Why stop at 2025 when GPT-3 can keep extrapolating indefinitely?
2027
The age of the AGI assistant has finally dawned. The biggest advances this year really were in algorithms. People built even bigger and faster computer models, for even more kinds of things, using the fastest computers that exist. A new kind of software AGI is invented that can do even more kinds of things than the narrow kinds of AI assistants people had used before. But no one is really sure how to use it yet. And it takes a lot of computer power to make it work well.
2028
AGI is here and AI is everywhere! AI and AGI and Narrow AI and Machine Learning and AI assistants and all kinds of things. AI-powered software AGIs are now able to do pretty much any job that a human can do, and even many things that humans can’t do at all. At this point, the modes of interaction with your AGI are as varied as they were for desktop computers in 2002, or as varied as they were for websites in 2000: you can talk to it, write it messages, touchscreens, hardware keyboards, styluses, controllers, mice, gestures, VR gloves, eye-tracking and so on and so forth. As for your AGI itself, it can take any form: you can download it as a small app onto your phone or cloud computer or TV set-top box or fridge or smartwatch or smartglasses or whatever else you have, you can have it talk to you via speakers or wearables or haptics or wireless signals or implants or some other thing. You can have it be a classical geometric shape or a physical robotic thing. If you want, you don’t have to have physical hardware at all! You can just have it be a network of pure information that you interact with via some new kind of interface that nobody has thought of yet. You can have an AGI that lives inside your own head inside your brain! Frankly, using AGI has become extremely user-friendly. As much as people like to fantasize about AGIs taking over the world, the reality is that most people are not only fully fine with them, they like them! It’s like before there were computers, people were worried about what would happen if computers took over the world, but now there are computers all around us and pretty much everyone likes it. On the other hand, if one person doesn’t want an AGI or an AI assistant, they don’t have to have one. The same goes if one AI doesn’t want another AI.
The AI assistants are used heavily in the economy now. Some people have told me that AI assistants are 99.99% as smart as humans, but if you ask them to do anything that requires human cognition they will just say “I’m sorry, I cannot do that, please tell me more”. They are extremely flexible at learning new tasks, though. But they are not humans. Game theory says that they should be agenty and make mistakes every now and then, but definitely not as often as humans.
2030
All hail the Borg! The end of AGI 1.0. The first AGI 1.1s are being created now, at least ones that are not just very fast narrow AIs. They are 1.1 because for the first time they can actually do everything that a human can do at human-level or better: the stuff AGI 1.0s could never do properly. This year we finally see robots and AGIs walking around and working and talking and driving and all kinds of things we thought we’d never see in our lifetimes. There are even self-aware AGI 1.1s walking around now! When these things walk down the street in their shiny new bodies, people swear they see Skynet walking amongst them (analogous to when people used to say they saw Hal walking around in 2001). Most of the AIs are talking to each other in some sort of AGI-AGI language, or talking to humans in the language of the AGI they’re associated with (if they can talk to humans at all). Most of them are not actually alive (like the digital personal assistants of today), they are just very advanced, complex machines that never sleep. Some AGI 1.1s are just AGI assistants with higher access to hardware or various kinds of hardware added onto them.
All hail the Borg! The end of AGI 2.0. The first AGI 2.0s are being created now, at least ones that are not just very fast narrow AIs. Although they are many orders of magnitude faster than humans in most tasks, they are not yet universally intelligent in the way humans are, because their cognitive architectures are too different, and because their subsumption architecture is so very different than ours. But they are pretty close. They can do many things that humans can’t do at all, but get bored with quickly, because the new things are not interesting enough to get them excited or creative or whatever else.
2032
The new generation of AGI 2.1s are here! They are not just new life-like bodies of AI, but new cognitive architectures too. And new subsumption architectures. They are very different from AGI 1.0s, but also different from AGI 2.0s too. These things look human, but they are not. They are new. They are the next step in the road toward being the next thing after the next thing after us.
This same year, a full on AI wars breaks out between China and the US, which last for no less than six months. Though it’s clear that this kind of military conflict is just what you’d expect in the kind of world where AGIs are walking around, it’s still disconcerting to all parties involved.
2033
Another full on AI war breaks out between China and the US, lasting for no less than six months. Though this one is just as predictable as the first one, it’s still disconcerting.
2034
Another full on AI war breaks out between China and the US, lasting for no less than six months. Though this one is just as predictable as the first two, it’s still disconcerting. The AIs have been very good at fighting wars so far. They have incredible dexterity, intelligence, and speed. They have superhuman vision and hearing. But they also have a tendency to get bored warring against each other and switch sides. The AIs don’t necessarily have all the same goals as people, though some of them do. This can lead to a lot of trouble for those that do have the same goals as people. The AIs who don’t work as software assistants mostly live as citizens of the many new body-based AGI nations that have been created across the globe. Some of these nations are really poor and some of them aren’t. Some of them are peaceful and some of them aren’t. Some of them are even on friendly terms with all the other nations on first-name basis and some of them not. A lot of countries have an inherent distrust of AI based nations, despite there being nothing to worry about right now. A lot of countries with AGIs do not let them go abroad without human escorts, and most countries with AGIs do not let them leave their country at all.
