AGI And AI Superintelligence Are Going To Sharply Hit The Human Ceiling Assumption Barrier

In today’s column, I examine an unresolved question about the nature of human intelligence, which in turn has a great deal to do with AI, especially regarding achieving artificial general intelligence (AGI) and potentially even reaching artificial superintelligence (ASI). The thorny question is often referred to as the human ceiling assumption. It goes like this. Is there a ceiling or ending point that confines how far human intellect can go? Or does human intellect extend indefinitely and nearly have infinite possibilities?

Let’s talk about it.

This analysis of an innovative AI breakthrough is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here).

Heading Toward AGI And ASI

First, some fundamentals are required to set the stage for this weighty discussion.

There is a great deal of research going on to further advance AI. The general goal is to either reach artificial general intelligence (AGI) or maybe even the outstretched possibility of achieving artificial superintelligence (ASI).

AGI is AI that is considered on par with human intellect and can seemingly match our intelligence. ASI is AI that has gone beyond human intellect and would be superior in many if not all feasible ways. The idea is that ASI would be able to run circles around humans by outthinking us at every turn. For more details on the nature of conventional AI versus AGI and ASI, see my analysis at the link here.

We have not yet attained AGI.

In fact, it is unknown as to whether we will reach AGI, or that maybe AGI will be achievable in decades or perhaps centuries from now. The AGI attainment dates that are floating around are wildly varying and wildly unsubstantiated by any credible evidence or ironclad logic. ASI is even more beyond the pale when it comes to where we are currently with conventional AI.

Human Intellect As A Measuring Stick

Have you ever pondered the classic riddle that asks how high is up?

I’m sure that you have.

Children ask this vexing question of their parents. The usual answer is that up goes to the outer edge of Earth’s atmosphere. After hitting that threshold, up continues onward into outer space. Up is either a bounded concept based on our atmosphere or it is a nearly infinite notion that goes as far as the edge of our expanding universe.

I bring this riddle to your attention since it somewhat mirrors an akin question about the nature of human intelligence:

• How high up can human intelligence go?

In other words, the intelligence we exhibit currently is presumably not our upper bound. If you compare our intelligence with that of past generations, it certainly seems relatively apparent that we keep increasing in intelligence on a generational basis. Will those born in the year 2100 be more intelligent than we are now? What about being born in 2200? All in all, most people would speculate that yes, the intelligence of those future generations will be greater than the prevailing intelligence at this time.

If you buy into that logic, the up-related aspect rears its thorny head. Think of it this way. The capability of human intelligence is going to keep increasing generationally. At some point, will a generation exist that has capped out? The future generation represents the highest that human intellect can ever go. Subsequent generations will either be of equal human intellect, or less so and not more so.

The reason we want to have an answer to that question is that there is a present-time pressing need to know whether there is a limit or not. I’ve just earlier pointed out that AGI will be on par with human intellect, while ASI will be superhuman intelligence. Where does AGI top out, such that we can then draw a line and say that’s it? Anything above that line is going to be construed as superhuman or superintelligence.

Right now, using human intellect as a measuring stick is hazy because we do not know how long that line is. Perhaps the line ends at some given point, or maybe it keeps going infinitely.

Give that weighty thought some mindful pondering.

The Line In The Sand

You might be tempted to assume that there must be an upper bound to human intelligence. This intuitively feels right. We aren’t at that limit just yet (so it seems!). One hopes that humankind will someday live long enough to reach that outer atmosphere.

Since we will go with the assumption of human intelligence as having a topping point, doing so for the sake of discussion, we can now declare that AGI must also have a topping point. The basis for that claim is certainly defensible. If AGI consists of mimicking or somehow exhibiting human intelligence, and if human intelligence meets a maximum, AGI will also inevitably meet that same maximum. That’s a definitional supposition.

Admittedly, we don’t necessarily know yet what the maximum point is. No worries, at least we’ve landed on a stable belief that there is a maximum. We can then draw our attention toward figuring out where that maximum resides. No need to be stressed by the infinite aspects anymore.

Twists And Turns Galore

AI gets mired in a controversy associated with the unresolved conundrum underlying a ceiling to human intelligence. Let’s explore three notable possibilities.

First, if there is a ceiling to human intelligence, maybe that implies that there cannot be superhuman intelligence.

Say what?

It goes like this. Once we hit the top of human intelligence, bam, that’s it, no more room to proceed further upward. Anything up until that point has been conventional human intelligence. We might have falsely thought that there was superhuman intelligence, but it was really just intelligence slightly ahead of conventional intelligence. There isn’t any superhuman intelligence per se. Everything is confined to being within conventional intelligence. Thus, any AI that we make will ultimately be no greater than human intelligence.

Mull that over.

Second, well, if there is a ceiling to human intelligence, perhaps via AI we can go beyond that ceiling and devise superhuman intelligence.

That seems more straightforward. The essence is that humans top out but that doesn’t mean that AI must also top out. Via AI, we might be able to surpass human intelligence, i.e., go past the maximum limit of human intelligence. Nice.

Third, if there isn’t any ceiling to human intelligence, we would presumably have to say that superhuman intelligence is included in that infinite possibility. Therefore, the distinction between AGI and ASI is a falsehood. It is an arbitrarily drawn line.

