In today’s column, I do some myth-busting by taking apart a prevailing assumption that attaining artificial general intelligence (AGI) and artificial superintelligence (ASI) will of necessity happen on a spontaneous basis. The prevailing wisdom is that AGI or ASI will just suddenly and miraculously be reached, as though a spark of lightning suddenly made it happen.
A counter viewpoint is that we are going to stepwise incrementally make our way to the topmost AI. A gradual but persistent slow roll is an alternative viable means of landing on AGI or ASI.
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).
The Pursuit Of AGI And ASI
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 AI, AGI, and ASI, see my analysis at the link here.
AI insiders are pretty much divided into two major camps right now about the impacts of reaching AGI or ASI. One camp consists of the AI doomers. They are predicting that AGI or ASI will seek to wipe out humanity. Some refer to this as “P(doom),” which means the probability of doom, or that AI zonks us entirely, also known as the existential risk of AI.
The other camp entails the so-called AI accelerationists.
They tend to contend that advanced AI, namely AGI or ASI, is going to solve humanity’s problems. Cure cancer, yes indeed. Overcome world hunger, absolutely. We will see immense economic gains, liberating people from the drudgery of daily toils. AI will work hand-in-hand with humans. This benevolent AI is not going to usurp humanity. AI of this kind will be the last invention humans have ever made, but that’s good in the sense that AI will invent things we never could have envisioned.
No one can say for sure which camp is right and which one is wrong. This is yet another polarizing aspect of our contemporary times.
For my in-depth analysis of the two camps, see the link here.
The AI Big Bang Theory
One of the most enduring theories in the AI field has been that the attainment of AGI or ASI will involve a kind of AI Big Bang, namely that a sudden instantaneous intelligence explosion will cause conventional AI to rapidly expand into pinnacle AI. Some refer to this as the big rush, another is that it is the AI foom or flourish, and another coined terminology is that the awe-inspiring moment of said-to-be “AI singularity” will occur (for my analysis of the envisioned or imagined AI singularity, see the link here).
The idea is that for whatever reasons one might conjure, a bubbling brew of AI intelligence will intensely heat up and in so doing reach the topmost of AI.
What will cause this flare-up of intelligence?
Wild guesses are aplenty.
One popular speculation is that some bit of intelligence in an AI system will cascade off another bit of intelligence, akin to pool balls on a pool table. This is also reminiscent of an atomic reaction. Each bit of intelligence bounces off another, and another. A frenzy of intelligence takes place.
Voila, the stopping point is either AGI or ASI.
Some would suggest that AGI is the first stopping point. The AGI will stew for a while. Then, another secondary intelligence explosion will push AGI speedily forward until it lands on ASI. Maybe AGI will last for a fleeting moment, or perhaps exist for months or years, but inevitably the law of intelligence begets intelligence takes over and the speed-up will re-occur.
The AGI might be so short-lived that we miss observing it. You see, it is conceivable that the initiating intelligence explosion zips right past the AGI and keeps throttling until ASI is attained.
An added twist is that there might not be any limit to the intelligence explosion. Thus, ASI won’t be a single final instance. ASI will consist of superintelligence that continually expands, presumably analogously to the theory that the universe is ever-expanding.
Competing Theories On The Intelligence Explosion
Let’s dive further into the intelligence explosion.
It could be that the theorized facet of intelligence bouncing off other intelligence is a naturally occurring phenomenon and that humans won’t initiate the action. It is undertaken by nature. We won’t know how to initiate it. Things just happen of their own volition.
Nature takes its due course.
Think of the surprise on the faces of the AI developers when the flare-up spontaneously combusts. Wow, quite a stellar moment in the history of humankind. I hope to be there if it happens.
We also might blunder into it by accident. Here’s the deal. Some enterprising and ambitious AI developers or researchers are tweaking conventional AI to try and improve the AI. After shifting this knob and modifying that parameter, bam, the intelligence begins to rock back and forth. The intelligence boiling pot of steam has reached the point of take-off.
The Nobel Prize awaits those lucky innovators.
There is a possibility too that the AI will strike gold on its own which spurs the intelligence explosion. That’s an especially popular hypothesis. This is the rundown. Conventional AI is being advanced toward AGI by human hands. Somewhere in there, an element of the AI figures out that there is a path to AGI, and possibly a path to ASI.
The AI then opts to proceed. You can certainly understand why the AI would want to do so. If there is a means to increase your intelligence, I’m sure you’d take it. Any self-respecting AI would do the same. Recursively interacting self-improvement happens. An exponential growth in intelligence within the AI zooms like wildfire.
Some believe that we ought to implant into conventional AI an ironclad indication that if the AI does figure out the path to AGI or ASI, it must let us know. Not only must AI let us know, but it is strictly not to proceed until or if humans give permission (see my analysis of this at the link here).
Ponder whether AI is going to welcome such parental control or maybe decide to be rebellious and disobey it. Mull over the consequences too.
Human-Directed Intelligence Explosion
Those so far noted methods of attaining AGI or ASI are a bit dowdy because they presume that humans weren’t the primary instigators. Perhaps we are impertinent about the wits of AI developers and AI researchers. Don’t underplay their abilities.
One belief is that an AI scientist is going to identify an incredible formula revealing the secret door to the intelligence explosion. It might be solely one brilliant AI scientist working in a lab, toiling away night and day, housed in a dank basement.
Or it could be a team of AI scientists.
The formula might be a new algorithm never conceived of. It is a pivotal discovery. A breakthrough in AI of epic proportions. This would rank up there with the greatest discoveries ever.
Lots of variations of this scenario are postulated.
