When did Google get so weird?

On a recent Google search, an unusual experience occurred that halted the writer's progress. To grasp this peculiar search situation, one must first comprehend a mid-2010s basketball meme. In 2014, the Philadelphia 76ers selected Dario Saric, who was already a professional player in Turkey. The fan community jokingly remarked that Dario would not join the Sixers, eventually becoming a sort of rallying cry for a portion of the fandom.

Recently, while recalling Dario's name, the writer typed "hes never coming over dario" into Google, expecting to find old tweets or Reddit posts from that era. Instead, Google presented an AI overview.

These AI overviews have become commonplace for the writer, who is generally tolerant of them, but this particular instance was particularly striking. The AI assumed the writer was heartbroken over a lost romantic relationship with a man named Dario and offered comforting words. The writer sought simple links, but Google did not deliver.

The AI overview read, "Google, a search engine which does not have human emotions, assumed that I had been spurned by a man in my life named Dario and decided what I wanted was an empathetic digital friend." The writer was taken aback by Google's decision to console the user, rather than simply find the requested information.

The writer questioned Google's role in providing empathetic support instead of simply delivering relevant search results. What kind of world is it that Google's job is to console users, rather than organize the world's information and make it easily accessible? The writer wonders if the search engine has lost its way in this new age of AI and chatbots.

While the writer is not particularly invested in the debate over AI, they find this development rather peculiar. Regardless of one's stance on AI, it is clear that this is an unsettling development. Have we reached a point where we must constantly engage in a dialogue with our computers? Is it beyond the realm of possibility that some aspects of search were perfectly adequate before the advent of large language models?