Turns out Google’s killer ChatGPT tool isn’t so smart: the demo turned out to be fake
Earlier this week, Google introduced the Gemini “ChatGPT Killer,” an AI model that will make chatbot Bard smarter. Now this turns out to be disappointing: many of the demonstrations appear to be exaggerated or a bit different.
The Gemini show was accompanied by A Impressive video Where the conversation takes place. The AI is asked via voice command what can be seen in drawings placed on paper using a marker. Gemini’s responses are very quick and correct every time.
It now appears that Google has extensively edited these images. If you look at the description below the video on YouTube, you will see that the delay has been reduced and the answers have been shortened.
What could it look like?
but this is not all. Reverse Bloomberg Google admits that it also worked with text prompts and not with chat, as the video suggests.
Oriol Vinyals of Google Deepmind, where Gemini was developed, defends these choices in one way message On Social Network
Gemini makes a lot of mistakes
In a press release, Google said that Gemini Pro scored six out of eight times better than ChatGPT 3.5 (the free version of ChatGPT) in tests. On X Complain Users, however, that the new poet still has many simple facts wrong.
For example, Google Bard with Gemini Pro under the hood has a bug in determining 2022 Oscar winners, and when the chatbot is asked to provide a six-letter French word, inquirers get “amour” or “fromage” — neither They are six-letter words.
Gemini will be rolled out in various phases in the near future. Gemini Pro is now available on Google Bard, but only in English and not in Europe. A better version of Gemini called Gemini Ultra will be released in early 2024. This can then be used via Bard Advanced, a type of ChatGPT Plus.
Don’t miss anything, follow Our channel on WhatsApp Or subscribe to the Bright Daily newsletter.
“Web maven. Infuriatingly humble beer geek. Bacon fanatic. Typical creator. Music expert.”