Musings on designing experiences & (re)engineering complexity

Jul 2024

AI As Accessibility Tooling

Do you need AI, or just a better memory?

In some past engagements, one of the superpowers exercised has been that of relational-associative memory. Specifically, the power to remember the threads between decisions, even having some assets to remind others of those decisions. When one looks at the landscape for some of the uses of artificial intelligence, we see it being leveraged as a replacement for memory. That gets us down the line of asking if we are replacing the wrong aspects of ourselves with these tools, instead of using these tools to strengthen that muscle?

For some people, their schooling years were filled with the reinforcement of memorization. You read a story and you had to remember specific quotes. Or, you took a test and were asked about obscure aspects of a story or piece of history that really required one to develop some type of framework for memory. For some, they held onto those frameworks, developing into the types of leaders or managers who are known for their artifacts and memory palaces of necessary items. They may have specific behaviors they utilize in order to trigger contextual memory; and at the same time may have a series of knots which prevent them from untethering memories from specific behaviors or places.

For others, even though those school years were filled with memorization, they never quite jumped the chasm of getting that framework. Instead, they relied on other senses in order to get through the doldrums of academic work. When academic work left, there was no longer need for memorization as a muscle. Those facilities which were already underdeveloped (???), were left aside for other facilities. For these, the rise of general purpose transformers (GPT) make up for the parts of memory replaced with other psychological or social aspects. So it seems that this is a type of intelligence, when really it’s more of an accessibility aid.

If language models, and the prompting/spells which generate outputs are looked at as an accessibility tool for memory or types of thought, how could we take a perspective of its biases, its wrinkles, its properties, and then shape the tool before the tool completely reshapes us? If folks need an aide for memory, then what are we rightly or not rightly giving people before they use the tool? And if AI is a method, more than it is a tool, who and what shapes the containers that memory needs to exist within? Whose narrative is worth remembering?