Saturday, November 9, 2024

Using Gemini Live was faster than Google, but also more awkward

Must read

A week ago, I was on a three-day road trip when all the audio cut out in my car. From podcasts to directions — even the turn signal wasn’t clicking. My copilot struggled to find a solution via Google. It took me at least five minutes. Today, it took Gemini Live, the just-keep-talking-to-it version of Google’s voice assistant, about 15 seconds to come up with a solution that worked. Naturally, the moment felt magical. In The Verge’s first interaction with the assistant, rolling out today, it felt like the promised cleverness of digital assistants is finally being delivered.

But then Gemini Live kept talking. And talking. The Verge team was packed in a glass booth, and as Gemini Live droned on, a friendly Google employee encouraged me to “go ahead and interrupt it.”

It felt weird! I don’t mind interrupting Google Assistant in my car. In fact, I can be downright abusive to most of these bots. I call them names and interrupt them with ease. But Gemini Live felt different. The pleasing masculine tone of the voice, the easy way it spoke. It felt a little too human for me to interrupt.

My next question led to a similar interaction. I asked for ideas on how to entertain my dog, and Gemini Live just started talking. The only way I could get it to stop was to interrupt it. Which I did repeatedly. It was like talking to my 9-year-old godson. Like him, Gemini Live doesn’t know how to read the cues on my face, doesn’t know when to acknowledge that, actually, I don’t care as much about the subject at hand as it does.

I found myself getting caught up in these little interactions with Gemini Live more than I found it useful as a partner to brainstorm with. Its ideas for my dog weren’t especially inventive. When I interrupted it to ask how I could make a doggy obstacle course in my apartment without pissing off the neighbors, it just kept spitting out more ideas. In frustration, I accused it of mansplaining to me.

Gemini Live quickly apologized, rationalizing its mansplaining before offering to change its tone. Embarrassed, I ceded the demo to my colleague Sean Hollister. More familiar with these bots, he had no problem interrupting it at the drop of a hat. There was no hesitancy on his part. No being lulled into a sense of human interaction as I had been. He asked it to create a Dungeons & Dragons campaign, and while it did conflate that with his second question about investment advice, it came up with a clever hybrid idea. He asked it for stock tips, and it demurred, but it admired the Boglehead strategy. Then he asked it if he should invest in crypto, and it cautioned him to be careful and noted it “personally” wouldn’t invest.

Sean didn’t have the hang-ups with the familiar human tics of Gemini Live that I did. He’d practiced with other, more human-like digital assistants. He could easily ignore social training and just demand answers. As Gemini Live launches, I’ll be curious to see how many people get hung up on the human qualities and how many just treat it like a nicer-sounding Alexa.

You can berate it yourself if you have a Gemini Advanced subscription and a powerful enough Android device. iOS users will need to wait a little longer. We’re eager to compare it to OpenAI’s ChatGPT voice assistant and whatever smarts Apple eventually packs into Siri.

Latest article