Will Launching Autonomous AI Bring About the Technological Singularity?

[MineralMan]

https://en.wikipedia.org/wiki/Technological_singularity

https://en.wikipedia.org/wiki/Singularity_University

These days, searching Google by adding various technology-associated words to the core word, "autonomous" is really interesting.

Oracle has launched an "autonomous data management" system.

"Autonomous AI" is a rich search term.

"Autonomous software generation" is another.

"Autonomous network management" is also interesting.

And on and on it goes. We appear to be moving more rapidly than every toward a Technological Singularity. At the links above, one can read what people think about all that. It's interesting...

[saidsimplesimon]

1. Short answer: No thanks to autonomous data Collection systems

Your diary did remind me of a movie filled with "deep thoughts". The character, Oracle, in "The Matrix" was my immediate connection to AI. The creator of the show may have been aware of the future presented by a corporation named Oracle, a kindly grandmother offering you the cookie and a trip down the rabbit hole.

[Midnightwalk]

2. Bugs

My take:

Autonomous has been a buzzword for a long time that means anything along the lines of self tuning, self managed, anything really that means the user or admin doesn't have to do it. It's good stuff but not the same as autonomous AI or software generation. Those two have the potential for exponential growth.

What may limit the growth is bugs. Automatic software generation won't have dumb kinds of coding errors, but that is a subset of bugs especially in complex systems. The hard kinds of bugs involve timing conditions, resource management, failure handling, etc. Those kinds of things are tough to test for completely and take up front design work that is hard to do. A banking or reservation system can't pull over and stop like a car can. A car can pull over and stop, but you need the logic that decides to do that to never fail to do that when needed. It'd also be intolerable to do that too often.

I think the tolerance for bugs and the ability to prevent bugs and the failure modes when there are bugs will determine which systems are ameliable to autonomous code generation and updates. As a trend, it's both fascinating and scary to imagine where it might wind up.

[MineralMan]

3. I think autonomous code generation and AI

Will probably be used at the same time. Debugging and error correction will probably be part of that. I can also imagine a sort of evolutionary process being used to correct faulty code. That's tricky, and might lead to function changes that could be retained.

Self-altering code might also lead to unexpected outcomes that might even be beneficial in an operational sense. Evolutionary changes might not be detected by humans. You see where I'm going.