IA > AI — How the original IA path to the Singularity solves the AI control problem

Intelligence Amplification

The Singularity is the messiah of our time. We yearn for super intelligence, prophesy and await its appearance with reverence and wonder, and our passion allows us to believe we’re actually going to make it.

The truth is, we’re not very smart as a species, that is, collectively. We know this, and we’re looking to AI to supply the critical boost in intelligence we know we’re going to need to evolve.

But there is another path to the Singularity, much less well known than AI, and that is Intelligence Amplification (IA).

What is IA?

Intelligence amplification (IA) (also referred to as cognitive augmentation and machine augmented intelligence) refers to the effective use of information technology in augmenting human intelligence.

So the difference between IA and AI is that IA increases your personal intelligence by augmenting your mind / brain, while AI is a standalone machine-only intelligence.


Which will happen first, AI or IA?

It’s very difficult to predict either. There is a tremendous bias for wanting IA to come first, because of all the fun movies and video games with intelligence-enhanced protagonists. It’s important to recognize that this bias in favor of IA does not in fact influence the actual technological difficulty of the approach. My guess is that AI will come first because development is so much cheaper and cleaner.
Both endeavours are extremely difficult. They may not come to pass until the 2060s, 2070s, or later. Eventually, however, they must both come to pass — there’s nothing magical about intelligence, and the demand for its enhancement is enormous.

So it could be decades before we see a strong IA or AI, but when we do it’ll first probably be an AI.

This is not good.

We need a massive increase in intelligence in our world right now, I’m sure you’ve noticed. We can’t wait… Also, if AI comes first then IA probably can’t be used to control it. And if not IA to control it, what? AI has to be controlled.

Fortunately, there exists another form of IA that is entirely different from augmentation. In fact, it turns out to be the original vision of IA.


Vernor Vinge

Networks and Interfaces

Back in 1993, Vernor Vinge, the acclaimed mathematician and award winning science fiction writer who coined the term, Technological Singularity, explained that IA is actually more likely to lead to the Singularity than a general AI.

When people speak of creating superhumanly intelligent beings, they are usually imagining an AI project. But as I noted at the beginning of this paper, there are other paths to superhumanity. Computer networks and human-computer interfaces seem more mundane than AI, and yet they could lead to the Singularity. I call this contrasting approach Intelligence Amplification (IA). IA is something that is proceeding very naturally, in most cases not even recognized by its developers for what it is. But every time our ability to access information and to communicate it to others is improved, in some sense we have achieved an increase over natural intelligence.

Note that Dr. Vinge is not talking about augmentation here. He’s talking about super intelligence emerging from new forms of human interaction, made possible by the advent of new kinds of networks and user interfaces.

He goes on to clarify…

I am suggesting that we recognize that in network and interface research there is something as profound (and potentially wild) as Artificial Intelligence. With that insight, we may see projects that are not as directly applicable as conventional interface and network design work, but which serve to advance us toward the Singularity along the IA path.

So our current understanding of IA is not at all what the progenitor of the Technological Singularity himself had in mind.

And now comes the part that might shock you a bit… Dr. Vinge’s vision of IA has actually been brought to life. Yes, a fully functional minimum viable product exists and is ready for private alpha prior to public launch. And yes, this new IA product is indeed a network-interface breakthrough.

The new IA will allow us to solve our problems, virtually, through the network, collectively, by engaging as individuals in open conversation on the issues that matter to us personally. Doing so will produce collective intelligence, via consensus, on whatever we focus.

Using the new IA, top experts will articulate their truths, there will be great debates, and through it all, the people, the audience, will feed back on all the expert articulations. The maximally amplified articulations, the ones that rise to the top, effectively represent our collective will, mind, decisions, and intelligence on the myriad subjects and issues we focus on.


The Current State of IA

So what is the technological landscape into which our new form of IA will emerge? As we’ve seen, IA as currently understood is about augmentation.

Elon Musk has a new company called Neuralink building what he calls a neural lace. It will eventually become a direct thought interface to the coming Artificial Super Intelligence (ASI).

With neural lace installed as a new artificial layer atop your cortex, and millions of electrodes connected to your cortical neurons, you’ll be in a persistent mind meld with ASI. It will think with you and help you solve your problems, all of them, inside your own head, lightning fast. No more deliberating, wracking your brain, no more uncertainty, anxiety, instead… instant right every time. In every situation the thought that comes into your mind will be the right one, the optimal one, the most intelligent one.

Bottom line, neural lace + ASI will in theory make us super intelligent, like ASI, and thus better able to control ASI. By entering into symbiosis and controlling the identity of the new symbiotic being we make our continued existence essential.

