Superintelligence Follow Up
This idea of organizations as superintelligence has had me thinking more and I managed to use the right combination of keywords and found a fabulous series of articles on this subject by David Rostcheck. The links are:
Exploring Aggregate Intelligence. Superintelligence in Organizations | by David Rostcheck | Medium
The Effect of Connectivity on Aggregate Intelligence | by David Rostcheck | Medium
Man and Superintelligence. Exploring the Limits of Human Power | by David Rostcheck | Medium
So, it looks like I am not the only one thinking about this idea.
The thing I’ve been thinking about lately related to this idea is the impact of burn out of individuals to the superintelligence function.
I’ve been literally vibrating with positive and negative excitement about these ideas in relation to leadership:
Complexity / cognitive overload
Effect of burn out on the superintelligence
Lateral communication
Superintelligences in competition
Nodal heterogeneity
I apologize in advance for a bit of a downer in some of the downside thinking, but as engineers we are conditioned to try and spot the weaknesses where things will fail, and to attempt to mitigate those risks.
Lets take a really quick look at these one by one.
1. Complexity / cognitive overload
a. A major path to burn out is cognitive overload by those in leadership.
b. If the organization is flat and has a well functioning superintelligence, this is a resource that leaders can and should engage to cope.
2. Effect of Burn Out on Superintelligence
a. This is a big one. A superintelligence is composed of a set of nodes processing tasks and in communication with one another.
b. The level of function (ie: intelligence) of the superintelligence will be dependent on the speed and quality of communication as well as the processing power of the nodes.
i. If one person becomes burnt out a path of communication is either interrupted, or possibly even corrupted where faulty information is transmitted.
ii. Work load is moved to other individuals as the burnt out person is no longer functional.
1. The burn out is contagious, and can spread further impacting and potentially accelerating.
c. Overall though I see a positive. It is possible through the application of sound leadership to set up constructive resonances, and through the effect in #1, mitigate burn out risk.
3. Lateral communication
a. Rostcheck’s modelling work indicates that flatter is better, especially in smaller organizations. For those wanting to go down an interesting rabbit hole, take a look at Haier’s org structure and the RenDanHeyi philosophy.
b. I’ve written about this in the past with the comparison to rugby and how the extended periods of improvisational play are dependent on this lateral communication.
c. Again this is a positive. We have it in our grasp to be good communicators, and to develop others as good communicators.
4. Superintelligences in competition
a. Continuing from rugby, what happens in sport where teams (ie: superintelligences) are in competition? There are rules and a referee. An exception is ultimate frisbee. Maybe the culture of this sport should be studied further. In pick up basketball the game is also self policed. However, most sports include a referee with a mandate for (and in the following order as priorities):
i. Player safety
ii. Fairness
iii. The actual rules (or in rugby the laws of the game)
b. In the corporate world we have the imposition of regulation by government (another superintelligence) acting in the role of referee. But as with any referee, government can become corrupted.
c. Examples of competition:
i. This is a potentially interesting way to reframe the cold war as a conflict between two superintelligences: a communist one and a capitalist one.
ii. Corporations in competition in the market with each other.
d. What kinds of neural adaptations happen when there is this kind of competition? Surely not all of these adaptations are going to be to the benefit of humanity. I think there are infamous instances of how individual humans have been treated as disposable by these superintelligences. 20th century communism anyone? The late 19th / early 20th century corporate robber barons.
e. War.
f. In all of the above competitive scenarios, the character of head coaches and team captains matters a lot. Whether there will be cheating. Atrocities?
g. Overall I’m negative on this one.
5. Nodal Heterogeneity
a. One of the comparisons that people have talked about is the swarm or hive intelligence. This book is on my reading list: The Wisdom of Bees: What the Hive Can Teach Business about Leadership, Efficiency, and Growth: O'Malley, Michael: 9781591843269: Amazon.com: Books
b. Hive insects have a pretty finite set of roles / morphs. With this limited number of node types they are able to manifest very interesting group behaviour and incredible outcomes. Here I am thinking of termite mounds or bee hives in the wild.
c. I think human beings bring at least an order of magnitude more heterogeneity to the roles and sets talent and skills we bring to the world.
d. Do we have some people that will function as a “super-node”. I am certain the answer to this is yes. David mentions one leader who I consider a hero: Teddy Roosevelt. He was not your typical president and did some very unconventional things and made a very big difference. Teddy was certainly a “super-node”.
i. Could the leader be a type of amplification effect? The truly GOAT athletes make the people around them better. I think the same can be said for leaders in other oganizations.
e. Overall I’m neutral on this, and see this as an interesting aspect to look into further.
David Rostchek explores the ideas around whether leaders are controlled by, or control the superintelligences they supposedly lead. There are examples of leaders who have controlled the beast. In doing a startup we create the beast. I think leadership does represent a different type of node in the neural network of a superintelligence.
I don’t have a lot of answers here, but a lot of questions, and concerns. If AI agents are being trained on ideologically tainted data and information, and these agents are combined into a swarm or hive along with human intelligence that is also potentially ideologically captured, a new kind of superintelligence results. To rosily think that this is going to come without massive tail risk and the pain of failure is frankly irresponsible.
Our history of technological development is replete with examples of increasingly unexpected impacts with increasingly powerful technology:
1. Internal combustion engine and leaded gasoline
2. Fertilizer + pesticide + herbicide and the destruction of healthy soil ecosystems
3. Antibiotics + prophylactic use in animal husbandry as a growth promoter and the rise of super-bacteria
4. Social media and it’s impact on mental health
As we have developed forms of increasingly efficient communication thereby increasing our existing superintelligences the results have not always been good. Is this part of why and how the State has become larger and more powerful? When the State has become corrupted the effects become more profound: 20th century horrors committed by State actors pale in comparison to almost all other historical examples. As corporations have grown and become more powerful when they go wrong the effects are more profound: More and larger financial crises, pollution, various corruption scandals with increasingly globalized impacts.
The amelioration of some environmental problems (acid rain, air pollution) were not initiated through the benevolence of the corporations, nor the sound judgement of government. The initiation came from the collective efforts of a mobilized environmental movement that functioned as another superintelligence in conflict with both government and corporations.
I am quite convinced we need to be looking to history to understand events through the lens of superintelligence, reevaluating successes and failures, doing more advanced modelling, knowing that modelling, especially advanced modelling of complex systems involves making simplifying assumptions and that results of models need to be treated with caution. That said models can still provide insights, especially when we try to examine the boundaries and failure mechanisms.
While Teddy Roosevelt was a leadership hero, we have villains like Hitler and Stalin. Leadership matters. I think charismatically chosen leaders generally do not have a good record. How likely is it that with the help of AI we end up with a “super-charismatic”, super-influential leader? What will the super-consequences of this be?
Just as our current landscape is closer to Brave New World than it is to 1984, I think the advent of new forms of superintelligence will have a pervasive and subtle effect rather than a Matrix or Terminator character.
Closing
We will likely need to create governmental-regulatory and citizen-centric superintelligences to referee the private-sector superintelligences that are on their way. In the past regulatory oversight has lagged, usually requiring some kind of a disaster or social movement to be implemented. For the record I actually tend to be a bit of a “minarchist” favouring minimal intervention per my axiom: “as many rules/policies as needed and as few as possible”. I think the possible future disaster is so bad that we need to get ahead of this one.
Leadership matters, and I suspect that it will matter even more in this brave new world of superintelligence.