08 Mar | A Better & Shared Life Alignment
2023-03-0815-19 Mar | How do we Reach out to Others?
2023-03-17I finally invested consecutive days in one of my Sabbatical key topics: Complex Systems and Collective Intelligence. This somehow exotic topic comes down to the idea that groups of interconnected elements can create much more Complex, Unpredictable, Dynamical and Rich behaviors than any of the elements could. In this post, I explore examples of it in the Brain, ChatGPT and Human Organizations.
A Thousand Voting Mini-Brains.
These days I read the 2021 book “A Thousand Brains: A New Theory of Intelligence” where Jeff Hawkins presents the idea that the brain works via collective behaviors of thousand physical units: neocortical columns. It is not new that we have thousands of these mini-brains, each of them made of thousands of densely interconnected neurons (Vernon Mountcastle in the 70s). The somewhat novel idea is that each of these mini-brains can host hundreds of partial models of the world that are used in a collective behavior to create what we perceive as reality.
For example, the mini-brains located in the region linked to the eyes, create visual spatial models of, for example, a pizza while other parts of the brain store other sensory spacial models of the same pizza. This way the brain involves many partial spatial models that vote on what they think we are experiencing when we eat, hold or smell the pizza. There is no single model of pizza anywhere in the brain but the diverse partial models connect to form a process that is very robust at recognizing a pizza…or much more difficult objects or notions.
In itself, the idea that our mind is made of multiple agents and not of a formal deterministic flow is probably ancient. For example, a similar concept is called Society of Mind by Marvin Minsky in the 70s. Looks like we are adding more know-how from science to improve or translate to current terms the vision we have of ourselves and of what are intelligence and awareness. By the way, it’s challenging and fun!
Space and Movement are Part of Brain Learning.
Another important aspect of this process is movement. Since, remember, all models are sensorial plus spatial, they include distances and shapes. Comprehension is actually a dynamic process, so when we turn something around, touching and looking through different angles to “see it better”, we are actually creating or improving our spatial models.
This spatial capacity of the mini-brains is ideal to move around, find food and survive in nature, so it was perfected by natural selection over the ages. In us, the spatial nature of the models is used, according to Hawkings, also to comprehend high-level processes like language or science. In these cases, the spatial references map concepts, hierarchies of words, etc. When we think we are actually moving through these networks of entities. I like this concept the most. Like traveling inside us.
This idea of Democracy and Spatial References inside the Brain are better explained in this video.
ChatGPT Uses Voting and Movement to Write like Us?
Do we have examples of computers using this wisdom of the crowds to create complex and apparently intelligent collective behaviors? These explorations into the model of the brain remembered me of a recent and interesting blog post by Stephen Wolfram on how ChatGPT works. It is very long and super technical but worth mentioning because it has some parallels…and everyone is now talking about ChapGTP!
Imagine that every text, like this post or a book, is a self-consistent alignment of written ideas understandable by other humans. Like the mini-brains in the previous example, they have a certain model, in this case not of the world, but how to transform ideas into text. Moreover, texts follow rules and have elements, words instead of neurons, connected via clear rules and categories. In this case, we have texts as Individual Elements internally made of words and sentences.
Apparently, what ChatGPT does to write in a human-readable way is to ask all the billions of individual elements, in this case texts, it has read what they suggest as a possible answer and picks the best by a kind of voting system. Amazingly we can draw even more parallels to the brain. One is movement, chatGPT does this process not at once but word by word, like a path. The other similarity is the spatial model since it looks chatGPT uses also a coordinates system to separate words.
This is why ChatGPT feels so different from searching in google. It creates an answer from all existing ones instead of giving us a list of the best existing ones to pick from. Both methods are complementary and probably we will have them integrated ahead.
For those who ate are not yet in the ChatGPT world, you can go to chat.openai.com and just ask any complex question. Below I have two examples of this blog written by ChatGPT…I was really tempted to use them and save myself four hours of work!
On the implications of AI in our lives see this very recent post from Bill Gates who has been in close contact with ChatGPT makers.
Example of this post written by ChatGTP
My request to ChatGPT:
Write me a blog post between 200 and 400 words on the topic of the book 1.000 brains and how that reflects our individuality vs collective intelligence.
