Researchers from Loughborough University, MIT and Yale have introduced the concept of “collective AI.”
The researchers propose to document their ideas in a perspective paper published in Nature Machine Intelligence Shared Experience Lifelong Learning (ShELL) as a framework for creating decentralized AI systems consisting of multiple independent agents, or “collective AI”.
These individual AI units operate like a “hive mind,” continually learning and sharing knowledge throughout their lives, posing a challenge to centralized monolithic architectures.
If collective AI is developed, it could mirror the capabilities of “The Borg” from Star Trek and various other science fiction concepts similar to “The Get” from Mass Effect or “The Replicators” from Stargate SG-1.
By enabling agents to learn from their very own experiences and the knowledge shared by others, ShELL systems can learn faster, exhibit improved performance and greater flexibility within the face of adversity – just like biological organisms.
Top computer scientists from all over the world say the long run of artificial intelligence is analogous to Star Trek🖖🤖
Read: https://t.co/JgKhqvUDmG pic.twitter.com/kd1QAyGxGB
Dr. Andrea Soltoggio from Loughborough University, the study's lead researcher, described the vision of the study: “Instant knowledge sharing across a collective network of AI entities able to constantly learning and adapting to latest data will enable rapid responses to latest situations, challenges or threats.”
Soltoggio also highlighted the potential of decentralized AI by drawing an analogy to the human immune system, where multiple components work together to mount a coordinated defense against threats.
“It could also result in the event of disaster response robots that may quickly adapt to operational conditions, or personalized medical agents that improve health outcomes by combining cutting-edge medical knowledge with patient-specific information,” Soltoggio explained.
The study mentions several potential practical applications:
- Space exploration: ShELL's decentralized learning and adaptation capabilities could possibly be precious in space missions where communication with Earth is proscribed and autonomous systems must overcome unexpected challenges.
- Personalized medicine: ShELL could power distributed medical AI systems that constantly adapt to patients' changing needs and medical knowledge, enabling more targeted and effective healthcare.
- Internet security: The collective learning and knowledge sharing of ShELL agents could possibly be used to create decentralized defense systems that quickly detect and disseminate details about latest threats, enabling faster and more robust responses to cyberattacks.
- Disaster relief: The paper proposes that ShELL systems could possibly be used to coordinate autonomous agents in disaster scenarios, enabling more efficient and effective response efforts by leveraging the group's collective intelligence.
- Multi-agent detection: ShELL could enable the coordination of swarms of agents to create 3D world models for tasks similar to search and rescue missions or anomaly detection in military reconnaissance.
Here's a possible idea for constructing a democracy with plenty of AI, diversity and resilience versus a handful of huge entities controlling probably the most powerful technology of the twenty first century: https://t.co/Eg7R9X4WEp Are there any takers?
Interest in decentralized AI itself is growing, as evidenced by the recent resignation of Emad Mostaque, CEO of Stability AI
Systems grow to be significantly more resilient once they work together each collectively and independently, which we see in natural systems similar to schools of fish and the coordinated movements of birds and insects.
There has been some research into decentralized AI prior to now. For example, a startup Sakana recently founded by former Google engineers raised $30 million for “swarm” AI This is conceptually just like what’s proposed on this latest study.
Building collective AI
How could collective AI work? Researchers suggest several possible mechanisms:
- Lifelong machine learning: Allows AI agents to step by step learn multiple tasks without affected by catastrophic forgetting. Techniques include replay methods (storing and replaying previous experiences), regularization (limiting model updates to forestall overwriting old knowledge), and parameter isolation (assigning separate model components for various tasks).
- Federated learning: A distributed learning paradigm by which multiple agents train a model together while keeping their data localized. Each agent calculates model updates based on its local data and only shares those updates with others to take care of privacy.
- Multi-agent systems: Study of autonomous agents interacting in a shared environment. ShELL agents work decentrally and make decisions based on their individual goals and knowledge.
- Edge computing: Perform calculations and store data near the info sources, e.g. B. on devices or edge servers, fairly than in centralized cloud systems. ShELL agents work on edge devices, enabling low latency processing and reducing communication costs.
Researchers are also aware of the potential risks of collective AI systems, similar to the rapid spread of incorrect, unsafe or unethical knowledge between entities.
To address this, they propose encouraging the autonomy of every AI unit inside the collective and ensuring a balance between collaboration and independence.
Collective AI builds on recent futuristic developments in AI, similar to: bio-inspired AI Architectures that effectively simulate analog synaptic structures and AI models running on them real brain cells.
