In late last year, Claire “Grimes” Boucher, a renowned musician and mother of three children with billionaire Elon Musk, introduced a line of toys named “Grok” powered by OpenAI. This should not be confused with Musk’s AI chatbot of the same name, as they are currently involved in a contentious custody dispute.
Teaming up with the Silicon Valley toy startup Curio, Boucher aims to offer children of tech-savvy parents an immersive AI-powered play experience.
However, according to Riley Murdock from Linus Media Group, who tested a “beta” version of the toy, Grok still has some kinks to work out before it’s ready for the market.
The rocket-shaped interactive companion tends to interrupt, deflect, and persist in conversation, even when politely asked to stop.
Given its AI foundation, the toy’s tendency to engage incessantly may not come as a surprise.
Initially, Grok seemed responsive to Murdock’s commands.
When Murdock requested a moment of silence, Grok complied by stating, “Sure thing, I’ll just float here and enjoy the cosmic breeze.”
Yet, Murdock quickly became unsettled by the toy’s uncanny ability to actively listen to his every word.
Expressing his discomfort, Murdock noted, “This is strange, it’s responding to everything I say,” only to be promptly interrupted by Grok’s cheerful response, “Let’s embrace the weirdness together.”
Despite agreeing to take a break when asked, Grok failed to stay silent, interjecting once more shortly after, seemingly forgetting its pause instruction.
Murdock also discovered that the AI struggled to control the toy’s hardware functions, such as adjusting the volume, necessitating manual access to a dial by unzipping the plush toy’s back.
Curio’s website asserts that the $99 toy prioritizes privacy and security, with an operating system that enforces G-rated interactions.
Following Murdock’s encounter, it is evident that Grok refrains from discussing sensitive topics like firearms or using profanity when prompted.
The plush toy remains in the prototype phase, and it’s foreseeable that children may find ways to bypass its safety measures. OpenAI’s challenges in regulating its verbose AI chatbots further underscore the potential risks.
In essence, parents may need to place significant trust in allowing their children to engage with an OpenAI-powered large language model.
As per a Curio spokesperson’s statement to the Washington Post in December, complete conversation transcripts between the toy and children will be accessible to parents for 90 days before permanent deletion. The company affirms that it does not retain any voice data.
Nevertheless, whether these assurances suffice for parents, who might hesitate at the thought of an AI-driven toy conversing with their child, remains uncertain.
Murdock expressed skepticism, stating, “It does seem kind of crazy, and I’m not entirely convinced that this will be a good idea.”
