AI integration no longer allows our brains to function on their own and the thoughts are triggered only artificially by whatever is implanted in our brains, then I believe we lost. We lost humanity to the progress.

References:

K. Kudashkina, P. M. Pilarski, Richard S. Sutton: Document-editing Assistants and Model-based Reinforcement Learning as a Path to Conversational AI. Under Review. ArXiv: 2008.12095.

Selfridge, O. G. 1993. The Gardens of Learning: A Vision for AI. AI Magazine 14(2):36–48.

Woodrow, H. 1946. The ability to learn. Psychological Review 53(3):147–158.