Singapore FM Vivian Balakrishnan warns against overusing AI after building Claude AI brain

Singapore Foreign Minister Vivian Balakrishnan has warned that human intelligence still remains critical even as AI systems become more powerful and widely adopted. Speaking at the AI Engineer conference in Singapore, Balakrishnan cautioned against relying too heavily on large language models (LLMs) for solving every problem.

“We should beware of just trying to throw every problem, and every step in a solution, at a large-language model,” he said while addressing engineers, startup founders and technology operators.

According to him, traditional AI systems programmed by humans still remain important because large AI models come with limitations, including high costs and massive computing requirements.

He also pointed out that the human brain remains significantly more efficient than many modern AI systems.

“The human brain is often far more efficient than the energy-gobbling systems which we have today,” he said.

He says AI works best alongside human expertise

Balakrishnan stressed that the real value of AI comes when it is used to enhance human expertise rather than replace it.

“People who know their jobs and are empowered by these tools, that’s how you create real value for society and for the economy,” he said.

The minister has also built his own AI assistant

Interestingly, Balakrishnan is not just talking about AI from the sidelines. Last month, he published NanoClaw — a self-hosted personal AI assistant built using Anthropic’s Claude models. According to Balakrishnan, the assistant helps him with research, updates and drafting speeches and briefing notes.

The project was published on GitHub on April 21, 2026, allowing others to replicate the setup themselves.

At the conference, he also explained how he built the Claude-powered agent on the NanoClaw platform using the Raspberry Pi 5 setup.

The AI assistant runs on a Raspberry Pi 5, a compact single-board computer often used by developers and hobbyists for experimental projects.

The system also includes persistent memory and integrations with platforms such as WhatsApp, Telegram, Slack and Discord. It can process voice notes and even handle scheduled recurring tasks.

“AI agents have crossed a threshold”

Speaking about the rapid progress of AI tools, Balakrishnan said AI agents are becoming genuinely useful in daily life rather than remaining just technology demonstrations.

“AI agents have crossed a threshold I didn’t expect so soon. Not just impressive demos, but practical tools for daily use,” he said.

According to him, falling costs and easier access to AI tools now mean that more people should start experimenting with the technology themselves.

“Because the barriers for entry have come down so dramatically, everyone should embark on their personal experiments,” he said.