Life Tips Miscellaneous Travel & History Zen and Life Tips Read Classics Philosophy Navigation of this blog
Artificial intelligence technology from the Tao
Lao Tzu, also mentioned in ‘Living like water – the path at the root of Lao Tzu’s thought’, was an ancient (6th century BC) Chinese philosopher and is regarded as the founder of the ‘Taoist’ school of thought. Central to Lao-Tzu’s thought is the concept of ‘Dao’ (道), which is regarded as the fundamental principle of the universe and the basis of all things. Although the Tao is invisible and cannot be fully expressed in words or concepts, Lao Tzu believed that by pursuing the Tao, people could live according to the natural order.
The Tao’s philosophy emphasises the generation of natural data and algorithms that transcend ‘inaction and nature’ human activity and produce complex results with few elements, change and adapt according to the situation, as well as valuing the harmony of the whole system with algorithms that are less wasteful, simpler and more aware of the balance and coordination of the whole system The Tao philosophy is a key element of the Tao system.
In this article, we consider artificial intelligence (AI) technology from this Tao philosophy. Considering AI technology from the Tao’s philosophy is because it may provide unprecedented inspiration for the role of AI and its design principles. As Tao emphasises ‘natural flow’ and ‘harmony’, and ideally adapts to the environment and circumstances without strain, the following perspectives are also important in the way AI should be designed.
1. ‘no action, no nature’ AI design: applying Tao’s ‘no action, no nature’ philosophy to AI, the ideal system would be one in which artificial intelligence does not rely on ‘artificial instructions’ or ‘forced goal attainment’, but rather produces optimal results naturally. For example, the aim is for ‘spontaneous AI’, where AI adapts itself to the environment and user needs via data and provides support and suggestions naturally; specifically, a system where AI self-adjusts and provides optimal support based on the user’s behaviour history and situation.
2. simple and harmonious algorithms: tao should respect simple and lean structures, and AI algorithms should not be over-complicated, but should be designed to be natural and harmonious. This leads to an approach where AI learns efficiently while using minimal computational resources, depending on the nature of the task and data. Technologies that maximise effectiveness with minimal structure, such as sparse modelling and lightweight neural networks, are in line with Tao’s philosophy.
3. self-organising and self-adaptive: by incorporating the ‘flow’ and ‘natural adaptation’ of Tao’s philosophy into AI, self-organising and self-adaptive AI systems can be built. This is a system where AI responds to constantly changing data and environment and optimises its structure and knowledge autonomously, e.g. ‘evolutionary neural networks’ and ‘dynamic network structures’, where the neural network architecture changes dynamically in response to data. This is expected to allow AI to function with Tao-like fluidity rather than a fixed model.
4. harmony-oriented multi-agent systems: Tao emphasises the concept of harmony, and in AI systems it is ideal for different agents to interact with each other and achieve harmony as a whole. Based on this, co-operative multi-agent systems (MAS) can be effective in AI technology, with each agent having an individual role, but with a design that emphasises interaction and co-operation so that the overall result is naturally harmonious. This will create a system where AI can respond to diverse tasks in a natural flow. For more details, see e.g. ‘Overview and implementation examples of multi-agent systems using graph neural networks’.
5. natural decision-making and intuitive learning: Tao’s philosophy includes ‘natural decision-making without overthinking’, and in AI decision-making, it is ideal to avoid excessive computation and complex reasoning, and to learn and make decisions intuitively based only on necessary information, and reservoir computing and lightweight The use of reservoir computing and lightweight neural networks can achieve useful pattern recognition and decision-making with simple computation, and AI with a Tao-like naturalness can be aimed for. See also ‘About reservoir computing’ for more information.
6. designing AI in harmony with ethics: As the Tao philosophy values harmony between humans, nature and society, AI could also be considered ethical and in harmony with human society. For example, ‘Compassionate AI’, in which AI understands human intentions and emotions and supports them in their situation, and ‘Empathetic AI’, in which AI effortlessly follows the intentions of the user, are being considered, as well as maintaining transparency in the data and algorithms used by AI and building natural trust relationships. For more information, see also ‘Towards compasionate and empathetic AI’.
7. lean data processing: Tao’s philosophy of ‘less is more’ requires a lean approach to AI data processing. Even when dealing with vast amounts of data, learning based on the minimum necessary information and building simple, efficient models is in line with Tao’s thinking. These machine learning approaches include compression learning, data sampling or data sorting to optimise information processing.
