AI washing: The big new deception in Silicon Valley
Would you mind looking over all the previous pitch presentations and substituting the term ‘crypto’ with ‘A.I.
Would you mind looking over all the previous pitch presentations and substituting the term ‘crypto’ with ‘A.I.’?
This headline, featured in a New Yorker comic illustrated by Benjamin Schwartz, succinctly encapsulates the emerging trend of AI washing within Silicon Valley.
AI washing may seem like just another marketing tactic, but in reality, it is a sophisticated and multifaceted occurrence. It is crucial for all individuals consuming this piece—be it tech executives, advertisers, developers, users, or IT experts across all sectors—to grasp the exaggerations, skewed emphases, and blatant falsehoods that saturate not only marketing and sales practices, but also the narratives we encounter based on industry assertions.
Understanding AI washing
AI washing entails a deceitful marketing strategy that exaggerates the significance of artificial intelligence in the product or service being endorsed. This term draws inspiration from “greenwashing,” a concept introduced by environmentalist Jay Westerveld in 1986, where consumer goods are marketed as environmentally sustainable irrespective of their environmental impact.
Items utilizing traditional algorithms are tagged as “AI-powered,” exploiting the absence of a universally accepted definition of what qualifies as AI and what does not. Emerging companies develop applications that integrate with publicly accessible generative AI APIs and promote them as AI applications. Grand AI initiatives intended to showcase technological advancements often rely on human intervention behind the scenes, as humans are essential for the successful operation of such ambitious AI solutions.
Let’s delve deeper into the latter point.
AI: The human element
Retail behemoth Amazon launched 44 cutting-edge stores dubbed Amazon Go and Amazon Fresh, which (commencing in 2016) employed the company’s “Just Walk Out” suite of technologies. (I initially reported on this endeavor in 2017.)
Amazon’s vision entailed establishments where patrons could enter, select items from shelves, then exit without encountering a cashier. Sensors, including cameras, fed data to AI systems capable of discerning purchasers’ selections and billing them accordingly—completely bypassing any checkout process. It somewhat resembled shoplifting, but remained legal.
The system hinged on advanced computer vision systems that monitored customers and their selections. Shelf-mounted sensors relayed information on the weight of items taken, verifying the type and quantity of products identified by the cameras. RFID-tagged items further enriched the data pool. Sophisticated machine learning algorithms analyzed the input from cameras and sensors to identify products and link them to specific shoppers. Electronic entry and exit gates determined the identities and movements of individuals.
These algorithms underwent training on millions of AI-generated images and videos to recognize products, human behaviors, and actions.
For seven years, Amazon eagerly discussed these components of its Just Walk Out technologies. However, the tech giant was reticent about the approximately 1,000 human operators engaged to ensure the system’s functionality—an acknowledgment only revealed following media revelations. Even then, Amazon obscured the specific tasks these employees undertook, disclosing only that they did not review footage.
Despite having 1,000 staff members overseeing operations at 44 stores (vetting three-quarters of transactions, according to reports), the technology experienced issues, including delayed receipts, mishandled orders, and hefty operational expenses.
This year, Amazon has been phasing out the Just Walk Out technology from its primary outlets but still provides it as a service to external entities.
Another striking example of human involvement behind the AI facade is found in the realm of autonomous vehicles.
Alphabet’s Waymo (formerly Google’s self-driving vehicle venture) operates a NASA-inspired command center where employees monitor auto functions through cameras and remotely intervene during anomalies. (Here’s a timelapse video I recently captured during a trip in a Waymo vehicle through San Francisco.)
General Motors’ subsidiary, Cruise, acknowledges that its self-driving taxis necessitate human intervention every 4 to 5 miles on average, with each remote control session lasting about 3 seconds.
Several other self-driving enterprises rely even more heavily on human remote operators. For instance, a German firm named Vay directly employs human operators to drive vehicles remotely. The company recently introduced a valet parking service in Las Vegas, where a remote operator steers the car to the customer, who then drives it to their destination. Upon arrival, the customer exits, and a remote operator handles the parking.
