Google Definitely Had Its ‘Hands-On’ That Gemini AI Demo

Google releasedGemini , its newest productive AI chatbot that promises to bebetter than ChatGPT , alongside a splashy demo video showcasing its capabilities . A Google developer web log post let out that the viral display case was edited , sped up , and used dissimilar prompts than the 1 shown in the video .

“ For the purposes of this demo , response time has been reduced and Gemini outputs have been castrate for brevity , ” enjoin Google in theGemini demo videodescription , but that does n’t describe the full amount of redaction here .

The spill of Gemini was a large marketing present moment for Google ’s artificial intelligence team , which has faced pressure from investors tocatch up with OpenAI . Google ’s main AI chatbot Bard has jug behindChatGPTand other models in 2023 , so Gemini needed to help Google return to the cutting edge of the AI conversation – and it did . However , Gemini ’s demonstration video seems to have been an over - magnification of its factual capability . Google crowd back theoriginal launchof Gemini because Sundar Pichai take for it was n’t ready , and now we ’re ascertain why .

Screenshot: Google

The demonstration video appeared as if Gemini work with spoken language , picture visuals , and paw motion , but that ’s not how Google got those answer . Behind the scene , Gemini call for more careful , longer prompts with text and figure to get those reply in the video , according to aGoogle web log post . “ Hands - on with Gemini ” appear as if the AI was reacting as a existent - sentence assistant with video and articulation capabilities , but that was deceptive , and user will not have those features anytime soon . The telecasting was likely filmed with Gemini Ultra , the most advanced version of Google ’s AI that wo n’t be released until at least next class .

Google ’s claims that Gemini is good than ChatGPT seemmisleadingas well , as Bloomberg points out . Google touted a model of Gemini that does n’t exist yet ( Ultra , hail 2024 ) against outdated manakin from OpenAI ( ChatGPT-4 , from March 2023 ) . The bench mark does n’t compare toChatGPT-4 Turbo , OpenAI ’s most advanced AI modeling that was discharge last calendar month . Google ’s mannikin that ’s available today , Gemini Pro , does n’t tick ChatGPT-4 in Massive Multitask Language Understanding – it loses by over 7 % .

As users got their first chance to expend Gemini Pro on Wednesday , many mat the experience miss the luster of Google ’s merchandising campaign . Gemini Pro seems to ‘ hallucinate ’ a lot , a common problem where chatbots confidently give imitation answer . One user posted on X that Gemini Pro could n’t answer who wonbest actor at the 2023 Oscar . Another user incur that Gemini Pro tell them to “ Google it ” when need questions about Israel and Gaza .

Benchmarks comparing Gemini to top generative AI models