Collaborative Intelligence
Research work on the notion of Human - AI Collaborative Intelligence.
Human - AI Collaborative Intelligence Basics
Collaborative Intelligence (CI) is a human-machine partnership that combines the strengths of both humans and AI to achieve better results than either could on their own. CI systems are designed to be co-creative, meaning humans, and AI work together to solve problems and make decisions. On the other hand, Industry 5.0 is a new industrial revolution focused on human-centric manufacturing. In this new era, AI can augment human capabilities rather than replace them. CI is essential for Industry 5.0 because it allows humans and AI to work together effectively.
Human - AI Interaction Paradigms
In classic AI systems, the human operator (often a data scientist) is involved in targeted steps of the AI lifecycle to perform specific tasks:
- Data labelling: People often label data for ML models. This is especially important for tasks that are difficult or time-consuming for machines, such as identifying objects in images or extracting text from documents.
- Model training: People can help to train ML models. This can be done by providing feedback to the model during training or by manually adjusting the model's parameters.
- Model deployment: People can also deploy ML models in the real world. This can involve monitoring the model's performance and adjusting as needed.
The above can be considered a basic paradigm of CI. But CI for Industry 5.0 goes beyond data labelling and model deployment, fostering actual human -AI collaboration in hybrid teams, to achieve:
- Machine learning with human feedback: This is a type of CI where humans provide feedback to AI systems to help them learn and improve. For example, humans might be asked to label data or correct errors in the AI system's output. The clearest example of this category is the HITL solution.
- Co-creation: This type of collaborative intelligence is where humans and AI systems work together to create new products or services. For example, humans might provide AI systems with their ideas and feedback, while AI systems provide humans with access to new data and insights. The clearest example of this category would be the emerging Large Language Models (LLMs)
- Augmented intelligence: This type of CI uses AI systems to augment human capabilities. For example, AI systems might provide real-time assistance to doctors, help pilots fly planes or any other related activity.
Advantages of the Human-in-the-Loop and CI Approaches
Several advantages are associated with successfully integrating the CI paradigm in current practices, such as:
- Improved Accuracy: Human expertise can help correct AI mistakes and enhancing overall system performance.
- Ethical Decision-making: Humans can provide context, empathy, and ethical judgment in complex situations where AI alone may struggle.
- Adaptability and Flexibility: Human in the Loop (HITL) allows AI systems to learn and adapt quickly, incorporating new information and addressing unforeseen scenarios.
- Data Quality Assurance: Humans can validate and curate datasets, ensuring high-quality inputs for AI training and avoiding biases.
Requirements of CI Systems
Effective collaboration in hybrid human - AI teams relies on clear communication and mutual comprehension of goals and expectations but also on other high level requirements:
| Requirement | Description |
|---|---|
| Need for trust | Humans and AI must be able to trust each other to work together effectively |
| Need for data | CI systems require a lot of data to learn and improve. In some cases, we are talking about very gigantic amounts of data needed to train the systems, which is not always possible |
| Need for ethical guidelines | CI systems must be developed and used ethically and in accordance with the laws |
| Need for continuous communication | Human operators must be able to interpret and understand the outputs and decisions made by AI systems, while AI systems should be able to accurately comprehend human intentions and context |
Challenges of Collaborative Intelligence
There are still some important challenges to overcome in the context of a CI system and in hybrid human-AI teaming environments:
- Cost and Scalability: Integrating humans can increase operational expenses, and scalability may become challenging.
- Training and Expertise: Ensuring human contributors have the necessary training and expertise to provide accurate input to AI systems.
- Workflow Integration: Establishing seamless workflows between humans and AI systems, minimising delays and bottlenecks.
- Balancing Human-AI Interaction: Striking the right balance between human involvement and AI automation, optimising system performance.
Tools and Solutions
In the following Table, we look at concrete examples of software solutions that facilitate collaboration between operators and machines in the manufacturing context.
| Tool Name | Purpose | Description | Link |
|---|---|---|---|
| Orchestra | Real-time data exchange and integration between different systems | Orchestra is intended to serve as a bridge between siloed systems, enabling a collaborative environment where data flows without restrictions. This orchestration is intended to enhance productivity and reduces manual intervention, minimising errors and simplifying processes.
Using a HITL system like Orchestra will improve project the management processes, resulting in increased accuracy and consistency, less time coordinating projects, and more expert time in the areas they are uniquely positioned to work on. In addition, Orchestra's flexibility extends beyond data integration. It also offers various customisable automation capabilities, allowing human operators to streamline complex workflows and decision-making processes. In this context, this solution can intelligently route tasks, prioritise workloads, and adapt to changing conditions in real time. This adaptability enhances efficiency and empowers organisations to stay agile and responsive in fast-paced landscapes. In summary, whether orchestrating data or tasks, this solution offers a new level of efficiency and collaboration |
https://www.b12.io/orchestra/ |
| Teaming AI | Platform for human-AI collaboration in Industry 4.0 | Teaming.AI is a cutting-edge platform designed to foster collaborative interactions between human stakeholders and artificial intelligence systems within the context of Industry 4.0 [7]. The primary objective of Teaming.AI is to address the limitations of current Industry 4.0 practices, particularly the lack of flexibility, by establishing a human-centred AI collaboration model. This approach seeks to ensure that humans remain in control and autonomous while leveraging the capabilities of AI to achieve more efficient and effective industrial processes.
The platform encompasses a comprehensive suite of software components that facilitate seamless interactions between the Teaming.AI system and all human-AI team members. At its core, the platform is tailored to cater to specific industrial facilities, referred to as the target system, which hosts various production processes, such as manufacturing and quality control of products |
https://www.teamingai-project.eu/ |
Other Info
- Research by Software Competence Centre Hagenberg - SCCH