TechSpot means tech analysis and advice you can trust. Read our ethics statement.
A hot potato: It is hard to say whether Google was being deliberately deceptive but there are billions of dollars at stake in the race to be No. 1 in generative AI. Anything that smacks of being second-best in this race will hurt. Unfortunately for the tech giant, that could include a demonstration video perceived to be faked.
With much fanfare, earlier this week Google unveiled Gemini AI, its most advanced AI model so far. As part of the launch, it released several videos on YouTube, X / Twitter, and a post describing Gemini's attributes and performance. But now Gemini's story is moving in a new direction; one that Google surely doesn't like and probably didn't anticipate. The six-minute demonstrative video has been seized upon as painting a misleading picture of Gemini's capabilities, namely that some of the depictions of Gemini in action didn't actually happen in real time even though the video suggested that they did.
For example, the video (watch below) featured conversations between a user and a Gemini-powered chatbot and also showcased Gemini's ability to tell the difference between visual pictures and physical objects. The reality is that Google used still images and fed text prompts to Gemini to get these results, according to The Information.
It doesn't appear that Google was trying, or at least trying very hard, to be deceptive. For starters, there is a disclaimer on the video that says, "For the purposes of this demo, latency has been reduced, and Gemini outputs have been shortened for brevity." Also, a Google developer detailed the prompting approaches it used with Gemini in the demonstration video.
In response to an inquiry from CNBC, Google told the publication that "The video is an illustrative depiction of the possibilities of interacting with Gemini, based on real multimodal prompts and outputs from testing. We look forward to seeing what people create when access to Gemini Pro opens on December 13."
How Google created its Gemini AI demo video pic.twitter.com/tAJnrPysh9– Trung Phan (@TrungTPhan) December 8, 2023
But whatever its intentions, for a global conglomerate like Google this was a serious misstep. Once the realization about the demonstration video hit, headlines and posts decried Google for "faking" part of its presentation with some suggesting that perhaps the company was already behind OpenAI. Indeed, billions of dollars are at stake in the race to deliver the best generative AI tools and in this race speediness will be a key factor. Anything suggesting that a tool requires complicated prompts or is not as fast as its competitors would be a death blow.