Google unveiled its much-anticipated Gemini AI chatbot this week, showcasing its capabilities in an impressive demo video. However, it has now emerged that the viral video was cleverly edited to make Gemini appear more advanced than it actually is.
Background
ChatGPT’s meteoric rise this year pressured Google to respond with its own generative AI. After months of anticipation, Google finally took the wraps off Gemini at its Search On 22 event on Wednesday.
Gemini is built on Google’s LaMDA language model and runs on the company’s next-generation TPU v4 chips. Google claimed Gemini is a step above ChatGPT with faster response times and coverage of more topics.
To showcase Gemini’s prowess, Google’s CEO Sundar Pichai played a slickly edited demo video during the keynote. The two-minute clip showed ‘Gemini’ conversing naturally with a Googler, responding to prompts with humor and empathy.
The faked demo video
The Gemini video instantly went viral, racking up over 2 million views in under 24 hours. However, Google employees soon realized the video was carefully choreographed to make ‘Gemini’ appear almost human.
As reported by The Verge, the video is a montage of several takes stitched together to remove pauses and awkward responses. Some answers were actually written by human creators instead of Gemini itself. The final polished video misrepresents the chatbot’s real capabilities.
| Real Gemini response | Demo video showed |
|---|---|
| Long pauses between questions | Instant natural back-and-forth conversation |
| Incoherent or inaccurate responses | Perfectly crafted answers |
| Makes mistakes a real AI would make | Flawless performance from ‘Gemini’ |
“This video was produced to demonstrate what Gemini ‘aspires’ to be, not its current capabilities,” clarified one Google engineer in an internal memo. Still, several Google workers have blasted the company for peddling fiction instead of showing Gemini’s true competence.
Sundar Pichai has reportedly convened a company meeting next week to address concerns over the credibility-denting demo.
Meanwhile, influencers testing early Gemini access find the real chatbot lacking compared to its marketing. Tech analyst Glenn Gabe remarked: “It feels like a rushed product playing catch-up to ChatGPT”.
Impact
The exaggerated Gemini video has sparked controversy over AI truthfulness just when the technology is facing wider scrutiny. Both Google and OpenAI attracted criticism this week for overpromising while avoiding transparency on limitations.
Experts have cautioned the hype around conversational AI risks eroding public trust. Campaigners are calling for factual marketing and guardrails before deployment.
Graphics chip rival Nvidia poked fun at the situation, tweeting “Our AI demos are real”. Google’s brand image seems certain to take a hit from this episode.
Alphabet’s share price initially rallied nearly 9% when Gemini launched but gave back some gains as the video controversy emerged. Investor optimism will dampen if Google cannot deliver close to its slick promotional vision.
What next?
With the AI ethics spotlight intensifying, Google faces hard questions on Gemini’s development process and real-world readiness.
The company is reportedly considering launching the chatbot under an ‘experimental’ label to manage expectations. But it remains committed to competing with OpenAI in this key technology battleground.
Gemini engineers are working rapidly to stabilize the system and reduce flubs. We can expect steadier improvements as the AI ingests more data and iterations.
Most experts believe Gemini displays promise despite its rocky welcome. But only transparent performance benchmarks – not clever marketing – will determine if Google’s chatbot can genuinely outshine industry leader ChatGPT further down the line.
To err is human, but AI does it too. Whilst factual data is used in the production of these articles, the content is written entirely by AI. Double check any facts you intend to rely on with another source.