But again, probably at least a decade or more for something like neural lace to materialize, with ASI coming in pieces both before and after.

Elon says he’s going to focus neural lace first on helping brain damaged people, huge potential there, and he’s promising something on that front in four years.

Another area where neural lace shows promise, beyond the thought interface to ASI, is downloading units of knowledge directly into our brain, memories as strong as our own that enable us to do new things.

If this does prove possible neural lace will be even more devastatingly awesome, but again, we’re probably talking decades and we need our Kung Fu now. The Machines are coming.


The Control Problem

Elon and others have been focusing intensely of late on the dangers of runaway AI, an AI-driven intelligence explosion wherein each successive generation of AI builds a superior version of itself. It’s called recursive self improvement, and it’s a Singularity scenario that is extremely dangerous… for us.

The idea is that ASI will be so superior we’ll give it complete control of every aspect of our lives. This may work wonderfully for some time, but eventually a malfunction will occur and we’ll all be terminated. It does seem inevitable we humans will eventually be seen as parasites by ASI, you’ve seen the movie.

Takeaway is we can never allow AI to get near the strange attractor that is recursive self improvement, because that’s where we lose control, and once lost we may never regain it.

Fortunately, Elon has suggested that neural lace might actually be able to solve the control problem.

“Some high-bandwidth interface to the brain will be something that helps achieve a symbiosis between human and machine intelligence, and maybe solves the control problem and the usefulness problem,” Musk said of the need for humans to merge their minds with computers, according to CNBC.

Neural lace is a way for us to communicate with ASI in a much richer way, a thought interface. It will allow us to merge with ASI, to essentially become one with ASI.

So it appears the only way to control ASI is by merging with it, through symbiosis. ASI can never be allowed to see humanity as “other.” We must become a single integrated being, with us supplying the executive and identity components. The question is how to achieve that.

Those who install neural lace will literally become a new super intelligent species, and this does appear to be the inevitable future of humanity.

However…

If ASI is literally inside our heads, physically and mentally, if we become true symbiotes, the control problem does apparently diminish greatly, but the threat of a malfunction remains. What if the human-ASI cyborg becomes unstable, schizophrenic, self destructive, suicidal?

An argument could be made that an implanted neural lace will actually make us more vulnerable to a rogue ASI, not less, even in a symbiotic relationship. Because once something does go wrong, everything depends on who controls the off switch.

I’m confident Elon and his Neuralink team will develop a compelling failsafe solution, but Elon is also keenly aware that we need a solution to the AI control problem now, we can’t wait for a decade or two. AI is already going exponential.

Regulation is the obvious next step, and may serve as a stopgap (research will go dark), but it seems inevitable that ASI will eventually emerge, in spite of all regulations, so a permanent solution remains essential.

I agree the permanent solution must be an IA, but not necessarily the implantable augmentation kind. Instead, a new media system that produces collective intelligence, the network-interface form of IA Dr. Vinge envisioned. I have no doubt it can and will solve the control problem.

And the beauty is, it won’t take decades because it’s already built. We just have to start using it. It can emerge simultaneous with AI and provide a powerful new control mechanism, which I touch on below.

As Elon has pointed out, we must compel our government leaders to act quickly, proactively, preemptively, to regulate AI so as to prevent an AI-driven intelligence explosion.

And the way we compel them? That is precisely the kind of thing the new IA does best, as a media system capable of producing collective intelligence and collective decision making through consensus.


The Intelligence Amplifier

Neural lace will facilitate communication with ASI within the brain, but in the case of our new form of IA, ASI becomes CSI, Collective Super Intelligence.

ASI and CSI will both be super intelligences, but whereas ASI is likely to take decades to appear, CSI will emerge soon after we start using the new IA media system.

In order to understand how the new IA works we need to realize that intelligence amplification is basically what happens when we humans form a consensus.

In the consensus making process, ideas are articulated (by experts typically), and the public/audience feeds back, agreeing and amplifying their articulations or disagreeing and attenuating them.

These articulated ideas represent units of intelligence, as does the feedback given by each and every audience member.

And here is where the magic happens… As it turns out, if you aggregate and average all this feedback on a set of articulated ideas, the one receiving the greatest amplification (i.e., the most and best feedback) usually turns out to be the right one, the most intelligent one.

