Google is building a new AI system called GEMINI AI that will outperform Open AI's ChatGPT. At least that's what Demis Hassabis, CEO of Google's DeepMind, thinks about its upcoming Gemini AI. Hassabis says that Google's new AI system is revolutionary and is still in the developing phase. He added that it could take several months to complete the AI. Hassabis says the development cost of Gemini AI is estimated to be in the hundreds of millions. In other words, Google is putting a lot of resources into the AI system. That is, the company wants to beat ChatGPT at all costs.
What is GEMINI AI?
It's not just the money that's making Google think that the Gemini AI will beat ChatGPT, but it's all about the core of the AI system. This system is a next-generation AI architecture that will eventually replace PaLM 2. Let us tell you that PaLM 2 is the current AI model behind all of Google's AI services. It also includes Duet AI and Bard chatbots in workspace apps. Google is making Gemini AI more powerful than ChatGPT-4. OpenAI's GPT4 model is already a large language model that can generate text and images. This investment by Google is one of the key factors in Google's belief that Gemini AI can outperform ChatGPT's GPT 4 model.
Will Gemini AI enable the innovation of the future?
Google announced Gemini during the Google I/O conference. During the event, Google said that Gemini is fundamentally built to be highly efficient and enable future innovations in multi-model, tool, and API integration. According to Google, the Gemini AI will come in various sizes and capabilities. This means that it is flexible and can be changed as per the requirement.
Problem-solving and planning skills
Hassabis says that Google's engineers are currently using the technology of AlphaGo. Now, what is AlphaGo? AlphaGo is an AI that was powerful enough to beat the champion of the board game Go. Google is using the same technology that powered AlphaGo to power the Gemini AI. This means that the Gemini AI will be adept at problem-solving and planning.
On that note, AlphaGo was built using the Reinforcement Learning process. This process forces the software to make repeated attempts to solve a task. It can change its performance and feedback. At the high level, Hassabis says, you could see Gemini with the language capabilities of larger models with the same power as an AlphaGo-type system.
Google is planning to beat ChatGPT
The company believes that Google doesn't want to copy something that already exists, so offering performance similar to GPT 4 is not the main concern of Gemini AI. Instead, it is being made better.
Google will avoid repeating Bard's mistake
Bard Open AI's ChatGPT was a hastily answered answer. But in its haste, Google made the mistake of noting that its AI model was not ready. After all, Bard's factual mistake in the very first demo cost a company $100 billion in market value. Now the company would like to avoid repeating this mistake. Google has learned a lesson from this, and that is why the company is not in a hurry to enter Gemini AI.
(pc amarujala)