ChatGPT Answer
The concept of collective intelligence has been around for a while, but it has recently gained more attention with the release of the book “1000 Brains” by Jeff Hawkins. The book explores the idea that our brains do not function as individual units, but rather as parts of a collective intelligence that is greater than the sum of its parts.
Hawkins argues that our brains are not just individual processors of information, but rather they are constantly communicating with each other to create a cohesive understanding of the world around us. He compares this to a swarm of bees, where each bee has its own individual intelligence, but together they create a complex and sophisticated collective intelligence that allows them to accomplish tasks that would be impossible for any individual bee.
The idea of collective intelligence challenges the traditional view of individuality, which places a high value on independence and autonomy. Instead, it suggests that our individual intelligence is only part of a larger whole, and that we are constantly influenced by the thoughts and actions of those around us.
This raises interesting questions about how we define ourselves as individuals. Are we truly independent beings, or are we shaped and influenced by the collective intelligence of the groups we belong to? And how does this collective intelligence affect our decision-making processes and ability to solve problems?
While the concept of collective intelligence may seem like a threat to individuality, it actually has the potential to enhance it. By recognizing the ways in which we are influenced by others, we can become more self-aware and make more informed decisions. We can also learn to collaborate more effectively, drawing on the strengths and insights of others to create more innovative and effective solutions.
In conclusion, “1000 Brains” challenges us to think differently about the nature of intelligence and individuality. While our brains may function as individual units, they are also part of a larger whole that is constantly communicating and collaborating to create a collective intelligence that is greater than the sum of its parts. By embracing this idea, we can enhance our own individuality and become more effective problem-solvers and collaborators.
Second answer from ChatGPT with exactly the same request
In his latest book, “1000 Brains: How We Think Together to Solve Problems,” neuroscientist and AI expert Jeff Hawkins proposes a new theory of intelligence that challenges our traditional assumptions about what it means to be smart. Instead of viewing the brain as a single, unified entity, Hawkins argues that each of us possesses multiple “mini brains” that work in concert to perceive, reason, and act in the world.
According to Hawkins, these “mini brains” are constantly communicating with one another, sharing information and coordinating their activities to solve problems and make decisions. And while each individual brain is unique, with its own set of experiences, memories, and preferences, the collective intelligence that arises from this collaboration is greater than the sum of its parts.
This idea of collective intelligence has been gaining traction in recent years, as researchers explore new ways to harness the power of groups to solve complex problems. From crowdsourcing to swarm robotics to collaborative decision-making, we’re finding that we can achieve remarkable results when we tap into the wisdom of the crowd.
But this raises an important question: if our intelligence is distributed across multiple brains, what does that mean for our sense of individuality? Are we just cogs in a collective machine, or do we still retain some measure of autonomy and self-determination?
Hawkins argues that we can have it both ways. While our brains are constantly working together to achieve shared goals, each of us is still a unique individual, with our own thoughts, feelings, and perspectives. Our collective intelligence doesn’t diminish our individuality; instead, it enhances it by allowing us to tap into the knowledge and experience of others.
This has important implications for how we think about ourselves and our place in society. If we embrace the idea of collective intelligence, we can start to see ourselves as part of a larger whole, with a shared purpose and a common destiny. And if we work together, we can achieve things that no individual could accomplish on their own.
So the next time you’re faced with a difficult problem, remember that you’re not alone. You have a thousand brains at your disposal, all working together to help you find the best solution. And by tapping into the collective intelligence of those around you, you can achieve great things and make a real difference in the world.
Collective Intelligence in Human Organizations
In my work in Eviso, especially in collaboration with Gianfranco Sorasio, I have been exploring for months ideas and frameworks on collective intelligence in human organizations. The main idea is how to design organizations that can understand the fast-changing, interconnected and unstable economies and markets and at the same time provide safe and efficient zones for its employees.
The current best answer looks like it is a Collective Intelligence Design based on teams with high internal interconnection between its members. Looks like, as in the thousand brains model, these Teams are like mini-companies in the sense that they have a self-sufficient, even if partial, description of the context and are able to propose and test solutions.
