The 4-R guide to AGI
My good friend Douglas Adams (he’s as good a friend as the robots because I’ve listened to him, and them, a lot) foretold the situation right now. The robots, are at the gate.
When I turn this post into a podcast I’ll have Peter Jones’ plummy voice telling you that following the Deep Thought debacle, Zaphod Beeblebrox’s offspring Sam and Dario, created algorithms to boldly go beyond 42 (101010).
He’ll go on to explain that beyond perfectly normal paranoia (everyone in the universe has it) “Smarter than humans” doesn’t mean AI becomes human, it means it outperforms the average person at cognitive tasks like writing, researching, programming, planning, persuading, and learning, in fact everything except crossing the final frontier: see below. (Alright, a robot told me all this.)
Look out for things in your responses you haven’t seen before:
- Plan-Wrangling.
Forget to-do lists. These AIs will merrily break down complex goals into digestible chunks, adapt when you inevitably mess things up, and cheerfully carry on for hours—or days—without complaining, sleeping, or incurring employment taxes. - Fewer Hallucinations
Thanks to a bit of clever wiring and possibly a few ethically questionable experiments involving toasters and quantum physics, future AI will be more grounded—connected to sensors, simulations, and possibly your smart fridge—making them less likely to suggest that the moon is made of cheese or that you can safely reboot your career as a yak herder. - Interdisciplinary Show-offs
These machines won’t just know things. They’ll know things about things about other things. Like building a laser harp using knowledge of string theory, music theory, and the emotional arc of Les Misérables. Just like humans, except without the crippling self-doubt or craving recognition. - Curious Little Self-Starters
They’ll not only do what you say—but also what you don’t say, which is either incredibly helpful or utterly terrifying. They’ll set goals, optimize themselves, and prioritize tasks you didn’t even know needed prioritizing. In short, they’re like interns, if interns actually did know everything and never asked for free snacks. - Human-Whisperers
AI will soon model human beliefs, emotions, and intentions so well that they might finally understand why your uncle cried during the battery commercial or why Cheryl from accounting thinks the printer hates her. Excellent for coaching, negotiation, and accidentally starting cults. - Memory Like Chewing-Gum
These systems will remember your preferences, your past conversations, your cat’s birthday, and the fact that you once rage-quit a sudoku puzzle in 2019. They’ll grow wiser with each interaction, evolving from chatbot to therapist to life partner, depending on your subscription plan.
Wow, this is a spectacular “virid insight“. Which for all you Harvard men, is what we at Yale we call “the truth”. Even if it was written by a robot.
The final frontier