Amazon’s stores and self-driving vehicles represent just a couple of examples illustrating a broader trend.
Rationale behind AI washing
The upper echelon of technologists crafting AI systems have unwavering faith in AI’s potential to address incredibly complex dilemmas—and theoretically, it can. They assert to their superiors that their objectives are attainable. Those leaders, in turn, assure their boards that the goals are feasible. Corporations communicate to investors that these feats are within reach. And collectively as an organization, they relay to the public that their ambitions are achievable.
However, there arises a small issue: these claims often fall short.
Most companies experience a sense of obligation to uphold grand assertions, leading them to downplay the extent to which their products or services rely on human intervention behind the scenes to navigate challenges, make decisions, and facilitate the purported “magic”.
Companies with brazen attitudes often remain undeterred by evidence revealing that their AI capabilities are not as robust as they proclaim or believe, subsequently persisting in reiterating their claims time and again. Tesla CEO Elon Musk exemplifies this stance.
In October 2016, Musk pledged that Tesla would demonstrate a fully autonomous cross-country drive from Los Angeles to New York by the end of 2017.
By April 2017, he forecasted that within about two years, occupants would be able to doze off in their vehicles while they autonomously traversed.
In 2018, Musk moved up the timeline for fully autonomous Tesla models to the conclusion of 2019.
In February 2019, Musk assured that full self-driving capabilities would be realized “this year.”
In 2020, Musk asserted that Tesla would deploy over 1 million self-driving robotaxis by the year-end.
Even this year, Musk expressed optimism that fully self-driving Teslas might become a reality “later this year.”
However, this vision is unlikely to materialize. Musk is deceiving himself and his consumers. Musk epitomizes the epitome of AI washing.
The real dilemma with AI washing
The collective consequence of AI washing is that it leads both the public and the technology sector astray. It nurtures the misconception that AI can accomplish tasks beyond its capabilities. It fosters the belief that AI is a universal solution to varied challenges—or conversely, a precipice into a dystopian era, depending on one’s outlook.
AI washing encourages inferior solutions replete with fanciful claims rather than emphasizing quality. A claim asserting that your dog-washing hose is “AI-powered” does not equate to a cleaner pet; it merely denotes an inflated price tag on a hose.
AI washing distorts investment trends. Presently, Silicon Valley investments predominantly sway towards authentic AI solutions and AI-washing facades. Even seasoned investors may overlook instances of AI-washing exaggerations and falsehoods, reassured by the notion that AI narratives readily resonate in the market owing to gullible buyers.
Nonetheless, the principal concern lies not in the overzealous salesmanship within the industry but in self-delusion. Proponents of AI solutions regard human assistance as a form of disgrace, whereas I firmly believe human involvement would be welcomed.
Consumers, in fact, desire human engagement in their shopping and driving experiences.
What we truly necessitate is more human involvement and less machine autonomy. Presently, an inundation of AI-generated content—ranging from cringeworthy scripts to fallacies and unsettling images—is inundating the landscape. Google’s eagerness to transition from a search portal to a response engine often results in unconventional outcomes, such as adhesive on pizza.
The public yearns for a search tool that directs us towards human-crafted content, or at the very least, a PageRank system that favors human work over AI-generated creations.
The AI-washing phenomenon is founded on deception. It hinges on the notion that individuals desire a society dominated by automated entities, which is not accurate. It asserts that the mere addition of AI enhances any entity automatically, which is untrue. It propagates the fallacy that the involvement of people signifies a technological shortfall, which is unfounded.
Enough with the duplicitous AI washing! Vendors must delineate the truth regarding AI. Consumers must demand concrete evidence that any AI integrated into the products and services they procure serves a practical purpose.
I am confident that I articulate the sentiments of the technology sector, the tech consumer community, and the tech media in imploring Silicon Valley: Cease manipulating perceptions concerning AI.
About Author