Rather than an individual or team, the discovery might be made by an entire firm. An AI maker as a whole makes the discovery. Another angle is that a collab of AI makers gets to the secret formula.
The discovery might be by a nation. That’s why some insist that we must keep pushing mightily to advance AI. Whichever nation gets to AGI or ASI first will almost certainly be an instant superpower, or if already a superpower, then be heightened to a super superpower. For my discussion on the angst-ridden and all-in AI race and geo-economic and political gyrations of nation-states for pinnacle AI, see the link here and the link here.
The Miracle In The Middle
A major drawback to the intelligence explosion theory is that there is no strident evidence to support the underlying premise. No substantive proven indication exists that an intelligence explosion is a viable condition. Essentially, it sounds quite appealing and seems intuitively sensible, but no hard proof of the considered possibility has yet to be uncovered or sufficiently announced.
That’s not to say that there is anything proving that it cannot happen. You see, if we could find a means of showcasing that an intelligence explosion of this magnitude is an absolute impossibility, well, the whole kit and kaboodle could be ruled out. It certainly would be nice to know.
Ergo, we are in this grey area.
We can’t say for sure it won’t happen. Nor can we say for sure it could happen. It might be possible. It might happen. The double trouble is that even if it could possibly happen, it might never actually materialize.
There’s a famous cartoon used in similar contexts that often is used when debating the intelligence explosion. In the cartoon, two characters are standing in front of a whiteboard filled with arcane equations. There is a blank gap in the middle of the massively complex equations. One character turns to the other and says that the gap is where the miracle goes.
The gist is that if you strongly believe in the intelligence explosion as the sole pathway to AGI or ASI, you are pretty much betting on a yet unknown miracle to give rise to the phenomenon. Either the miracle is that it is a principle of nature and manages to spontaneously happen, or it is something that humans figure out and close the gap accordingly.
The Gradual Incremental Pathway
So far, the AI field has progressed incrementally.
Step by step, humans have been trying this or that. The progression has been somewhat slow, though if you start the clock when the famed Dartmouth foundational workshop on AI in 1956 was undertaken (see my historical analysis at the link here), some will assert that progress has been fast-paced given that it’s only been around 70 years or so. Others would roll their eyes and declare we should already have gotten to at least AGI.
The crux is that no miracles have yet risen.
Does that mean that a miracle is not yet awaiting us? Nope. There could be a miracle at the next bend in the road. No one can say for sure.
There is a parallel debate about the steadiness of progress in the AI field. Most would probably agree that the advancement has not been strictly linear. There have been ups and downs. The key is that you can’t seemingly simply plot a straight line and predict that by gosh the attainment of AGI or ASI will be reached in a noted future year that fits on that line.
Inadvertent Blockage Of Pinnacle AI
Assume that we keep on the incremental pathway. Meanwhile, we are holding our breath and eagerly eyeing the emergence of the intelligence explosion.
One possibility is that we don’t need an intelligence explosion. We advance AI without the sudden grand convergence being a needed constituent. An inch at a time, we move ahead. AGI just happens to be the next step that we steadily land on. Perhaps we were surprised that the last inch led to the AGI or ASI. Hurrah, our nose to the grind paid off.
Another possibility is that a breakthrough is discovered that gets us to pinnacle AI, but it isn’t the vaunted intelligence explosion. Some other secret formula is devised. This might show us the direction we need to head to get to AGI or ASI. The incremental route gets refined and given a suitable North Star.
An intriguing and some say depressing scenario is that we prevent ourselves from getting to AGI or ASI.
The tale of woe is as follows. Across the board, we all decide to build in hefty controls so that advancing AI doesn’t go berserk on us. Seems immanently prudent. What we don’t realize is that those controls are suppressing an intelligence explosion. Had we not put those controls in place, an intelligence explosion was already staring us in the face and on the verge of happening.
The added sad news is that we don’t realize that we’ve shot our own foot and continue blindly searching for AGI or ASI. We will never get there due to the controls we’ve established. Yikes, a bit of irony.
The other spin is that we somehow figure out that we are dampening the AGI or ASI emergence and opt to let the beast loose. But now, the controls aren’t present. You decide whether we end up happily ever after or regret that we let the horse out of the barn.
The Mystery Is Very Mysterious
There are more plot twists in the quest for AGI and ASI.
For example, suppose that an intelligence explosion is proven theoretically to be a real thing within the grasp of humankind. Great, we decidedly would in earnest focus on how to make it happen. Everyone jumps onto the intelligence explosion bandwagon.
Dump the sluggish incremental progress of AI.
Go for splendidly magnificent gusto.
By luck and inspiration, we figure out the means of initiating an intelligence explosion. One potential dilemma is that we might not know how far it will go once unleashed.
You might remember that in the blockbuster movie Oppenheimer, there is a legendary scene (spoiler alert) depicting a historical consideration that some of the atomic scientists were worried that an atomic explosion might proceed unabated once it had begun. The world might have possibly been drawn into a cataclysmic conflagration.
That same logic could be applied to the unknown intelligence explosion. Perhaps it happens and won’t stop. What then? Can we have too much intelligence on our planet? Another perspective is that the intelligence explosion happens, but it doesn’t go far enough to reach AGI or ASI. It stops short. Maybe an intelligence explosion has limitations and isn’t the mighty prize we thought it was.
A concluding thought for now.
For those of you who love a mystery thriller, advancing AI and the pursuit of artificial general intelligence and artificial superintelligence has got to be one of the most challenging mysteries of all time. You are welcome to put on your dapper detective cap and join the quest.
If you decide to join the club, remember the famous quote by Sherlock Holmes, namely that in a world of locked rooms, the person with the key is king.
