Why it is so complicated to advertise AI

Many tech companies, including the largest ones, are experiencing difficulties in promoting their AI-related services, especially when it comes to chatbots, ending up producing advertisements considered at best unsuccessful and at worst entirely misleading. Meta, for example, last June released an ad that sparked much discussion about the company’s AI suite. In one scene of the advertisement, a young man asks the AI to imagine Little Italy in 1954 to show the result to his grandfather. The clip was heavily criticized because, according to many users, it doesn't make much sense to ask an AI to generate images of the famous New York neighborhood in the 1950s when a simple Internet search would provide real photos. Additionally, according to critics, the result shown in the ad was no better than any available online photo. But Meta's case wasn't isolated. For the Paris Olympics, Google created a campaign to promote Gemini, which is practically equivalent to ChatGPT. The ad – broadcast in the United States and online – is titled “Dear Sydney”: it’s about a girl who dreams of becoming a professional athlete and wants to write a letter to her sports idol, Sydney McLaughlin-Levrone – the world record holder in the 400-meter hurdles. To help her, her father decides to turn to Gemini, asking it to generate the letter. In the ad, it is said that this is a special occasion, and thus the letter must be perfect – implying that it deserves to be created by an AI, not by a human being.

Within a few days, the advertisement attracted a lot of criticism, especially online, to the point that Google pulled it. The ad was described as cold, as it downplayed moments that should instead foster closeness between parents and children, entrusting them instead to a machine. «It’s hard to think of anything less inspiring than instructing an AI to tell someone how inspiring they are», commented the U.S. tech magazine TechCrunch. But Microsoft also, during the Paris Olympics, promoted its chatbot Copilot with an ad deemed inappropriate by many. The ad explains how the service can help athletes train. In one scene, an athlete asks the AI to analyze data on his heart rate. However, users pointed out that – at least so far – the numerous attempts by tech companies to create chatbots capable of providing medical assistance haven't yielded great results. These systems are known for slipping into unpredictable errors, also called “hallucinations,” which could have non-negligible health consequences.

What’s unconvincing about AI ads

It may not be a coincidence that three of the most powerful tech companies in the world, with ample marketing resources, have had similar problems advertising their AI systems. First of all, the ads focus on situations that many people wouldn't want to delegate to technology (at least theoretically). Additionally, the ads were considered unsuccessful because they don't truly showcase applications of artificial intelligence that weren't already known. At the same time, however, the aspects where these technologies can make a difference are the same ones that generally expose companies to the most criticism. Consider this: an ad about a chatbot capable of replacing a software developer’s job – a still remote possibility – would not likely be well received. The same applies to ads about students using chatbots, which risk portraying these technologies as tools that allow students to put less effort into school. And the fact that the tech industry revolving around AI is grappling with the great expectations generated in recent years doesn’t help. The applications of these tools, in fact, are proving to be less exciting than expected, and some wonder if artificial intelligence is just a big bubble. It’s no coincidence that the most critical observers argue that using AI for very simple tasks, like those shown in these ads, would be – as the Wall Street Journal puts it – like taking a Lamborghini to deliver a pizza.