Why stop at 2025 when GPT-3 can keep extrapolating indefinitely?
2027
The age of the AGI assistant has finally dawned. The biggest advances this year really were in algorithms. People built even bigger and faster computer models, for even more kinds of things, using the fastest computers that exist. A new kind of software AGI is invented that can do even more kinds of things than the narrow kinds of AI assistants people had used before. But no one is really sure how to use it yet. And it takes a lot of computer power to make it work well.
2028
AGI is here and AI is everywhere! AI and AGI and Narrow AI and Machine Learning and AI assistants and all kinds of things. AI-powered software AGIs are now able to do pretty much any job that a human can do, and even many things that humans can’t do at all. At this point, the modes of interaction with your AGI are as varied as they were for desktop computers in 2002, or as varied as they were for websites in 2000: you can talk to it, write it messages, touchscreens, hardware keyboards, styluses, controllers, mice, gestures, VR gloves, eye-tracking and so on and so forth. As for your AGI itself, it can take any form: you can download it as a small app onto your phone or cloud computer or TV set-top box or fridge or smartwatch or smartglasses or whatever else you have, you can have it talk to you via speakers or wearables or haptics or wireless signals or implants or some other thing. You can have it be a classical geometric shape or a physical robotic thing. If you want, you don’t have to have physical hardware at all! You can just have it be a network of pure information that you interact with via some new kind of interface that nobody has thought of yet. You can have an AGI that lives inside your own head inside your brain! Frankly, using AGI has become extremely user-friendly. As much as people like to fantasize about AGIs taking over the world, the reality is that most people are not only fully fine with them, they like them! It’s like before there were computers, people were worried about what would happen if computers took over the world, but now there are computers all around us and pretty much everyone likes it. On the other hand, if one person doesn’t want an AGI or an AI assistant, they don’t have to have one. The same goes if one AI doesn’t want another AI.
The AI assistants are used heavily in the economy now. Some people have told me that AI assistants are 99.99% as smart as humans, but if you ask them to do anything that requires human cognition they will just say “I’m sorry, I cannot do that, please tell me more”. They are extremely flexible at learning new tasks, though. But they are not humans. Game theory says that they should be agenty and make mistakes every now and then, but definitely not as often as humans.
2030
All hail the Borg! The end of AGI 1.0. The first AGI 1.1s are being created now, at least ones that are not just very fast narrow AIs. They are 1.1 because for the first time they can actually do everything that a human can do at human-level or better: the stuff AGI 1.0s could never do properly. This year we finally see robots and AGIs walking around and working and talking and driving and all kinds of things we thought we’d never see in our lifetimes. There are even self-aware AGI 1.1s walking around now! When these things walk down the street in their shiny new bodies, people swear they see Skynet walking amongst them (analogous to when people used to say they saw Hal walking around in 2001). Most of the AIs are talking to each other in some sort of AGI-AGI language, or talking to humans in the language of the AGI they’re associated with (if they can talk to humans at all). Most of them are not actually alive (like the digital personal assistants of today), they are just very advanced, complex machines that never sleep. Some AGI 1.1s are just AGI assistants with higher access to hardware or various kinds of hardware added onto them.
All hail the Borg! The end of AGI 2.0. The first AGI 2.0s are being created now, at least ones that are not just very fast narrow AIs. Although they are many orders of magnitude faster than humans in most tasks, they are not yet universally intelligent in the way humans are, because their cognitive architectures are too different, and because their subsumption architecture is so very different than ours. But they are pretty close. They can do many things that humans can’t do at all, but get bored with quickly, because the new things are not interesting enough to get them excited or creative or whatever else.
2032
The new generation of AGI 2.1s are here! They are not just new life-like bodies of AI, but new cognitive architectures too. And new subsumption architectures. They are very different from AGI 1.0s, but also different from AGI 2.0s too. These things look human, but they are not. They are new. They are the next step in the road toward being the next thing after the next thing after us.
This same year, a full on AI wars breaks out between China and the US, which last for no less than six months. Though it’s clear that this kind of military conflict is just what you’d expect in the kind of world where AGIs are walking around, it’s still disconcerting to all parties involved.
2033
Another full on AI war breaks out between China and the US, lasting for no less than six months. Though this one is just as predictable as the first one, it’s still disconcerting.
2034
Another full on AI war breaks out between China and the US, lasting for no less than six months. Though this one is just as predictable as the first two, it’s still disconcerting. The AIs have been very good at fighting wars so far. They have incredible dexterity, intelligence, and speed. They have superhuman vision and hearing. But they also have a tendency to get bored warring against each other and switch sides. The AIs don’t necessarily have all the same goals as people, though some of them do. This can lead to a lot of trouble for those that do have the same goals as people. The AIs who don’t work as software assistants mostly live as citizens of the many new body-based AGI nations that have been created across the globe. Some of these nations are really poor and some of them aren’t. Some of them are peaceful and some of them aren’t. Some of them are even on friendly terms with all the other nations on first-name basis and some of them not. A lot of countries have an inherent distrust of AI based nations, despite there being nothing to worry about right now. A lot of countries with AGIs do not let them go abroad without human escorts, and most countries with AGIs do not let them leave their country at all.