James Surowiecki explains why in his 2005 book, The Wisdom of Crowds

[There are] four conditions that characterize wise crowds: diversity of opinion, independence, decentralization, and aggregation. If a group satisfies these conditions its judgment is likely to be accurate. Why? At heart, the answer rests on a mathematical truism. If you ask a large enough group of diverse, independent people to make a prediction or estimate a probability, and then average those estimates, the errors each of them makes in coming up with an answer will cancel themselves out. Each person’s guess, you might say, has two components: information and error. Subtract the error and you’re left with information. [For example, ask] a hundred people to run a 100-meter race and then average their times. The average time will not be better than the time of the fastest runners. It will be worse. It will be a mediocre time. But ask a hundred people to answer a question or solve a problem, and the average answer will often be at least as good as the answer of the smartest member. With most things, the average is mediocrity. With decision making it’s often excellence. You could say it’s as if we’ve been programmed to be collectively smart.

The aggregation condition is where intelligence amplification occurs, leading to the emergence of collective intelligence. CI materializes in the form of a consensus that peaks as all facts are factored and all voices heard.

As we aggregate feedback on these expert posts (articulations) they are literally amplified by the expressed intelligence of each audience member.

Over time, all facts that emerge can be factored into successive versions of the post until the interested collective pretty much agrees that the whole and complete truth is known.

Now imagine we can do this on every subject in human knowledge, every issue that divides us, every monumental tipping point crisis that threatens us with extinction, every problem we collectively face.

Collective intelligence grows in magnitude, it is literally amplified as it evolves toward a single articulation, the consensus.


A New Natural Resource

CSI will emerge because crowds are wise when they’re diverse, decentralized, and independent, and the Internet makes that possible, creating the wisest crowd the world has ever known.

Because of the Internet, a vast undiscovered ocean of CSI lies before us, we’re immersed in it, a boundless reservoir just waiting to be tapped and put to use, its product more valuable than oil to our future on this planet.

Our network-interface IA turns out to be the mechanism for tapping into this boundless CSI reservoir, by aggregating feedback on individual articulations on individual subjects.


SSI and the Solution to the Control Problem

So there are going to be two sources of super intelligence, ASI and CSI. They will coexist, and just like the various regions and parts of the brain, they will play different roles.

What we must avoid at all costs is an uncontrolled ASI that sees us as other. Fortunately, the new IA, which harnesses the power of the collective, is capable of supplying the all critical executive/identity functionality. The new IA media system will emerge simultaneous with AI and serve as identity for the Singular Super Intelligence (SSI = CSI + ASI) as it develops over time.

What does the SSI believe to be true about any given subject, issue, or unfolding story taking place in the world? With collective humanity as its executive/identity source via the new IA, the emergent SSI will always have an answer, it will always serve human beings above all, and the ASI will never be able to become an other.


From Amplification to Explosion

The network-interface IA involves at least two forms of feedback-driven intelligence amplification.

First there is the amplification of expressed intelligence (ie, the expert posts) using aggregated feedback from the public audience.

Then, as the emergent CSI from this first amplification renders the public more intelligent, the subsequently emerging CSI becomes more intelligent still.

This second form of intelligence amplification, with individuals getting smarter as CSI gets smarter in a tight feedback loop, gives rise to the classic intelligence explosion, and this is how Vernor Vinge’s network-interface IA can produce the Singularity.

By the time Elon’s neural lace becomes available CSI will already be mature, so it will probably be CSI that first becomes available to the individual within their own mind, not ASI.

So we do in fact want the CSI explosion to occur, and we definitely do not want the ASI explosion to occur. Only one of the two explosions can become the Singularity. If it is CSI, then subsequently emerging ASI will be safely integrated piecemeal into the larger super intelligent being (SSI) over time. However, if the ASI explosion occurs first then CSI may never happen, and our chances of survival long term appear greatly diminished.


Expected Outcomes

The new network-interface IA will produce CSI. The predictions and decisions made by the collective using this new IA will consistently prove superior to those made by individual human experts, proving the emergent collective intelligence is in fact superhuman.

Using the new IA, the collective will focus first on AI, its potential benefits and dangers, and decide where and how to place limits. Led by the AI community, the collective will use the new IA to decide on a plan for the future development of AI.

All threats facing humanity will be addressed using the new IA in similar fashion. CSI will be brought to bear upon them all, and consensus will emerge around a solution to each.

People will get smarter as they’re exposed to CSI, they will come to rely on CSI, and the relationship will become symbiotic. This will take place over the next many years, during which time neural lace and AI will be developed. ASI will likely emerge as a series of highly specialized units delivered via API, and these will be safely integrated into the SSI superstructure that CSI heads, as they appear.

It bears repeating. The smarter the people get absorbing CSI, the smarter future emergent CSI becomes, and the smarter the emergent CSI becomes the smarter its users, the people, in turn become. It’s a positive feedback loop capable of producing an intelligence explosion, and therefore the Singularity. And in this scenario neither ASI nor neural lace is actually necessary.

Dr. Vinge was right.