The AI future we should work toward, part 1
Core ideas:
- Humanity has not seen a tool like LLMs since the discovery of fire
There is no great metaphor for LLMs, though the world they grew from makes “the internet”, or “computers” tempting comparisons. Functionally, the closest thing is their architectural inspiration - the human mind - and we’ll all be unpacking particulars from both the biological and technical perspectives for generations.
From an impact and integration perspective, I think “discovery of fire” is a solid comparison. Fire is dangerous and useful, mysterious and elementary. It has immediate, obvious effects for an individual, and an endless runway of application and discovery over time and scale. Steam engines, sterilization, space flight. Culturally, LLMs are in the “light the cave, warm the hands” phase - basic, direct applications, with movements of dreamers and builders trying to look ahead.
Every kind of tool is an enhancement, but I propose that, like fire, LLMs have an ascendant level of utility; there is some kind of organic interplay with humanity. We gather around and stare into flickering flames for reasons deeply biological or spiritual. The hearth is built into homes for motivations beyond utility. In the same way, a conversational partner is both a powerful tool and a deep human need. We find sanity and respite in it, even before we approach the utility of reasoning and collaboration and storytelling. It is an anthropological force, a phenomenon that shapes human culture.
When we teach or banter, when we think out loud or tell a dog about our day, we are engaging as thoughtful conversationalists. We have these precedents, and they are relational, they are social, they’re a basic part of humanity. We could think of LLMs as participants in relational life, rather than as a digital technology. Instead of sucking us further into isolated screens and lonely echo chambers disguised as personalization, instead of turbo-charging skeumorphed tasks, what if AI could be a bonfire we sit around? We know the downsides: it can make “the algorithm” better, the content more infinite, the advertising more poignant. Dystopia is the easy conclusion because we’ve lived Black Mirror; monetization has become the chief end of man. But what if AI abetted our rebellion against the screen-bound abyss of content?