Google admits that a Gemini AI demo video was staged

L4sBot@lemmy.worldmod to Technology@lemmy.world – 249 points –
Google admits that a Gemini AI demo video was staged
engadget.com

Google admits that a Gemini AI demo video was staged::Google admits that it staged part of a Gemini AI demo video, in which there was actually no voice interaction as implied.

26

Nothing matters anymore. Lie on your resume. Don't shovel the snow. Be absolutely naked. Throw your careers to the wind. Bathe in the pale moonlight amongst the corpses of oligarchs.

They probably should've disclosed that beforehand, or as part of the video, but anyone with any experience with AI (ChatGPT, Midjourney, etc) knew the voice was staged to make for a better presentation.

They did. On their Gemini webpage it has the marketing stuff, the marketing video (the one that everyone saw), and linked to blogs about how they performed the tasks in the video. So Google hasn't admitted anything - they stated it from the start. We could argue that they should have stated it in the video but what marketing material does? Eg. Redbull's stuff suggests that their product gives us wings.

Google hasn’t blown OpenAI out of the water because they don’t want to. They’re an ad company and they can’t move away from that.

I think you aren't aware of how much Google sucks at even simple software.

They really do a surprisingly bad job at things.

I wish they could compete with apple on quality, but it's just not there.

They value engineering too much and design too little. It results in not only shittily-designed products, but also shittily-engineered ones too because they over-engineer everything.

5 more...

That's funny.

https://killedbygoogle.com/

They've always experimented with new stuff.

Yeah but LLMs eat into the ad model as I see it. Less visiting ad links when the bot spits out your answer for you

Put ad links into the bot. Simple really.

Damn it all players are switching from LP to CD, better stop trying to develop and give up

5 more...

That's not quite correct. It says it wasn't real time because the guy was typing his questions, not speaking them.

However the responses from the AI were all correct. Maybe just not as fast. Maybe it took a few seconds to respond, but that is acceptable at this stage.

They wanted to show you how amazing it really is, and could be if it were hooked up to speech recognition and was able to resolve quicker. Which one day will be the case.

Visit Gemini's webpage, there are links to Google blogs that show exactly what the prompts and responses actually were.