Google Develops AI with Human-Like Reasoning Abilities

Google is working on creating artificial intelligence capable of reasoning like a human, according to Bloomberg, citing sources.

In recent months, several teams within the company have made progress in this area, the agency notes.

To develop this “thinking” AI, the tech giant aims to employ a technique known as “chain of thoughts,” journalists have learned. Before responding, the neural network pauses for a few seconds to consider a series of queries related to the main one, then synthesizes them to provide the best answer.

In mid-September, AI startup OpenAI introduced a new large language model, o1, trained with reinforcement learning to perform complex reasoning. The neural network is said to “think”—it can create a long internal chain of thoughts while analyzing a question, the company claims.

Some employees of Google’s AI division, DeepMind, have expressed concerns that the firm is lagging behind competitors, a Bloomberg source reported.

TrueMedia founder Oren Etzioni noted that the tech giant remains a formidable player in the AI field.

“From a technical standpoint, Google’s capabilities have always been top-notch. They are simply more conservative in their implementation. It’s a marathon, and anyone can win,” he added.

In July, DeepMind released two AI models focused on mathematics. Later, 200 employees of the division voiced dissatisfaction with the corporation’s involvement in military initiatives.

In August, Google unveiled a new line of Pixel 9 smartphones with the AI assistant Gemini built-in by default, the Gemini Live voice-mode chatbot, watches, headphones, and other AI-focused products.

That same month, the company released three new models of open-source generative AI, positioning them as “safer, more compact, and transparent” compared to others.