AI: The New Frontier of Human-Like Technology

Artificial Intelligence (AI) is rapidly evolving, becoming increasingly sophisticated and capable of performing tasks that were once considered the exclusive domain of humans, such as making crucial decisions. This raises many questions, particularly around responsibility and accountability as AI systems make their own decisions.

Hosted by AI.SE, and led by Patrick Couch, AI expert at HPE led a discussion featuring Christina Ramm-Ericsson, Chief Economist and Head of Business Policy at TechSverige, Viktor Rosenqvist, Business Developer at PostNord, and a representative from Futurion, a think tank, discuss the challenges and opportunities of AI.

AI in Action: Decisions with Consequences

Patrick started the panel discussion by presenting three cases of AI systems that made decisions, previously made by humans, which turned into conversations about transparency and responsibility:  

• Self-driving cars: In San Francisco, a self-driving vehicle from Cruise, a subsidiary of General Motors, was involved in an accident. Who is responsible for the crash? Cruise, General Motors, or the individual passengers?   
• Apple Pay Card: Apple's credit card uses an algorithm to assist with credit scores. This AI-powered tool reportedly provided better credit recommendations for men than for women. This raised concerns about fairness and discrimination. Apple did not know how the algorithm had made the decisions nor why.
• Air Canada's Chatbot: Air Canada's AI chatbot offered discounted fares to customers. The chatbot was - by Air Canada - initially considered a legal entity but was later deemed not to have legal standing.

Lowering the Barrier to Entry: AI Becomes More Human-Like

AI models are becoming increasingly sophisticated, making them easier to adopt and integrate into various applications. This progress is evident in two key areas:

• Human-like interactions: AI systems are becoming more adept at interacting with humans naturally and intuitively.
• Creativity: AI demonstrates the ability to perform creative tasks, such as writing and generating art.

Transparency and Explainability: Demystifying AI

AI models are becoming more transparent, allowing users to understand their decision-making processes. This shift is crucial for building trust and ensuring that AI is used responsibly.

Viktor Rosenqvist of PostNord highlights the importance of explainability:

"Just a few years ago, AI models were largely black boxes. Now, we can explain how these models work in plain language. This gives us more control over the technology. No more black boxes.”

The Scalability Challenge: Maintaining Control

As AI models become more powerful, it is essential to keep control over their operations. This requires clear governance and accountability mechanisms.

Viktor Rosenqvist emphasizes the need for control:

"The threshold for scaling AI is control. We cannot accept that things just happen. Businesses must demand to know how AI works."

PostNord's Approach to AI Governance

PostNord has implemented a governance framework to ensure responsible AI use. This framework includes:

• Defining clear roles and responsibilities: Identifying who is accountable for the AI model's decisions and actions.
• Validating the model: Ensuring that the model is accurate and reliable before deployment.
• Regular monitoring and review: Continuously evaluating the model's performance and addressing any issues that arise.

Sweden's Position in the Global AI Landscape

Despite its reputation for innovation, Sweden lags in AI adoption. According to the Global AI Index, Sweden ranks 17th, indicating significant room for improvement.

The AI Act: Mitigating Risks and Ensuring Responsible Development

The AI Act, a new regulation from the European Union, aims to mitigate the risks associated with AI and ensure its use aligns with European values.

Panelists' Perspectives on the AI Act

The panelists expressed their support for the AI Act and recognized the need for clear guidelines and regulations to govern AI development and use.

Patrick Couch highlights the importance of the AI Act:

"The only way to steer development is to steer the direction. However, it is difficult to regulate something difficult to define. What is AI -really?"

Christina Ramm emphasizes the need for concrete measures:

"We need to make the AI Act measurable, concrete, and follow-up."

Conclusion: Embracing AI Responsibly

The rapid advancement of AI presents both challenges and opportunities. By adopting a responsible and transparent approach, we can harness the power of AI to drive innovation and improve our lives while mitigating potential risks. The AI Act provides a valuable framework for achieving this goal.

Want to learn more about the AI Act? Check out our events in the fall: ideon.se/events

READ MORE

About The AI Act

The AI Act aims to strike a balance between encouraging innovation in AI and protecting citizens from potential harm.

The Act aims to mitigate the risks of AI and ensure its use aligns with European values. Here are some of its key points:

• Classifies AI systems based on their risk level (from minimal to high risk).
• Sets out requirements for high-risk AI systems, such as risk assessments, data governance, and transparency.
• Prohibits certain high-risk uses of AI, such as social scoring systems that unfairly discriminate against individuals.
• Promotes responsible development of AI that is fair, safe, and unbiased.

Read more about Sweden as number 17 on the Global AI Index and Ai.se’s suggested strategy and approach:

https://www.vinnova.se/nyheter/2024/052/experter-varnar-sverige-behover-en-tydlig-riktning-i-ai-utvecklingen-for-att-inte-halka-efter/

https://strategy.ai.se/