Artificial Intelligence: Where is it heading?

Stuart Esplin

June 2024

AI has three main types—artificial narrow intelligence (ANI), artificial general intelligence (AGI), and artificial superintelligence (ASI).

Narrow Ai is the only kind of AI available today. When you search for something on Google, AI is working behind the scenes to give you the best result possible. ANI offers us prompts of words and phrases when composing our email. It has auto complete and spell check functions for our word processors.

What of the future?

A 165-page paper, titled Situational Awareness. The Decade Ahead was released this June 2024 by a whistleblower and brilliant young programmer from Open AI named Leopold Aschenbrenner.

In this document, he discusses how rapidly AI is going to leap across many OOMs, (Orders of Magnitude) to improvements in intelligence and speed.

In the very near future, by 2027 he predicts AGI will be achieved. These will be systems that will be as smart or smarter than a human with upper IQ.

This means essentially, that every cognitive function, everything that involves thinking that is carried out by humans in the current market can probably be performed just as well, or better, by AGI.

Insurance, medicine, finance, bookkeeping, government, data processing will all be able to be AI automated by 2027. Not that they will be automated by that time, but automation will begin to sweep through all aspects of industry and commerce that can be automated.

That could mean that there will be no need for radiologists to interpret radiology images, no need for any insurance claim adjusters, receptionist to take doctors’ appointments. All cognitive jobs will be heading for obsolescence from 2027, according to Aschenbrenner’s report. So, AI will be your co-worker. Or your staff if you are an independent business owner. Or maybe your boss!

Now, that’s just AGI—artificial general intelligence.

That’s not superintelligence! Superintelligence (when it occurs) will create an intelligence explosion! And superintelligence is achieved when AGI systems are able to function as AI researchers and developers themselves.

So, imagine you have a giant server farm of general intelligence systems, each given the task to improve themselves and become smarter and more capable.

They are instructed to: “Modify your algorithms, experiment with new approaches, change the overall architecture, until you have breakthroughs.”

“Then when you have breakthroughs, share your breakthroughs with all the other AI systems in the facility, and let them use them to make more breakthroughs.”

Then what you have is a feedback loop of superintelligence, taking order of magnitude leaps in capabilities, making each other smarter and more capable.

And this would all be happening at the speed of compute, with no longer any need for human developers.

Here is a link to the summary page of the Situational Report.

There are links to all other pages in the report from this introductory page.