Google claims that its new AI is better than OpenAI’s ChatGPT-4
Google’s ‘Gemini’ is being built with the intention of utilising advancements in reinforcement learning to complete tasks that current language models find challenging.
Highlights
- Google’s DeepMind has unveiled Gemini, an advanced large language model (LLM)
- Gemini will be capable of planning, problem-solving, and text analysis
- It will be more powerful than OpenAI's ChatGPT, says DeepMind’s CEO
Due to its novel use of AI for human interaction, ChatGPT has gained widespread attention. Because of this reason, a number of AI firms have entered the fray and introduced similar chatbots to rival ChatGPT. Recently, Google DeepMind announced that it would challenge the ChatGPT platform. DeepMind’s CEO Demis Hassabis now claims that his developers are employing AlphaGo principles to create Gemini, an AI system that will be more powerful than OpenAI's ChatGPT.
What is Gemini?
DeepMind, a division of Google's AI research, has unveiled Gemini, an advanced large language model (LLM) that is competing to be the next major development in artificial intelligence (AI) after making an impact with its long language model (LLM), PaLM 2.
The chatbot is being built using principles from DeepMind's AI programme AlphaGo, which made history by defeating a ‘Go’ champion named Lee Sedol in 2016. ‘Go’ is an east Asian game known for its limitless move possibilities, and strategic thinking. These same approaches are being used to create Gemini.
Gemini has not yet been made public, but DeepMind CEO Demis Hassabis has already asserted that the multimodal model would be more effective than GPT-4, the technology that supports the well-known AI chatbot, ChatGPT.
What's the mission of Gemini?
Gemini's objective is to create a chatbot that is competitive with or even better than OpenAI's ChatGPT. According to Demis Hassabis, the chatbot would be able to plan out scenarios or find solutions, in addition to analysing text.
Gemini is being created with the intention of using advancements in reinforcement learning to carry out tasks that current language models find challenging. The chatbot's development aims to capitalise on the generative AI industry.
According to a recent study by Grand View Research, Inc., the size of the worldwide generative AI industry is expected to reach USD 109.37 billion by 2030. The market is anticipated to grow at a CAGR of 35.6% from 2023 to 2030.
Can Gemini outperform ChatGPT?
According to Hassabis, the developers working on Gemini are employing many of the strategies used by DeepMind's AlphaGo. Gemini attempts to outperform other chatbots by giving them special problem-solving abilities and producing original output that exists outside of their training set.
"At a high level, you can think of Gemini as combining some of the strengths of AlphaGo-type systems with the amazing language capabilities of the large models."
In contrast to its rival GPT-4, DeepMind's Gemini will be able to handle any data or job without the need for specialised models. Once fully developed, Gemini will have the capacity to provide more imaginative replies as well by producing 'off-script' material that is not constrained by its training data and is based on the structures it learned during training.
Despite the fact that GPT-4 technology can presently analyse pictures, audio, text, and video, Gemini can only produce text, which means that it will be less prepared to handle complex demands.
However, this is not to imply that ChatGPT or other chatbots don't have a chance. Since GPT-4 is currently accessible to the general public, its output will be more accurate and dependable than prompts produced by Gemini.
