Understanding 'AI Washing' and Its Implications
Amazon’s Just Walk Out technology, used in Amazon Fresh and Amazon Go stores, allows customers to pick items and leave without traditional checkout processes. This system, leveraging AI through facial recognition, sensors, and cameras, automatically bills customers for their chosen products. However, reports surfaced in April revealing that instead of solely relying on AI, around 1,000 workers in India were needed to manually verify almost 75% of transactions.
Amazon quickly refuted these claims, stating that the reports were inaccurate and that the Indian staff were only reviewing the system, not monitoring video footage from all stores. They emphasized that human reviewers are a standard component in any high-accuracy AI system. Despite this, Amazon announced plans to reduce the number of stores utilizing the Just Walk Out technology.
This incident highlights a growing concern: companies potentially exaggerating their use of AI, a phenomenon known as “AI washing,” akin to environmental “greenwashing.”
To understand AI washing, it’s crucial to grasp what AI entails. While there isn’t a precise definition, AI enables computers to learn and solve problems by training on extensive data sets. Generative AI, a type that’s gained significant attention, creates new content such as text, music, or images, with popular examples including ChatGPT, Google’s Gemini, and Microsoft’s Copilot.
AI washing manifests in various forms. Some companies claim to use AI when relying on less advanced technology, overstate the effectiveness of their AI solutions, or present their AI as fully operational when it isn’t. Others may simply add an AI chatbot to their existing non-AI software.
According to OpenOcean, a UK and Finland-based tech investment firm, the percentage of tech startups mentioning AI in their pitches increased from 10% in 2022 to over 25% in 2023, with expectations for this figure to exceed a third in the current year. Sri Ayangar from OpenOcean attributes this rise to intense competition for funding and the need to appear innovative. He notes a significant gap between companies claiming AI capabilities and those delivering genuine AI-driven results.
This issue isn’t new. A 2019 study by MMC Ventures found that 40% of self-proclaimed AI startups used minimal to no AI. Simon Menashy from MMC Ventures explains that while cutting-edge AI is now accessible at standard software prices, many companies opt to overlay a chatbot interface on non-AI products instead of developing comprehensive AI systems.
Dougal Dick from KPMG highlights that the lack of a unified AI definition exacerbates AI washing. He points out that differing interpretations of AI allow for broad and ambiguous usage of the term, enabling companies to mislead. This can lead to businesses overpaying for technology, failing to meet operational goals, and investors struggling to identify truly innovative firms. Consumer trust in authentic AI advancements may also be undermined.
Regulatory bodies are beginning to address AI washing. The US Securities and Exchange Commission (SEC) recently charged two investment advisory firms for making false AI usage claims, signaling stricter enforcement ahead. In the UK, the Advertising Standards Authority (ASA) enforces a code of conduct against misleading marketing, including AI claims.
Michael Cordeaux from Walker Morris notes that AI-related advertisements are increasingly scrutinized by the ASA. Misleading claims, such as an app overstating its AI photo enhancement capabilities, have faced regulatory action.
Sandra Wachter, an Oxford University professor, argues that the current AI hype overlooks practical considerations. She questions the necessity of AI in products like electric toothbrushes and emphasizes the environmental impact of AI, which contributes significantly to climate change.
Advika Jalan from MMC Ventures believes that AI washing may diminish as AI becomes ubiquitous, losing its novelty as a branding tool. Eventually, “AI-powered” might be as commonplace and unremarkable as “we’re on the internet.”
In summary, AI washing poses significant challenges for businesses, investors, and consumers, but increased regulatory scrutiny and the eventual normalization of AI could mitigate these issues over time.
