Can generative artificial intelligence help build a global hive mind?

On this planet of science fiction, the concept humanity will in the future join our brains collectively into a worldwide “hive thoughts” dates again to Olaf Stapledon’s 1930 novel The Final and First Males. It tells the story of a fictional “future story” by which humanity biologically evolves right into a species that may talk telepathically and type collective minds of extraordinary intelligence.

In the actual world, this pursuit known as collective superintelligence, and it does not require telepathy or different fancy units. As a substitute, it is going to use new generative AI applied sciences to attach massive human teams into real-time dialogue programs that can permit us to resolve advanced issues utilizing our mixed information, knowledge, and understanding in highly effective new methods.

This pursuit has been my private focus as a synthetic intelligence researcher for the previous decade, and I imagine it has the potential to create superintelligent programs that uphold human values, morals, and pursuits on the core of each understanding, analysis, or choice. In fact, to many individuals the concept of ​​massive human teams pondering collectively in real-time programs appears unnatural and even creepy, however Mom Nature would disagree.

The truth is, many social species have developed naturally on this route, growing the flexibility to make fast selections in massive teams that far exceed the flexibility of particular person members. Biologists name this swarm intelligence, and it permits colleges of fish, swarms of bees, and flocks of birds to rapidly clear up life-and-death issues at a stage of intelligence that far exceeds the psychological capability of their particular person minds.

One outstanding type of swarm intelligence that has impressed my educational analysis over the previous decade is the common-or-garden college of fish. Though it appears to be like easy at first look, the underlying dynamics permit 1000’s of members to make advanced selections and not using a single individual in cost. The truth is, colleges of fish could make good selections even when no member has sufficient info to resolve the issue. Contemplate the picture beneath:

The aforementioned college of fish is confronted with a hypothetical downside of utmost significance: three predators approaching from three completely different instructions. At this level, nobody is aware of about all three threats. The truth is, most are unaware of any threats. There’s a small group of fish within the higher left nook that’s conscious of the primary predator, a small group within the decrease left that’s conscious of the second predator, and a small group within the higher proper nook that’s conscious of the third predator. Most different fish are unaware of the hazard.

So how can this huge group, the place every member has restricted info (and nobody has all the knowledge), clear up this life-and-death downside and transfer rapidly in the suitable route? The very first thing to know is that fish have a particular organ on the edges of their our bodies referred to as the lateral line that permits them to trace the pace and route of motion of close by fish primarily based on the strain and vibrations within the water round them. Utilizing this organ, they’ll understand the intentions of their neighbors (that’s, the pace and route by which, in keeping with these fish, the varsity ought to be going). This communication is two-way, so we will take into account it an area dialogue by which small teams determine on the motion of their small a part of the varsity.

It is attention-grabbing, nevertheless it does not clarify how world selections are made. Finally, the group on the suitable, seeing a predator approaching, will doubtless determine that the varsity ought to transfer to the left. On the identical time, the 2 teams on the left are prone to determine that the varsity ought to transfer to the suitable. And the teams within the center, who’re unaware of the predators, are prone to proceed shifting within the route they have been already on. So how does this get resolved in a single fast answer that avoids pressing threats?

The magic occurs as a result of every fish within the college “consults” with a unique group of close by neighbors. Which means many “overlapping conversations” are occurring on the identical time, permitting info to unfold rapidly all through the varsity. As proven beneath, the result’s a fast and decisive collective answer to the issue.

Thus, education fish could make quick and environment friendly selections in massive populations, even when all members have restricted info. Such a talent could be much more highly effective for giant teams of individuals. In spite of everything, the issues confronted by teams of individuals are rather more advanced and have rather more prospects. This begs the query: Can massive teams of individuals cause in actual time with the effectivity of a faculty of fish and make optimized selections rapidly?

