I've been alluding recently to this concept about what we might start seeing in the 2020s: "cognitive autonomous agents", which would act as something akin to the neurons of a cognitive web..
The concept first came to me several years back (maybe in 2014) when I thought about a theoretical future search function that did something very, almost laughably simple: autocorrected to the most likely thing I'm looking for if I'm searching a specific URL.
For example, let's say I'm searching for FutureTimeline forums.
But what happens if I misspell it?
Today, you'd just get an error message.
A functioning cognitive web would "understand" that I'm actually going to the first one and correct it for me. This is simple, yes, but it would work even for much longer strings. As long as you're roughly close, a cognitive agent would figure out what you're searching for.
From there, I realized that a more semantic-based web that had agents that understood natural language and had the ability to follow search trees with some modicum of commonsense would basically make the internet seem intelligent. You could hold a conversation with a digital assistant, telling it over multiple posts about your preference in pizza. Then, the next day, you say, "I'm in the mood for pizza. Order me one from [insert pizza shop here]." It'll ask for certain specifics that it recalls from your conversation, and soon you'll be heating a hot pizza.
Except that isn't something too outrageous compared to what digital assistants today can do, so let's get crazier.
Let's say you download a program, but in order to run this program, you have to rename & move around multiple files into the correct folders. If even one file is out of place, the program will either run wacko or not run at all, so you typically have to read the Readme in incredible detail to make sure everything's in its right place— and sometimes, you might need updates and dependencies anyway.
So why not get a cognitive agent to do all that for you? Feed it the Readme.txt and it'll figure out what to do, what goes where, and if anything is still wrong, it'll search online to find the right dependencies, drivers, and updates for you.
One might argue it's lazy, but I don't see it as any lazier than the move from the extremely technical operating systems of the '70s and early '80s to the simplified graphical operating systems of the late '80s and '90s onwards. Yeah, remember when you needed to find and input the exact line of code just to run a single program and then had to continually input more code from sheets and booklets to keep the program going? And if you missed even a semicolon, that was it, you've failed. Now you simply click on a file and it runs.
Some programs have native features similar to this, but a cognitive agent would be able to do this for any program. For whatever it can't find, it'll go online to find another agent to assist it. And if there's a complete dead end, it ought to be able to tell you (in natural language) how screwed you are.
Artificial intelligence is no longer "just" an engine for software to function but genuinely cognitive agents that possess some level of real-world understanding. It becomes a "cognitive engine", actively capable of responding to us. It can identify problems to fix and then fix them for us. These agents are also capable of communicating with each other— if one agent can't fix a problem, it can search for another that can. The internet itself will seem intelligent. And as a result, the young have no concept of a Web that isn't their friend, a sort of godlike person they know rather than a medium of communication and entertainment.
You'd be able to talk to a chatbot that can also engage in certain actions. This chatbot can then engage with other bots that can engage in certain actions. It would appear to be something like a piecemeal AGI.