8. aim- and outcome-transcending AI: Tao emphasises surrendering to the natural flow of things, rather than being bound by objectives and outcomes, and in designing AI, mechanisms can be considered that adapt and flexibly change objectives according to the situation and environment, rather than only seeking to achieve a fixed goal. Specifically, this includes designing self-learning agents that are not goal-oriented, but allow for situation-adaptive decision-making, and AI that can flexibly set goals according to the environment.
AI technologies based on these Tao philosophies will aim for systems that are autonomous, harmonious, adaptive and flexible, and are consistent with the basic ideas for achieving strong AI as described in ‘Free Will, AI Technology and Zhuangzi Freedom’ and ‘Strong AI and Biomimicry’. AI with a ‘natural flow’, such as Tao, has the potential to build more humane and harmonious systems, with an emphasis on symbiosis with users and the environment.
Tao and GNN.
Let us now apply Tao’s ideas to the next generation of deep learning technology, the graph neural network (GNN). By applying Tao’s ‘Let Nature Take Its Course’ or ‘Inaction-Nature’ to GNNs, it is expected that it will be possible to design algorithms in which the network structure changes dynamically and knowledge propagation occurs in harmony with nature, aiming for a flexible and adaptive network model. Those specific approaches are described below.
1. self-organised interactions between nodes: Tao emphasises natural generation and the emergence of order out of disorder. From this perspective, it is conceivable to construct flexible, self-organising networks in GNNs, where the relationships and connection structures between nodes are not fixed, for example, by introducing algorithms where nodes dynamically create and delete links based on similarities and characteristics and adapt to an optimal network structure and natural information flow and propagation. This enables dynamic and tau-like networks that go beyond fixed graph structures.
2. information propagation with minimum energy: Tao emphasises waste, simplicity and harmony; in GNNs, too, mechanisms for efficient information propagation with minimum energy are ideal, e.g. by utilising ‘sparse propagation’, where only important nodes and edges are active and other nodes play a minimal role. Propagation’ can be utilised to reduce waste in information transfer. Sparse message passing and data importance-based sampling methods can be used to build energy-efficient GNNs.
3. adaptive and flexible network structure: the Tao philosophy includes flexible adaptation in response to the external environment; even in GNNs, the network can be flexibly reorganised in response to changes in the environment and data to naturally incorporate dynamic data and complex relationships, for example, dynamic Graph Neural Network (D-GNN) can be used to realise a dynamic network structure in which the state of each node and the relationships between them change over time. This enables GNNs with Tao ‘adaptability’ and ‘flexibility’.
4. information transfer with overall harmony: the Tao emphasises overall harmony and balance: in GNNs, the ideal state is one in which the entire network is in harmony, and the influence and update methods of each node and edge are devised so that information propagation between nodes proceeds in overall harmony. For example, when information at each node is updated, the influence of neighbouring nodes is taken into account and balanced information propagation is carried out so that overall harmony is maintained, and such a design helps to achieve both a stable learning process and high expressive power.
5. introducing a natural layer structure: from Tao’s perspective, the layer structure of a GNN should ideally also follow a natural distribution and naturally expand and converge to the extent necessary. For example, one can consider a layered structure of the network in which the stronger the relationships between nodes, the closer they are processed in the layers, and the weaker relationships are naturally moved away. This allows for flexible and adaptive overlapping of layers in accordance with the Tao-like flow, and prevents waste in the information processing of the network as a whole.
6. designing a target function that is aware of the ‘path’ as a whole: in order to reflect the wholeness of Tao in the learning goals of the GNN, it is conceivable to introduce a target function that takes into account not the individual losses of each node, but the harmony and balance of information propagation in the network as a whole. This would emphasise overall optimisation rather than local accuracy improvement, and allow the entire GNN to be designed to function in a natural way.
implementation example
Several examples of AI system implementations incorporating the Tao philosophy are given. The focus here is on code examples that avoid complex algorithms and huge computational resources in favour of a simple and efficient approach to designing for natural adaptation and harmony.