Yikes, it is quite a mind-bending dilemma.

Without some fixed landing on whether there is a human intelligence cap, the chances of nailing down AGI and ASI remain aloof. We don’t know the answer to this ceiling proposition; thus, AI research must make varying base assumptions about the unresolved topic.

AI Research Taking Stances

AI researchers often take the stance that there must be a maximum level associated with human intellect. They generally accept that there is a maximum even if we cannot prove it. The altogether unknown, but considered plausibly existent limit, becomes the dividing line between AGI and ASI. Once AI exceeds the human intellectual limit, we find ourselves in superhuman territory.

In a recently posted paper entitled “An Approach to Technical AGI Safety and Security” by Google DeepMind researchers Rohin Shah, Alex Irpan, Alexander Matt Turner, Anna Wang, Arthur Conmy, David Lindner, Jonah Brown-Cohen, Lewis Ho, Neel Nanda, Raluca Ada Popa, Rishub Jain, Rory Greig, Samuel Albanie, Scott Emmons, Sebastian Farquhar, Sébastien Krier, Senthooran Rajamanoharan, Sophie Bridgers, Tobi Ijitoye, Tom Everitt, Victoria Krakovna, Vikrant Varma, Vladimir Mikulik, Zachary Kenton, Dave Orr, Shane Legg, Noah Goodman, Allan Dafoe, Four Flynn, and Anca Dragan, arXiv, April 2, 2025, they made these salient points (excerpts):

• “The no human ceiling assumption: AI capabilities will not cease to advance once they achieve parity with the most capable humans.”
• “Our first claim is that superhuman performance has been convincingly demonstrated in several tasks. This provides a “proof of concept” that there exist concrete, clearly defined tasks for which human capability does not represent a meaningful ceiling for AI.”
• “Our second claim is that AI development exhibits a trend towards more generalist, flexible systems. Consequently, we expect superhuman capability to emerge across an increasingly large number of tasks in the future.”
• “Our third claim is simply that we observe no principled arguments why AI capability will cease improving upon reaching parity with the most capable humans.”

You can see from those key points that the researchers have tried to make a compelling case that there is such a thing as superhuman intellect. The superhuman consists of that which goes beyond the human ceiling. Furthermore, AI won’t get stuck at the human intellect ceiling. AI will surpass the human ceiling and proceed into the superhuman intellect realm.

Mystery Of Superhuman Intelligence

Suppose that there is a ceiling to human intelligence. If that’s true, would superhuman intelligence be something entirely different from the nature of human intelligence? In other words, we are saying that human intelligence cannot reach superhuman intelligence. But the AI we are devising seems to be generally shaped around the overall nature of human intelligence.

How then can AI that is shaped around human intelligence attain superintelligence when human intelligence cannot apparently do so?

Two of the most frequently voiced answers are these possibilities:

• (1) It’s a matter of sizing or scale.
• (2) It’s a matter of differentiation.

The usual first response to the exasperating enigma is that size might make the difference.

The human brain is approximately three pounds in weight and is entirely confined to the size of our skulls, roughly allowing brains to be about 5.5 inches by 6.5 inches by 3.6 inches in respective dimensions. The human brain consists of around 86 billion neurons and perhaps 1,000 trillion synapses. Human intelligence is seemingly stuck to whatever can happen within those sizing constraints.

AI is software and data that runs across perhaps thousands or millions of computer servers and processing units. We can always add more. The size limit is not as constraining as a brain that is housed inside our heads.

The bottom line is that the reason we might have AI that exhibits superhuman intelligence is due to exceeding the physical size limitations that human brains have. Advances in hardware would allow us to substitute faster processors and more processors to keep pushing AI onward into superhuman intelligence.

The second response is that AI doesn’t necessarily need to conform to the biochemical compositions that give rise to human intelligence. Superhuman intelligence might not be feasible with humans due to the brain being biochemically precast. AI can easily be devised and revised to exploit all manner of new kinds of algorithms and hardware that differentiate AI capabilities from human capabilities.

Heading Into The Unknown

Those two considerations of size and differentiation could also work in concert. It could be that AI becomes superhuman intellectually because of both the scaling aspects and the differentiation in how AI mimics or represents intelligence.

Hogwash, some exhort. AI is devised by humans. Therefore, AI cannot do better than humans can do. AI will someday reach the maximum of human intellect and go no further. Period, end of story.

Whoa, comes the retort. Think about humankind figuring out how to fly. We don’t flap our arms like birds do. Instead, we devised planes. Planes fly. Humans make planes. Ergo, humans can decidedly exceed their own limitations.

The same will apply to AI. Humans will make AI. AI will exhibit human intelligence and at some point reach the upper limits of human intelligence. AI will then be further advanced into superhuman intelligence, going beyond the limits of human intelligence. You might say that humans can make AI that flies even though humans cannot do so.

A final thought for now on this beguiling topic. Albert Einstein famously said this: “Only two things are infinite, the universe and human stupidity, and I’m not sure about the former.” Quite a cheeky comment. Go ahead and give the matter of AI becoming AGI and possibly ASI some serious deliberation but remain soberly thoughtful since all of humanity might depend on what the answer is.