For years, this aim appeared unattainable. It is because human conversations have been best in small teams of 4 to seven individuals and deteriorate quickly as teams get bigger. It is because the “airtime per individual” is step by step lowering, whereas the ready time for others to reply is steadily growing. By 12-15 individuals, the dynamic of the dialog adjustments from a considerate debate to a collection of more and more disjointed monologues. As much as 20 individuals, the dialogue ceases to be a dialog in any respect. This downside appeared intractable till advances in generative AI opened up new options.

The ensuing expertise known as Conversational Swarm Intelligence (CSI), and it guarantees to allow teams of doubtless any measurement (200, 2,000, and even 2 million individuals) to debate advanced issues in actual time and rapidly discover options with significantly enhanced intelligence. Step one is to divide the inhabitants into small subgroups, every sized for considerate dialogue. For instance, a gaggle of 1,000 individuals could be divided into 200 subgroups of 5 individuals, every of which works to its personal chat or video convention session. In fact, this does not create a single unified dialog—it creates 200 parallel conversations.

As described above, colleges of fish clear up this downside by overlapping native teams, permitting info to unfold quickly all through the inhabitants. Sadly, we people did not have the flexibility to have a number of conversations on the identical time. The truth is, if we attempt to concentrate to 2 conversations, we instantly get confused and can’t concentrate on both. That is generally known as the “cocktail social gathering downside” as a result of it usually occurs when small teams collect inside earshot of one another. For those who attempt to concentrate to the dialog subsequent door, you’ll instantly lose the that means of the dialogue you’re in.

So how can we overcome this human limitation?

CSI’s expertise solves this downside by inserting LLM-based “conversational surrogates” into every subgroup. These AI brokers are tasked with processing real-time human info in a delegated group and sharing that info with surrogate brokers in a number of different teams. Reception brokers specific the concepts they’ve obtained of their teams in a pure first-person dialogue. On this manner, every subgroup will get a synthetic member that seamlessly participates in overlapping conversations, guaranteeing that info flows freely all through the inhabitants.

Quite a few latest research present that this strategy is efficient. For instance, a 2023 examine from Carnegie Mellon College in contrast real-time conversations between about 50 individuals in conventional chat rooms and chat swarms. Through the use of the CSI framework, teams have been capable of have extra coherent conversations that rapidly converged on options. Moreover, every individual contributed 50% extra content material (on common) than members utilizing conventional strategies.

However does it improve group intelligence?

To discover this, a follow-up 2024 examine by researchers at Carnegie Mellon and Unanimous AI examined the flexibility of on-line teams of individuals to take IQ checks as a “hive thoughts” in actual time. The outcomes confirmed that teams of 35 individuals with a mean IQ of 100 (fiftieth percentile) might rating an efficient IQ of 128 (97th percentile) when utilizing CSI’s on-line platform referred to as Thinkscape. Though this examine used focus teams of solely 35 members, different latest research have efficiently examined teams of as much as 250 individuals.

Whereas the aforementioned research used text-based conversations, primary CSI strategies could be deployed for teleconferencing, video conferencing, and even VR conferences, permitting massive teams of tons of and even 1000’s of members to have interaction in sequential real-time conversations that successfully clear up issues, prioritize, brainstorm concepts and make selections, all with enhanced group intelligence. This has the potential to develop a variety of fields from enterprise collaboration and advertising analysis to civic engagement and deliberative democracy.

In the long run, this strategy may very well be used to create super-intelligent programs which are intrinsically aligned with human values, morals, knowledge and emotions. In concept, we might use CSI expertise to permit hundreds of thousands of individuals around the globe to “assume collectively” as a worldwide mind of brains to resolve probably the most advanced issues. To me, it is a safer route than counting on purely synthetic superintelligence, as AI programs could not help human values ​​or pursuits over time. This is the reason I imagine we want applied sciences like Conversational Swarm Intelligence and instruments like Thinkscape that use generational AI however hold individuals knowledgeable.

Louis Rosenberg is a longtime researcher within the fields of AI, collective intelligence, and blended actuality. He’s the CEO and Chief Scientist of Unanimous AI.