1. dynamically adaptive natural language processing (NLP) models
Implement a mechanism whereby the NLP model naturally adapts to the user’s intentions during the interaction with the user. This example shows a system that dynamically adjusts its response according to the user’s context.
import random
class TaoChatBot:
def __init__(self):
self.context = []
def update_context(self, user_message):
# Add user messages and keep the context up-to-date
self.context.append(user_message)
if len(self.context) > 5: # Holds up to five messages
self.context.pop(0)
def generate_response(self, user_message):
# Generate simple context-sensitive responses based on ‘flow’
self.update_context(user_message)
# Return simple contextual responses while avoiding over-calculation
response_options = [
"I see what you mean.",
"Can you tell us a bit more about that?",
"Understood. Is there anything else you would like to tell us?",
"Very interesting. Please continue.",
"Think about it in depth."
]
return random.choice(response_options)
# examples showing the use (of a word)
tao_bot = TaoChatBot()
print(tao_bot.generate_response("I want to talk about the application of AI technology."))
print(tao_bot.generate_response("I would like to know more about the philosophy of Tao."))2. self-adaptive recommendation systems
Based on Tao’s ‘no action, no nature’ approach, we design a recommendation system that naturally adapts to the user’s preferences. Here, the items selected by the user are learnt by a lightweight feedback loop and reflected in the next recommendation.
import random
class TaoRecommender:
def __init__(self):
self.user_preferences = {}
def update_preferences(self, item):
# Simple update of user preferences based on item categories.
if item in self.user_preferences:
self.user_preferences[item] += 1
else:
self.user_preferences[item] = 1
def recommend(self):
# Simple recommendations based on natural flows and taking user preferences into account.
if self.user_preferences:
top_preference = max(self.user_preferences, key=self.user_preferences.get)
recommendation = f"{top_preference}’ is recommended for you."
else:
recommendation = "The current recommendation is for ‘random items’."
return recommendation
# examples showing the use (of a word)
tao_recommender = TaoRecommender()
tao_recommender.update_preferences("book")
print(tao_recommender.recommend()) # Initial Recommendation
tao_recommender.update_preferences("film")
print(tao_recommender.recommend()) # Post-update recommendations.3. agent systems with natural interaction
A simple implementation of a multi-agent system in which agents co-operate with each other and act in harmony is presented. The agents do not have complex rules or action plans, but act in cooperation with other agents according to the situation.
class TaoAgent:
def __init__(self, name):
self.name = name
self.energy = 100 # Energy level of the agent.
def cooperate(self, other_agent):
# Share energy and harmonise with other agents
if self.energy > 50:
shared_energy = 10
self.energy -= shared_energy
other_agent.energy += shared_energy
print(f"{self.name} shared energy with {other_agent.name}.")
def act(self):
# Natural energy regulation for effortless action.
if self.energy < 50:
print(f"{self.name} is resting due to low energy.")
self.energy += 5
else:
print(f"{self.name} acts in harmony.")
self.energy -= 5
# examples showing the use (of a word)
agent_a = TaoAgent("Agent A")
agent_b = TaoAgent("Agent B")
# Co-ordination and flow of action
agent_a.act()
agent_a.cooperate(agent_b)
agent_b.act()
4. simple self-organising networks
Create networks that adapt to the complexity of the data and task and are self-organising with minimal structure. In this example, a simple graph structure that changes the weights of edges in a self-adjusting manner is used to naturally adjust the state of the network.
import networkx as nx
class SelfOrganizingNetwork:
def __init__(self):
self.graph = nx.Graph()
def add_node(self, node):
self.graph.add_node(node)
def add_edge(self, node1, node2, weight=1):
self.graph.add_edge(node1, node2, weight=weight)
def adjust_weights(self):
# Harmonious weighting and balancing of the network
for (u, v, d) in self.graph.edges(data=True):
d['weight'] += 1 # Simply increase weights and adjust adaptations
def display_network(self):
# Displays information on nodes and edges
print("Network nodes:", self.graph.nodes())
print("Network edges with weights:", self.graph.edges(data=True))
# examples showing the use (of a word)
network = SelfOrganizingNetwork()
network.add_node("A")
network.add_node("B")
network.add_edge("A", "B", weight=2)
network.adjust_weights()
network.display_network()reference book
Reference books for a deeper understanding of the relationship between Tao and artificial intelligence include the following.
2. “The Book of Tao and Te: A New Translation”
3. “The Alignment Problem: Machine Learning and Human Values”
5.”Complexity: The Emerging Science at the Edge of Order and Chaos“
6.”Graph Neural Networks: Foundations, Frontiers, and Applications“
7. “Deep Learning“
9. “MIND“
10. “Flourish: A Visionary New Understanding of Happiness and Well-being“
コメント