When does AGI arrive?
The most consequential timeline question in technology. Whether artificial general intelligence — AI that can do any intellectual task a human can — arrives in 4 years or 40 changes everything about how you invest, hire, and plan your business. The people building these systems disagree wildly on the answer.
Where They Stand
Before 2030
The CEOs of OpenAI, Anthropic, and Google DeepMind all believe AGI is just a few years away. Sam Altman has repeatedly said AGI could arrive by 2025-2027 and that it will be "the most transformative technology in human history." Dario Amodei published a long essay in October 2024 describing a world transformed by "powerful AI" within 2-3 years, covering biology, economics, and governance. Demis Hassabis has stated AGI could be achieved within a decade (from 2023) and possibly much sooner, pointing to the rapid progress from AlphaFold to Gemini. All three are investing billions on the assumption this timeline is real. Worth noting: they also have financial incentives to hype the timeline — their companies are valued on AGI being close.
2030–2050
The "it's coming but not tomorrow" camp. Andrew Ng, who co-founded Google Brain and Coursera, has consistently pushed back on hype cycles, arguing that while AI progress is real, AGI timelines are routinely underestimated by optimists. He emphasises that current LLMs still lack fundamental reasoning capabilities. Sundar Pichai has acknowledged AI's transformative potential but frames it in terms of decades of deployment, not years. Satya Nadella takes a pragmatic enterprise view — AI will gradually augment every job and business process, but the "singularity moment" framing is unhelpful for actual planning. These voices tend to focus more on narrow AI delivering real value now rather than speculating on AGI dates.
Much later or never as defined
Yann LeCun, Meta's Chief AI Scientist and a Turing Award winner, has been the most vocal sceptic of near-term AGI. He argues that current LLMs are fundamentally limited — they can't plan, reason about the physical world, or learn efficiently the way humans do. He believes entirely new architectures (what he calls "world models") are needed, and those are years or decades of research away. François Chollet, the creator of Keras and the ARC benchmark for measuring general intelligence, has demonstrated that LLMs fail badly at novel reasoning tasks. His position is that scaling alone won't get us to AGI — and the current definition of AGI is so vague that the goalpost will keep moving. Both argue the AI industry is conflating "impressive pattern matching" with "general intelligence."
Doesn't matter — danger is NOW
Geoffrey Hinton, the "Godfather of AI" who left Google in 2023 to speak freely about risks, argues that debating AGI timelines misses the point. Current AI systems are already capable enough to cause serious harm through misinformation, autonomous weapons, and concentrated power — whether or not they meet some academic definition of "general" intelligence. Connor Leahy, CEO of Conjecture, takes an even stronger position: the systems we have today are already outpacing our ability to understand, align, and control them. Waiting for "true AGI" to start worrying is like waiting for a fire to reach your bedroom before buying a smoke detector. Their message: stop arguing about dates and start building safety infrastructure now.
Patrick's Take
Here's the thing nobody on stage will tell you: the exact date AGI arrives doesn't actually matter for your business. What matters is the rate of capability improvement — and that's already fast enough to disrupt you. I've trained hundreds of Malaysian business owners on AI, and the ones who get stuck are always the ones asking "when will AI be smart enough to affect me?" Brother, it already is. GPT-4 can draft your contracts. Claude can analyse your financial reports. Midjourney can replace your graphic designer for 80% of tasks. You don't need AGI for AI to reshape your competitive landscape — you need to look at what's possible TODAY and ask whether your competitors are already using it. The real insight from this debate is about the people making the predictions. The CEOs building AI say it's coming fast (they need your investment dollars and enterprise contracts). The researchers say it's further away (they understand the technical gaps). The safety people say the timeline doesn't matter (the risk is already here). My advice to Malaysian SMEs: plan for AI that's 2x more capable every 18 months. That's conservative enough to not over-invest, aggressive enough to not get caught sleeping. And for God's sake, stop waiting for the "perfect" AI — start using the imperfect one that already exists.
What This Means for Your Business
If AGI arrives before 2030, any business model built on human knowledge work alone is at risk — legal, accounting, consulting, content creation, and customer service will face massive margin compression. If it's 2030-2050, you have time to adapt but should be building AI capabilities now while talent is still affordable and tools are cheap. If the sceptics are right and AGI is decades away, narrow AI still transforms every industry — the question is whether you're using these tools or your competitor is. For Malaysian SMEs specifically: don't wait for AGI to hire your first AI-literate team member. The companies winning right now aren't using AGI — they're using ChatGPT, Claude, and automation tools better than everyone else in their niche.
What to Actually Worry About
Don't worry about robot overlords. Worry about your competitor who figured out how to use AI to do in 2 hours what takes your team 2 days. The practical risk isn't AGI arriving — it's the steady, compounding productivity gap between businesses that adopt AI tools and those that don't. In Malaysia, where labour costs are relatively low, the temptation is to think AI automation isn't urgent. That's exactly the complacency that will bite you. Start with one process, one tool, one team member trained up. Measure the impact. Then expand. The AGI timeline debate is fascinating dinner conversation — but your 2026 P&L doesn't care about 2035 predictions.
Featured Minds in This Debate
Sam Altman
CEO, OpenAI
Dario Amodei
CEO & Co-Founder, Anthropic
Demis Hassabis
CEO & Co-Founder, Google DeepMind
Andrew Ng
Founder & CEO, DeepLearning.AI / AI Fund
Sundar Pichai
CEO, Alphabet / Google
Satya Nadella
Chairman & CEO, Microsoft
Yann LeCun
VP & Chief AI Scientist, Meta
Francois Chollet
Software Engineer & AI Researcher, Google
Geoffrey Hinton
Professor Emeritus, University of Toronto
Connor Leahy
CEO, Conjecture
Last updated: 2026-03-26
←Back to AI Minds Directory