What you need to know
- The Wall Street Journal’s senior personal tech columnist, Joanna Stern, recently took Microsoft Copilot, Meta AI, ChatGPT, and Gemini out on a 24-hour excursion in the woods.
- The AI tools were subjected to a variety of tasks, and findings indicated that their responses to queries were reminiscent of each other, though Google’s Gemini instructions on lighting a fire were a tad flawed.
- An attempt to befriend the chatbots depicted fake emotions and their inability to experience things such as touch, smell, and sight.
- While OpenAI’s ChatGPT depicted a more natural connection with humans, including referring to them using their nickname without prompting, Microsoft Copilot’s experience seemed a little over the top and annoying.
AI-powered chatbots are increasingly becoming human-like, and The Wall Street Journal’s Joanna Stern recently brought this premise to fruition. Ordinarily, you’d go hiking or camping in the woods with friends and family, but Stern audaciously brought OpenAI’s ChatGPT, Meta AI, Gemini, and Microsoft Copilot.
As you may know, chatbots ship with a customizable experience tailored to your specifications, including tone. For instance, Meta AI lets you customize your experience by allowing you to select the voice of a celebrity like Snoop Dogg, which it mimics while interacting with you.
Stern simultaneously put the chatbots through various tasks to assess their effectiveness and efficiency. Apple’s recently overhauled Siri failed to meet the cut because it fell short of expectations. When asked to introduce itself, Siri oddly asked whether the user wanted to restart their device.
First, Stern asked the chatbots what she’d need to start a fire in the woods. Interestingly, they provided almost identical answers, including detailed instructions describing how to chop kindling from trees using an axe. Stern indicates that Gemini provided poor fire-starting practices, but Meta AI, ChatGPT, and Microsoft Copilot salvaged the situation.
The next task included a prompt requesting the chatbots to recommend meatball recipes. While they managed to generate helpful instructions and required ingredients, they failed at simple tasks such as setting timers, alarms, or reminders. However, Meta AI indicated it had set a timer and would send a notification once the duration lapsed, but it didn’t.
Notably, Google’s Pixel buds and Meta’s AI glasses were handy throughout the process, especially where the user’s hands were engaged elsewhere. The accessories ship with built-in AI capabilities, fostering a seamless user experience.
“For sure. I think this is a new era, and it’s something that we have to embrace. We’ll come to trust our AI companions with lots of very important information and very important tasks in our lives,” indicated Microsoft AI CEO Mustafa Suleyman while responding to Stern’s question about AI chatbots being friends with humans.
The executive recently shared Copilot’s roadmap, including its evolution into a virtual companion that can become a friend and foster meaningful and lasting relationships with users. However, Copilot’s latest update has received backlash from users, citing a degraded user experience while highlighting their preference for the previous version. “It tries to be my friend when I need it to be a tool,” a concerned user indicated.
In The Wall Street Journal’s YouTube video, Microsoft Copilot came off as the annoying kid in class, seemingly trying a little too hard to appear friendly. “How are the vibes this afternoon? Just good vibes, peaceful vibes, forest vibes..” Copilot indicated.
On the other hand, OpenAI’s ChatGPT featured a more natural experience, including referring to Joanna Stern by her nickname (Jo) without prompting. To this end, it’s clear that AI chatbots can’t replace the human touch at any level due to their inability to feel, touch, and experience things and situations, including humor.
However, Stern made an amazing discovery that AI chatbots are extremely good at befriending each other and can even have human-like conversations with each other on a wide array of topics, acting more like a book club rather than AI chatbots.
