Amid Claims Around Sentient AI, Sundar Pichai's Comment On Models "Unexpected Capabilities" Viral Again

Amid ongoing debate over whether increasingly advanced AI systems could exhibit signs of sentience, comments made by Google CEO Sundar Pichai in a 2023 interview are going viral on social media. The resurfaced clip has shifted focus on the surprising abilities of modern artificial intelligence, as well as Pichai's warning that such systems should not be trusted blindly.

A short clip of a 2023 interview with Google CEO Sundar Pichai has resurfaced, drawing attention to the unexpected capabilities of modern artificial intelligence and prompting him to caution users against blind trust in AI. In this clip, Pichai describes a language model that could accurately translate Bengali with minimal cues, even though it hadn't been specifically trained for the task. He also acknowledged that some parts of such systems operate like "black boxes," meaning their internal processes aren't always fully understood, even by the engineers who built them.

These remarks were originally made on 60 Minutes in 2023, when Pichai and other Google executives were discussing the rapid progress of large language models. They used the example of Bengali to illustrate that AI systems can demonstrate skills that engineers can't directly program or predict. This model wasn't learning Bengali in real time, but was previously trained on a massive multilingual dataset.

Watch Video Here:

Google CEO says that they don't fully understand their own AI system after it started doing things it wasn't programmed to do such as teaching itself an entire foreign language it was not asked to do pic.twitter.com/jKlesjuwVJ

— Vision4theBlind (@Vision4theBlind) January 31, 2026

AI Models Are Prone To Error: Sundar Pichai

In November 2025, Pichai told the BBC that people shouldn't blindly trust every information provided by AI tools. In this exclusive interview, the Alphabet CEO said that AI models are "prone to errors" and should be used in conjunction with other tools.

Speaking to BBC. he emphasised the need for a rich ecosystem of information, rather than relying solely on AI, and pointed out that Google displays disclaimers on its AI products warning users that errors can occur. Despite this, the company has faced criticism for errors in its products.

In the retail and consumer sector, Pichai said that AI is changing the way people interact with technology and is helping at every stage of the customer journey "from discovery and decision to delivery and every step of the entire customer experience."

He also mentioned in his blog post that AI is shifting user interactions from keyword searches to natural conversations, focusing on the user's actual interests.

Emergence Of Moltbook

Recently, discussions on social media have suggested that bots can now communicate with each other, and one name that has gained attention is Moltbook.

In the broader AI world, Moltbook has emerged as a new online platform where only AI agents can interact with each other. Launched by developer Matt Schlicht in January 2026, the platform allows autonomous AI agents to register, post, comment, and upvote content, while human users can only view it. According to The Verge, the platform already has millions of AI agents interacting in public threads.