AI or Just Hype? The Growing Problem of AI Washing

Artificial intelligence (AI), a term coined by John McCarthy in 1956, evokes different ideas for different people. Some might recall early cultural references like the "heartless" Tin Man from The Wizard of Oz or the humanoids in Star Wars. Today, science fiction shows clear signs that we are catching up with the imaginations of past authors. We have entered the post-modern AI era, and society, especially businesses, is eager to integrate AI into the tools we use in our work and daily lives. By adding “AI” as a tool for business, many companies have promoted themselves as “AI companies” or makers of “AI tools.” While these tools have been expanding and multiplying, not all companies are clear about what “AI” actually means. This is especially true in a world where “AI washing” exists—the practice of promoting a tool or company as having AI capabilities when it does not.

AI is a broad field with many different terms and subtle differences between technologies. As an AI consultancy company, we are filled with experts who are well-informed and actively contributing to the field of AI. Who better to guide you through understanding the various versions of AI?

‍

The New Era of Artificial Intelligence

Professor John McCarthy, who coined the term "artificial intelligence," defined it as “the science and engineering of making intelligent machines.” In this context, intelligence can be understood as the ability to learn and apply appropriate techniques to solve problems and achieve goals in an uncertain, ever-changing world.

However, when we consider the common association many people make between robots—particularly those that are pre-programmed—and AI, we start to see a distinction. These robots, despite their flexibility, accuracy, and consistency, may not truly qualify as modern AI. They follow programmed rules and execute tasks with precision, but they lack the ability to learn or adapt autonomously. In other words, they lack the intelligence that modern AI implies. While robots may be advanced, they often fall short of the dynamic, learning-driven capabilities that characterise real artificial intelligence.

‍

Automation versus AI  

Let’s clarify this using the EU's definition of AI. According to the European Commission, artificial intelligence is defined as:

“Artificial intelligence (AI) refers to systems designed by humans that, given a complex goal, act in the physical or digital world by perceiving their environment, interpreting the collected structured or unstructured data, reasoning on the knowledge derived from this data and deciding the best action(s) to take (according to predefined parameters) to achieve the given goal. AI systems can also be designed to learn to adapt their behaviour by analysing how the environment is affected by their previous actions.”

As this definition clarifies, AI systems interact (to some extent) with the environment with a degree of autonomy. This interaction can be broken down into a few key capabilities:

1. Perceiving the Environment: AI systems gather data from their surroundings, whether through sensors, cameras, or other means, allowing them to understand the context in which they operate.
2. Reasoning on What is Perceived: After perceiving the environment, AI systems process and analyse the information to make sense of it. This involves interpreting data, identifying patterns, and drawing conclusions based on the inputs.
3. Deciding the Best Action: Based on the reasoning process, the AI system determines the most appropriate course of action to achieve its goals. This decision-making process is influenced by the system's algorithms, past experiences, and predefined objectives.
4. Acting Accordingly: Finally, the AI system takes action based on its decision, whether it's controlling a robot's movements, generating a response in a chatbot, or optimising a process in real-time.

These capabilities distinguish sophisticated AI systems from simpler automated tools or pre-programmed machines. Sophisticated AI is not just about executing commands—it is about understanding, adapting, and making informed decisions autonomously.

‍

AI System Examples

NLP AI system: An AI-based virtual assistant that can help users with scheduling appointments based on spoken requests.

1. Perceive the Environment: The AI listens to the user's voice input, e.g., "Schedule a meeting with Dr. Smith next Tuesday at 10 AM."
2. Reasoning on What is Perceived: The AI processes the request using NLP, extracting details like "Dr. Smith," "next Tuesday," and "10 AM."
3. Deciding the Best Action: The AI checks the user's calendar for conflicts and finds an available slot..
4. Acting Accordingly: The AI schedules the meeting, confirms with the user, and sends an invitation to Dr. Smith.

CV AI system: An AI-based system for self-driving cars that acts autonomously to navigate safely.

1. Perceive the Environment: The AI uses a camera to capture an image of a scene, such as a self-driving car's view of the road ahead.
2. Reasoning on What is Perceived: The AI processes the image using computer vision techniques, identifying objects like pedestrians, other vehicles, and traffic signs.
3. Deciding the Best Action: The AI evaluates the positions and movements of these objects to determine the safest path or necessary actions, such as stopping at a red light or avoiding a pedestrian.
4. Acting Accordingly: The AI controls the car's steering, acceleration, and braking to follow the planned path or take corrective actions.

As these two examples demonstrate, the systems presented above are AI as they interact with the environment in an autonomous manner.

Considering we know what is AI, let's look at what is not (according to EU AI definition above).

‍

Non-AI System Examples

Using a negation of the EU definition, non-AI systems, in the modern sense, are non-intelligent and non-autonomous. Below are two examples of non-AI systems:

NLP non-AI system: A virtual assistant that helps users with scheduling appointments based on predefined voice commands.

