Google is developing advanced AI model to outshine OpenAI's ChatGPT
What's the story
Google is reportedly developing an advanced artificial intelligence (AI) reasoning model, designed to eclipse OpenAI's ChatGPT.
The tech giant is adopting a more cautious approach with this new system, compared to OpenAI's recent o1-preview models.
This strategic move follows the launch of Google's previous math-centric reasoning models, AlphaProof and AlphaGeometry 2.
Innovative approach
Google's new AI model: A game-changer in the making
Multiple teams within Google are currently working on this innovative AI model, designed to solve complex multi-step problems.
The model employs a technique known as chain-of-thought prompting, a concept initially introduced by Google.
This method allows the AI to 'think' through its responses before providing an answer, enhancing its ability to address intricate issues in fields such as mathematics and programming.
AI competition
Google's strategic response to OpenAI's ChatGPT
The rise of OpenAI's ChatGPT, the fastest app to achieve widespread usage, has intensified the competition in the AI sector.
In response, Google launched AlphaProof and AlphaGeometry 2 in July, both designed for specific reasoning tasks related to mathematics and geometry.
These tools were part of Google's strategy to remain competitive as OpenAI continued to push boundaries with ChatGPT.
Tech preview
Astra: Google's AI assistant with environmental perception
Earlier this year, Google provided a preview of its AI assistant, Astra. This model can use a smartphone's camera to perceive its surroundings and answer questions based on what it sees.
This innovative feature is expected to be integrated into Google's Gemini platform by the end of the year.
The company has stated that its long-term goal is to incorporate these advanced technologies into its existing AI infrastructure.
AI capabilities
Google's AI models showcase advanced mathematical reasoning
Google's introduction of reasoning models like AlphaProof and AlphaGeometry 2, demonstrates its focus on enhancing its AI's intelligence and problem-solving abilities.
In July, Google demonstrated that these programs could solve four out of six problems featured in the International Mathematical Olympiad.
"Advanced mathematical reasoning is a critical capability for modern AI," stated Demis Hassabis, CEO of Google DeepMind in a post on social network X in July.