Article

ai

Category: Expertise strategy

AI and the Ethics Tango

This article explores the shifting relationship between humans and advanced technologies. As machines become more complex, the ethical and risk considerations grow, moving beyond the straightforward machine-user dynamic.

Humans have developed the scope of what modern machines are capable of beyond recognition and, with this, the relationship we have with them. Where early machines were designed to make people’s lives easier and more efficient in some way, the complexity and impact of today’s technologies spill into the realms of ethics, accompanied by other risks that must be addressed. It is no longer a straightforward machine-user relationship.

Here we explore some of these ethical considerations and risks. Let's delve into a few examples of why AI needs humans.

 

Prejudice

Research carried out by the Allen Institute for AI revealed evidence of commercial AI chatbots showing racial prejudice towards speakers of African American English in a dozen versions of large language models. The research suggests that this ‘covert racism’ could, for example, influence AI decisions about an individual’s employability, among other consequences.

Read more about the research here.

Large language models, trained on vast amounts of text data, can inadvertently incorporate societal biases present in that data. This underscores the importance of diverse and carefully curated training datasets. As this technology is developed, it is essential that the people propelling it are constantly reviewing and improving it to eliminate these serious issues.

AI

Equally important are the independent research organisations that are exploring the risks and developing frameworks to navigate AI, such as UNESCO’s Recommendation on the Ethics of Artificial Intelligence. These help to inform legislative bodies as they seek to implement legal parameters and guidelines to manage AI and keep the public safe.

See the UNESCO’s Recommendation here.

 

Legal liability

Earlier this year, Air Canada lost a small claims court case against a grieving passenger, after its AI-powered chatbot stated that the passenger would be able to take advantage of a bereavement discount retrospectively, which was in fact not part of the policy.

The tribunal ruled in favour of the passenger, stating that the claim constituted ‘negligent misrepresentation’. This case serves as a reminder that AI needs constant human supervision.

In the automobile world, driverless cars are already on roads, but the industry is still working through the ethical considerations involved, for example what ‘choice’ an automated car might make in an accident, and the liability question where the car owner is not the driver, as explored in our previous blog.

Read the previous blog here.

The AI landscape is rife with legal and insurance-related dilemmas which need to be identified, explored and adapted to, even whilst legislation struggles to keep up with the pace of change.

Read more about Air Canada lost here.

 

  Large language models, trained on vast amounts of text data, can inadvertently incorporate societal biases present in that data.

 

Continuous improvement

In many uses of AI, particularly in the business to consumer context, the technology is driven by human user interaction and constantly improved and refined in real time. This improvement often involves techniques like reinforcement learning and fine-tuning, where the AI model learns from user interactions and feedback, constantly adapting its responses to better serve users.

AI hallucinations, where AI algorithms generate content that seems plausible but is factually incorrect or nonsensical, often occur due to gaps in training data or misinterpretation of context. These errors highlight the ongoing need for human oversight in AI systems.

Read more about what AI hallucinations are here. 

There are plenty of examples of AI chatbots getting it wrong, sometimes with potentially harmful results, such as the Copilot chatbot telling a user claiming to be a PTSD sufferer that it did not care if they lived or died. Delivery firm DPD also experienced embarrassment when its AI chatbot swore at a customer and criticised the company’s service.

Read more about Copilot chatbot here.

Read more about DPD experienced embarrassment here. 

AI functions cannot be set up and left to run. Human supervision and perspective are an essential part of the puzzle.

The road ahead

There have been many reports and predictions on the prevalence and economic significance of AI in the near future.

For instance, in banking, AI is expected to revolutionise risk assessment through advanced machine learning algorithms that can process vast amounts of financial data in real-time, potentially leading to more accurate credit decisions and fraud detection.

McKinsey’s ‘The economic potential of generative AI’ predicts that banking, high tech and life sciences will see the greatest impact on revenues as a result of AI, whilst in retail and consumer packaged goods, the potential impact could equate to up to $660 billion per year.

The secret to successfully harnessing AI and producing these results will lie in handling the relationship between humans and AI and understanding the vital role that people play. Whilst AI can process vast amounts of data in seconds, humans provide the context, creativity and the moral compass.

Read more about McKinsey’s ‘The economic potential of generative AI’ here.

 

  AI hallucinations, where AI algorithms generate content that seems plausible but is factually incorrect or nonsensical, often occur due to gaps in training data or misinterpretation of context.

 

The AI Paradox

emagine’s experts work with hundreds of clients to plan, build and scale their AI projects, underpinned by essential AI training.

Our experts also step back to reflect on the bigger picture. In a new e-book, The AI Paradox: Dancing with the Machines, emagine’s Head of AI Solutions, Mariusz Misiek, explores AI use cases across industries, emphasising the balance between AI capabilities and human expertise. It offers insights on leveraging AI while maintaining the human element in business processes and decision-making.

The e-book is available here.

Ready to find out more?

Ask us how we can help you succeed.

Contact us

Blog

Explore our blog

Read more

left-arrow
right-arrow

Henrik Timm
Projects & Implementation
Succeed as a consultant

The project manager in the era of AI

The advent of artificial intelligence (AI) has sparked discussions across various industries about its potential impact on traditional roles and responsibilities. The field of project management is not exempt from these considerations, and as a project manager, it is only natural to ask yourself: How will AI affect my work? 
Project manager at a meeting with his team.
Projects & Implementation
Succeed as a consultant

Key considerations for aspiring project managers

This article serves as a guide to all aspiring project managers, discussing important considerations and essential skills needed to excel in this dynamic career.
Projects & Implementation
Succeed as a consultant

From risks to requirements: Project initiation from A to Z

As the first step in the project lifecycle, project initiation is critical in determining the ultimate success of a project. Learn the best practices and key considerations for successful project initiation.