Unmaking Sense

We consider the relationship between words, meanings and understanding in the circumstances where a word is new, and not one that we have met or used before. Perhaps even a circumstance when nobody has used it before, as happens when the human race invents a new word to deal with a new circumstance, such as the rise of the Internet. We consider the way we learn by using words even when we don’t understand them, and ask whether we are right to be dismissive of the idea that a chatbot doesn’t understand the words either when we consider all the ramifications of this example.

We look in more detail at the parallels between the way generative decoders produces text and the way human beings speak and write and generally live their lives by putting one foot in front of the other and deciding now what without necessarily having a definite idea of a final destination.

At the end of episode 47 of the series, I made the observation that if we could apply similar technologies to life itself, we might be able to find answers to questions like what next or now what? We explore this idea in a little bit more detail and consider its implications for the question how much of the past we should allow ourselves to take into consideration as we decide on the immediate and longer-term future.

We describe the process of tokenisation, encoding, and decoding in a way that explains the operation of software like GPT-3 and chatGPT from the beginning to the end of the text generation process. More details on the HuggingFace website, to which this episode is indebted.

The only example of self-awareness of which I have direct knowledge is my own. I grant gladly that other human beings and many other species enjoy something of the same kind of self-awareness, even if it is nonlinguistic or non-verbal. Since I interact with other entities - human and non-human - that I have only my assumption of awareness to base that relationship on, what manner of prejudice is it that would lead me to deny similar, or better, or simply different levels of self-awareness to something just because it consisted of a set of mathematical functions buried inside an electronic machine?

We explore the process of generating writing, or speaking in a way that satisfies the condition that we only know that we think we believe it when we hear ourselves saying it. We compare this with the generation of responses by chatbots using predictive text, probabilities and so forth. We ask, if that doesn’t involve understanding, then what does? What does human understanding really amount to other than the ability to produce words and perform actions that make sense to other human beings and to themselves?

We can now potentially download our brains into a chatbot using digital texts we have written. It would not be everyone’s idea of paradise, but that it is possible at all should give us pause. Derek Parfit, societies of competing chatbots with different sets of goods and values that might obviate the need for violent conflict; and how we just entered a new age.

Why the phenomenon of aphantasia and the possibility that some human beings have no inside story might mean that the boundaries between the human and AGI chatGPT worlds of awareness and self-understanding are already less clearly defined than we might like to think.

Just as chatGPT’s neural net is incomprehensible, consisting, as it does, only of a collection of matrices, storing numbers, unless it responds to a human being, so it appears our selves are unintelligible except in what Virginia Woolf calls the apparitions that emanate from us from time to time through which we are known. We include an extended quotation from section 11 of the first chapter of Virginia Woolf’s “To the Lighthouse.”

This episode needs to be taken together with 8.41. We are exploring the idea that a self need not be consistent, or a life be a continuous “horizontal” story, but is rather a sequence of irruptions from the depths that are as unintelligible unless they give voice to themselves as chatGPT’s digital neural net.