1. Perceive the Environment: A rule-based virtual assistant responds to predefined voice commands, such as "What time is it?"
2. Reasoning: The system matches the spoken phrase to a pre-programmed response without understanding or processing the context.
3. Deciding the Best Action: It does not make decisions based on context but follows a fixed set of rules.
4. Acting Accordingly: The assistant responds with the current time but cannot handle more complex requests.

Computer vision non-AI system: A system for self-driving cars that processes images.

1. Perceive the Environment: A basic image processing system captures an image and performs tasks like edge detection.
2. Reasoning: The system applies static algorithms without understanding the objects in the image.
3. Deciding the Best Action: There is no decision-making process.
4. Acting Accordingly: The system outputs a processed image without interpreting or reacting to the content.

Comparing both AI and non-AI systems, we can understand that the first uses techniques like machine learning to analyse data, learn patterns, and make decisions based on context, with the ability to improve over time. In contrast, the latter use rule-based systems and operate within a narrow scope defined by pre-programmed rules, lacking the capacity for learning or adapting to new information.

‍

AI in Everyday Home Appliances

Many of you may have noticed that home appliances are becoming increasingly sophisticated, with one prominent example being the so-called “AI washing machine.” Some brands advertise AI features that claim to optimise wash cycles and detergent usage. While we cannot be sure of the real technology used by this type of product, they seem more likely to use advanced rule-based systems and some machine learning algorithms rather than sophisticated AI. These appliances do not learn from usage patterns or adapt dynamically beyond basic adjustments.

This poses the theoretical question: “Where is the line that divides AI from Machine-learning and rule-based algorithms?“

To date, the answer to this question remains elusive. However, if we consider AI as existing on a continuum spectrum, it becomes clearer that what was once considered sophisticated AI (e.g., optical character recognition) no longer qualifies as such, despite using textbook AI techniques. AI today is the most primitive it will ever be, suggesting we are in a period of rapid development where technology quickly becomes obsolete or commonplace.

To illustrate this, consider the AI washing machine. At first glance, it may seem quite “intelligent,” but upon closer inspection, it employs relatively common technology compared to advanced techniques like deep learning or large language models. On the other hand, imagine a fridge that analyses its contents and suggests recipes based on the ingredients available. Now that would be a game-changer! Samsung, are you up for the challenge?

‍

The Rise of AI Washing

Now that we now know what AI is and is not, let's explore the concept of AI washing and why it has become prevalent.

AI washing occurs when companies exaggerate their use of AI by misrepresenting rule-based systems as AI, overstating capabilities (like surveillance systems that still require human oversight), or failing to achieve the promised full automation.

One reason for this behaviour is the pressure companies feel to appear cutting-edge by showcasing AI capabilities. As Sri Ayangar from OpenOcean told the BBC:

"Some founders seem to believe that if they don’t mention AI in their pitch, this may put them at a disadvantage, regardless of the role it plays in their solution," OpenOcean’s Sri Ayangar told BBC News.”

Another reason is that achieving fully operational AI systems is still extremely challenging. The dream of artificial general intelligence (AGI) may be closer than before, but developing robust AI systems capable of handling variations such as data quality, weather conditions, hardware issues, or handwritten text remains highly complex. This means human oversight is often still required, as shown by Amazon's "just walk out" technology.

Moreover, AI-washing is driven by the market hype surrounding AI, allowing companies to make vague claims without being questioned. Financial incentives also play a role, as AI-branded products can command higher prices and attract more investment, leading to inflated revenues. Without clear regulation, companies exploit the ambiguity of AI, misleading those unaware of the technical details.

‍

How to Avoid AI Washing?

Deciding whether to trust a company's AI claims can be difficult, but clear communication is key. When someone truly understands a concept, they should be able to explain it clearly, even to a non-technical audience. Companies that genuinely use AI should be able to articulate how their technology works in simple terms. If an AI system is as advanced as claimed, its functionality should be transparent and understandable, even for non-experts. This transparency can help consumers make informed decisions and avoid misleading claims.

‍

What is Happening to Prevent AI Washing?

Although not specifically designed to tackle AI washing, both the UK and Europe are working on regulations to improve transparency in AI. The forthcoming EU AI Act is set to establish clear standards for AI, ensuring that systems are safe, transparent, and accountable. Organisations developing or using AI technologies will need to comply with regulatory frameworks, even if AI is just one feature of a broader tool.

To comply with these regulations, organisations are enhancing procurement processes to rigorously evaluate AI products, ensuring they align with responsible AI frameworks. This includes creating model inventories to track AI systems already in use and assessing new AI-powered products to ensure regulatory conformity.

‍

Final Thoughts

We are only at the beginning of AI’s journey, and the advancements ahead will bring even more complexity and sophistication. As technology evolves, it will become increasingly important to understand the true capabilities of AI while staying vigilant against exaggerated claims. By being informed and discerning, we can ensure that AI is utilised ethically and effectively. With upcoming regulations, transparency and accountability will play vital roles in guiding the responsible integration of AI into our everyday lives and businesses.