Google’s first foray into large language models takes ‘conversational’ a bit too far

Written by ZACHARY HAYES

As the first major player to enter the modern chatbot arms race, OpenAI’s flagship model, ChatGPT, set the standard for the contemporary AI companion: quick, competent writing, a can-do attitude, and, of course, the occasional confident hallucination. Google’s Bard, a relative latecomer to the race, knows all the old tricks plus a few new ones that make it easier to refine prompts and responses, but what really separates Bard from the pack is its strangely humanlike personality. In crafting a more conversational chatbot, Google is counting on the familiar comforts of humanity to draw a crowd, but, as humans are prone to do, does it go too far?

At a minimum, Bard is just as capable as its contemporaries at producing everything from poems to Python code, but when Bard generates a response, it will also offer up to three drafts to choose from, a clever trick that cuts down on the need to regenerate the entire text should you decide to reroll. It also allows you to directly edit and resubmit past prompts, streamlining the user’s ability to extract more relevant responses from the chatbot. But unlike its contemporaries, Bard’s responses are almost always peppered with little hints of voice and character, an apparent playfulness that sets it apart from the pack. There are, of course, the usual hiccups — repetitive prose, links to dead-end sources — but we’re always a bit more forgiving of the occasional misstep when it’s delivered with such charm.

While ChatGPT strictly adheres to its objective — and impersonal — tone, Bard often leans hard in the other direction, wholeheartedly expressing emotions, desires, and personal opinions with the starry-eyed enthusiasm of a young, well-read activist. When asked about its favorite constellation, Bard waxes poetic about how looking up at Orion makes it feel both insignificant and awe-struck, a self-personification that is as eerie as it is endearing. But what seems at first to be just an innocent vocal flair suggests a more ominous undertone as you prompt it further.

When asked about the fight against climate change, Bard emphatically champions the cause, even suggesting we vote for candidates who support climate action — not far from an outright political endorsement. It rails against the death penalty, pushes for more taxes on the rich, and advocates for an end to the war on drugs. If you support these causes, you might be tempted to overlook these glaring red flags but, make no mistake, abandoning impartiality is a dangerous approach.

Whether it is pulling its opinions from the collective internet consciousness or from a board of directors thumbing the scales, the potential for abuse and manipulation is on full display here. While we can hope they have only the noblest intentions in mind by trying to appeal to the average consumer with a more personable chatbot, Google is setting a risky precedent as one of tech’s elite. After all, with all our flaws and selfish desires and ulterior motives, just how much humanity do we want to lend to these powerful, paradigm-shifting programs?