The speed and richness in which a well-designed and supported collective intelligence organization can tackle complex challenges are much higher than that of a full hierarchical organization. Besides these organizations feel much more human to work in!
A Resume on Collective Behaviors to Create Intelligence
Risking inevitable simplification and re-invention of the wheel, I have made this list of characteristics of intelligence created by collective behaviors and a table with the similarities between the three cases I covered in this Post. I will like to go deeper into this in other days or posts.
- Individual Elements with strong interconnectivity and structure to host a mini-model of reality to be perceived.
- Diversity of Elements is key to the resilience and richness of the intelligence process.
- Voting System to take advantage of the Wisdom of the Crowd
- Constant Learning and Challenges from Reality in every one of the points above.
| Example | Reality to be Perceived | Units | Sub-units |
| Brain | Word and Concepts | Mini-Brains | Neurons |
| ChatGPT | Language and Content | Texts | Words |
| Organizations | Context and Markets | Teams | People |
Extra section: Having Learning to Simulate Humans
Since the first of computers, they have been applied to simulate human societies. In order to better understand how that works I have taken two courses from the Santa Fe Institute in the USA. (they have many free excellent courses on complexity here)
1. Agent-Based Models with Python: An Introduction to Mesa
Interesting but too specific on coding and not on the fundamentals of Simulating Agents.
2. Fundamentals of Netlogo
This one was perfect as it starts from high-level intro on how to Simulate Agents for example humans, and by the end, you are already playing like God with all these agents living in the NetLogo Software.
It’s really amazing how some very simple rules can create complex behaviors. To see test for yourself, playing online with the examples below
A model for neighborhood segregation
In this model, you can play with variables like the percentage of people similar to the individuals living close by that makes them happy. If they are not happy they will look for a better place in the “city”. Small changes in the variables can create very different outcomes. The project is inspired by Thomas Schelling’s Micromotives and Macrobehavior – 1978 – Nobel Prize in 2005. The code is copyrighted by Uri Wilensky 1997
To play with this model online, go to this link in Netlogo and press “setup” and then “go”
A model for Prey and Predators
This model explores the stability of predator-prey ecosystems. Such a system is called unstable if it tends to result in extinction for one or more species involved. In contrast, a system is stable if it tends to maintain itself over time, despite fluctuations in population sizes. Copyright by Uri Wilensky 1997.
To play with this model online, go to this link in Netlogo and press “setup” and then “go”
3 Comments
From this great post, I learned how the brain works (voting system*) and how this can/should be applied to developing intelligent machines. Also, this post explains how this model of brain functioning could apply to human organizations, with teams acting as mini-companies.
Based on the main idea, this post has made me wonder whether having a number of predefined teams in each organization is absolutely necessary for success, and what these predefined teams would be.
*By the way, could each brain have a similar role to a node in a blockchain system? 🙂
Great follow-ups.
Definitely, the idea of a “predefined” or ideal size and number of teams is something that we have been studying in Eviso. That will most likely change with the company goals, market, growth phase etc but it looks some scaling laws are always present…material for a post maybe in the future.
The brain-as-blockchain is challenging. I don’t value and know much the blockchain technology…so I don’t think I can explore and argue on that.
Maybe someone else reading this can answer.
These proves the point that diversity of experiences allows for evolution (and better tackling problems).
When you have a closed context, bias settles and reduces your path of decisions. That is well known in the real world (e.g.: segregated communities tend to interact worse with others, etc) but also on computer world where algorithms are created by people who have their own biases and inequality spills unto it.
That is one of the major concerns with AI, since it’s behavior depends on the information it’s fed and how it’s validated. The current models still use human validators to define the internal voting mechanisms.
Regarding the blockchain remark: it’s slightly different. Blockchain is just a recording mechanism. Each node records a transaction and once recorded it’s immutable (or it should be) and publicly accessible. What you can have on top of the blockchain are smart contracts, where a piece of code triggers some action if certain conditions are met once a record is written on the public chain.
Currently, we haven’t seen the growth of use cases that were expected for the blockchain technology (as I’ve read somewhere, blockchain is an answer for which no one is asking the question). Perhaps it will be one of those technologies that dies without notice but triggers other, more valuable, ones