January was memorable in the media landscape with the news of Nvidia's stock dropping 17%—a reaction to the record-breaking app downloads of DeepSeek, a Chinese AI chatbot that outperformed ChatGPT in this criteria.
Why is it important? Because it is a critical point in the shift of the AI paradigm — from hype — toward a more realistic and weighted narration.
Numerous bylines appeared, in which experts talk about the overwhelming hype around LLMs.
Eric Siegel, writing in Forbes, argues that the hype around large language models (LLMs) inspired "a prevalent belief that technological progress will shortly arrive at artificial general intelligence, computers capable of almost everything humans can do." He says that while LLMs are groundbreaking, AI should not be viewed as a panacea. Specifically, he explains that the growing narrative around artificial general intelligence (AGI)—a machine intelligence with human-like cognitive abilities—is greatly exaggerated, and that the market's panic over AI’s potential disruption underscores it's reliance on speculative hype than on actual technological foundation.
Andy Kurtzig, writing in TechRadar, is even more critical towards AI, arguing that "in high-stakes fields like medicine, law, veterinary advice, and financial planning, AI isn’t just “not there yet,” it may never get there." The author provides statistics on AI errors as research found that AI makes mistakes in 52% of cases. He provides a more disastrous example which goes beyond academic consequences: A 14-year-old boy seeking mental health help from AI was advised to take his own life, which—unbelievable!—he did. Now, his family is now suing—and they’ll likely win—because the AI’s output wasn’t just a “hallucination”.
Cybersecurity, for sure, will be among critical areas in which hallucinations would be unacceptable. In this regard, I noted a timely campaign by Sophos exploring limitations and risks of AI in cybersecurity. The report "Beyond the Hype: The Business Reality of AI for Cybersecurity" addresses the mentioned issues and is notable with the introduction of a new risk category: over-reliance on AI.
A critical takeaway for communications professionals will be on the importance of anticipating shifts in trend dynamics at the very early stage and adjusting the messaging accordingly. It's notable that the skepticism toward AI had already started building before Nvidia’s stock drop on January 28. Just think that Sophos realized their report the same day! Occasional coincidence? Well-planned campaign? Well, while we don't know the answer to these questions, we know that preparing such a report takes more than one day, meaning that the company should have started its work months before.
One more takeaway is that we are entering a stage of evidence-based discussions on AI. Those who can anticipate and adapt to this new tone will be better positioned to lead conversations and build credibility for their brands.