Newsletter
So, Google just launched its multimodal juggernaut named Gemini. It’s the new extremely powerful AI model from the company, and it’s meant to go up against models like GPT-4. As part of all of the launch festivities, we saw a hands-on video showcasing Gemini’s capabilities. Well, Google admitted that its Gemini hands-on video was staged.
So, to catch you up, when Google launched Gemini, it showed a hands-on video where the person was showing off some of Gemini’s abilities. It gave the appearance that Gemini was processing real-time audio and video data. When the person would put an object into the camera, it would ask Gemini questions about what it “sees” and provide answers. We’d hear an AI-generated voice respond.
It’s a great showcase of Gemini’s capabilities… or it would be if it were REAL.
Google admits that its Gemini hands-on video was staged
A Bloomberg opinion piece spilled this bit of tea. It states that Google revealed that the video wasn’t 100% real. The real-time vocal interactions between the presenter and Gemini weren’t there. That was all through the magic of video editing. Also, the interactions were sped up in post, which made it seem faster than it actually is.
But, while the video wasn’t 100% real, we can’t say that it was 100% fake. It’s a showcase of Gemini’s abilities, and we’re still seeing its abilities. Google used “still image frames from the footage, and prompting via text.” So, rather than having a casual conversation with Gemini, the company fed still images into the model and typed what it wanted Gemini to produce.
In essence, we’re still seeing Gemini’s capabilities; we’re still seeing what it can produce given the input. Google used Hollywood magic to make it seem more powerful than it is. As for the speed of the responses, Google stated in the description that the responses were sped up for brevity.
Is the company wrong for doing this? Who knows? That’s a debate for the YouTube comment section.
The video was staged, and that’s a bit of a relief
Regardless of whether the video was faked, it’s still much more powerful than Bard. The model is smarter with more tokens and parameters, blah blah blah. No matter what happens, businesses will still have tools to speed up production and efficiency. There are also several ways to access Gemini.
However, the video got pretty scary for any creators watching. We literally saw Gemini create a cool tropical song in seconds, something that would take a composer much longer. We also saw it create images in seconds from yarn. Ever since DALL-E finally got good and since ChatGPT hit the market, human creators have been on the verge of being obsolete. The situation isn’t getting any better, and the hands-on video truly made it look like Google finally put the final nail in the coffin for creators.
However, the fact that it was staged, shows that the technology isn’t quite there just yet. Creators have just a bit more time. That’s all we can ask for at this point.