A quick trip in the DeLorean, and 10,000 years ago, your ancestors were conversing with imaginary beings. Spirits who would help find you a partner, or fix medical problems.
Sound familiar? Fast forward to today, and we've upgraded those spirits to algorithms. History doesn't repeat, but it sure does rhyme.
And the future is starting to look an awfully lot like the past.
Jump-cut to now, and if you are anything like me, your day-to-day is starting to look like a series of Pokémon battles. Except rather than types, it's problems.
Bug in the code, I choose you 'Debugging Script Prompt!'. A couple of clicks later, and presto, problem solved. Confused by quantum mechanics? "I summon thee, Richard Feynman Technique Prompt!" and bam, you're a physicist.
What's my point? I guess that humans are ludicrously adaptable. It's only been a year since GPT became accessible, and already it feels like I'm running a daycare for pint-sized AI toddlers.
Except these toddlers can write Python.
AI today isn't a full mind, but more specific thoughts that we weave together into a solution. We are the conductors of the orchestra, bubbling up what works, and popping what doesn't.
Which makes me think, what's the next act of the opera? The obvious answer would be a fully autonomous agent with a narrow scope, that replaces us in the middle. A system that is not able to "have" more than a single thought, string them together, review it's progress, and act on the result. Mabybe it's models that can see, and hear, and move? Perhaps learning by watching us do it.
But that seems far away until we build something like Archimedes or BabyAGI. I would put money first on people first building tools that make it easier to manage a thousand different thoughts. Surely, there's gotta be a better solution than listing them one-by-one down a sidebar of chats.
Or maybe Maggie is right.
And the best AIs won't look like chat apps at all.
Oh, and by the way - if you'd like to see all the prompts I use, add your email address below and I'll send you an invite.