Google's Gemini criticized as far inferior to GPT-4

At its launch event on December 6, Google showed a video of Gemini interacting with visual data through a camera mounted on a table, asking questions and solving problems like a human assistant would manipulate objects. The video showing Gemini’s ability to perform tasks smoothly quickly made a strong impression on users. However, Google admitted to Bloomberg that the video had been edited.

Not only that, Gemini continues to disappoint netizens by providing false information. This reminds many people of Google's "full of holes" launch of Bard AI in February when the chatbot gave the wrong answer right in the product introduction of CEO Sundar Pichai. Bard AI is currently running on Gemini Pro version.

Gemini's capabilities do not match user expectations

According to Insider , Bard AI repeatedly answered simple questions incorrectly. One user said the chatbot even made up information when asked who won the Best Actor award at the 2023 Oscars. In addition, Bard's language skills are also limited. For example, when asked to "give me a six-letter word in French", Gemini Pro gave an answer of only five letters.

When users ask for information on controversial topics, Bard AI will refuse to answer and advise users to search Google themselves. Meanwhile, rival chatbots such as billionaire Elon Musk's Grok, Bing, and ChatGPT Pro can provide detailed answers.

Google has previously warned that Bard AI may display inaccurate information, so users should double-check answers. The company said Gemini Ultra outperformed OpenAI's GPT-4, but the Pro version only outperformed GPT-3.5.

Google is rolling out Gemini in phases, with the company planning to launch the top-end Ultra version in early 2024.

Source link: https://thanhnien.vn/gemini-cua-google-bi-che-kem-xa-gpt-4-185231209054911276.htm