ChatGPT was the first AI chatbot to conquer the crowds. It effectively set a standard for all future competitors, while becoming the chatbot to beat. Today, two new chatbots claim to be better than ChatGPT, and one of them hasn’t even been released yet.
Baidu Inc, China’s leading search engine provider, has been working on a worthy competitor to ChatGPT since March, with the release of Ernie Bot. The company claims that the latest version of this chatbot, Ernie 3.5, now outperforms ChatGPT.
In a statement, the company says its chatbot outperforms GPT-3.5-based ChatGPT in “overall capability” and outperforms GPT-4 in “multiple Chinese language capabilities.”
Ernie will support plugins
To support its claims, the company cites an article from China Science Daily, a Chinese newspaper, which conducted a test using two benchmarks, AGIEval and C-Eval, to measure the performance of AI models.
Ernie 3.5 has been tested in public beta for three months. Since then, the company claims to have improved the “efficiency, functionality and performance” of its tool. In its statement, the company also said that the chatbot will support plugins, like ChatGPT.
Google is also in the running to steal the top spot from ChatGPT. The tech giant’s first attempt to create an equivalent to ChatGPT, Google Bard, failed to dethrone ChatGPT and delivered a disappointing performance.
Gemini to the rescue of Google
Even when Google upgraded Bard from a lightweight version of LaMDA to a much more advanced LLM, PaLM 2, the chatbot’s performance didn’t improve much.
In an interview with Wired, the co-founder and CEO of Google DeepMind, Demis Hassabis, says on the contrary that his next model will be more efficient than that of ChatGPT.
The model Hassabis is referring to is Gemini, which was originally unveiled during Google’s keynote, during which the company made numerous AI announcements.
Redesign of teams at Google
In the interview, Hassabis revealed that to build Gemini, DeepMind uses techniques from AlphaGo, the company’s high-performance AI system, which was the first to defeat a human professional Go game.
“Gemini combines some of the strengths of AlphaGo-like systems with the amazing language capabilities of large models,” Hassabis told Wired.
However, Gemini is still under development by DeepMind and it could take several more months.
The development of Gemini follows the merger last April of Brain Team, Google’s research team dedicated to machine learning and AI, and DeepMind to create a new group called Google DeepMind. .
Source: “ZDNet.com”