DeepMind CEO: Today's AI Can't Make Long-Term Coherent Plans

Demis Hassabis, CEO of Google DeepMind, addressed critical gaps in current AI systems at the India AI Impact Summit 2026. He highlighted the inability of AI to perform long-term coherent planning and to learn continually after being deployed. Hassabis also pointed out a major issue with consistency, where AI excels at complex tasks but fails at simpler versions of the same problems. The summit in New Delhi gathered global policymakers and experts to discuss AI's transformative potential for humanity.

Key Points: DeepMind CEO Demis Hassabis on AI's Key Limitations

  • Lacks long-term planning ability
  • Cannot learn continually after deployment
  • Shows inconsistent "jagged" intelligence
  • Needs improvement in contextual learning
3 min read

"Today's AI cannot make long-term coherent plans," says DeepMind CEO Demis Hassabis

At India AI Summit, Demis Hassabis says AI lacks long-term planning, continual learning, and consistency, highlighting gaps to true general intelligence.

"Today's AI cannot make long-term coherent plans," says DeepMind CEO Demis Hassabis
"Today's systems can get gold medals in the International Maths Olympiad... but sometimes can still make mistakes on elementary maths. - Demis Hassabis"

New Delhi, February 18

Founder and CEO Google DeepMind, Demis Hassabis, attended the India AI Impact Summit 2026 and shared his thoughts on AI Safety, Agentic AI, and various AI innovations.

"This Summit comes at a critical time as we start seeing more autonomous agentic AI systems," Demis Hassabis said, appreciating the government's efforts and willingness to inculcate AI in their systems.

Commenting on what AI systems need to build on more, he said, "When I look at the current systems and what's missing from them being a kind of general intelligence, I would say things like continual learning, so learning after they've been trained and put out into the world. In today's systems, we train them, we do various different types of training on them, and then they're kind of frozen and then put out into the world. But what you'd like is for those systems to continually learn online from experience to learn from the context they're in, maybe personalised to the situation and the task that you have for them, and today's systems don't do that."

Current models have shown difficulties in devising long-term plans. On this Hassabis said, "Also, they have difficulty with things like doing long-term coherent plans. They can plan over the short term, but over the longer term, as the way that we plan can plan over years, they don't really have that capability at the moment."

Consistency in processes and functionality is also an issue he highlighted. He said, "I think probably one of the biggest issues is what I would call consistency. So today's systems are kind of like jagged intelligences. They're very good at certain things, but they're very poor at other things, including sometimes the same things. So for example, today's systems can get gold medals in the International Maths Olympiad, really hard problems, but sometimes can still make mistakes on elementary maths if you pose the question in a certain way. A true general intelligence system shouldn't have that kind of jaggedness."

Demis Hassabis founded Deepmind in 2010 which was later acquired by Google.

The India AI Impact Summit 2026 is being held at Bharat Mandapam, New Delhi, from February 16-20, 2026.

The Summit has brought together government policymakers, industry AI experts, academicians, technology innovators and civil society from across the world at New Delhi to advance global discussions on artificial intelligence.

The India AI Impact Summit, the first global AI summit to be hosted in the Global South, aims to reflect on the transformative potential of, AI aligning with the national vision of "Sarvajana Hitaya, Sarvajana Sukhaya" (welfare for all, happiness for all) and global principle of AI for Humanity.

The Summit saw the participation of more than 110 countries, 30 International organizations, including about 20 HoS/HoG level participation, about 45 Ministers.

- ANI

Share this article:

Reader Comments

R
Rohit P
The long-term planning limitation is a huge deal for India. We have massive infrastructure projects, climate adaptation plans, and urban development that need 20-30 year vision. If AI can't help with that coherently, we still heavily rely on human expertise. Good that our government is engaging with these experts early.
A
Aditya G
Continual learning is the key. Our Indian education system also faces a similar challenge - we train students for exams and then the learning often stops. Maybe the solutions for AI and for our human systems have some parallels. Jai Hind!
S
Sarah B
As a tech professional in Bangalore, this is the most accurate assessment I've read. The hype is far ahead of the reality. We're integrating AI into our products, but the inconsistency Hassabis mentions creates real reliability issues. It's a powerful assistant, not an autonomous agent. Yet.
K
Karthik V
Respectfully, while the technical points are valid, focusing only on limitations from a Western lab perspective might miss grassroots innovation. Indian startups are already using AI for crop prediction and supply chain logistics with great success, even with current "jagged" systems. The summit should also highlight these practical applications.
M
Meera T
"Sarvajana Hitaya, Sarvajana Sukhaya" - this principle is perfect for guiding AI development. If the systems can't plan long-term or learn continually, how can they ensure welfare for all? The summit's focus on safety and humanity is crucial. Proud that India is leading this conversation for the Global South.

We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.

Leave a Comment

Minimum 50 characters 0/50