Nobody here talks about AGI

In San Francisco, every dinner conversation eventually lands on AGI timelines. When will it arrive? Will it be OpenAI or Anthropic? Should we be terrified or ecstatic? The discourse is breathless, philosophical, and almost religious in its intensity. In Singapore, nobody is having that conversation. They're too busy building invoice processors, logistics optimizers, and compliance automation tools. The gap between AI discourse and AI deployment isn't just a difference in priorities. It's a cultural divide, and the builders are winning.

The two conversations

If you spend any time in Silicon Valley's AI circles, or even just on X, the dominant narrative is about timelines. When does AGI arrive? What happens after? How do we align superintelligence? OpenAI CEO Sam Altman wrote in January 2025 that the company was "now confident we know how to build AGI." By August of that same year, he reversed course entirely, calling AGI "not a super useful term." The goalpost didn't just move, it evaporated. Meanwhile, in Asia, the conversation is entirely different. A Forrester report on 2026 predictions for the Asia Pacific region noted that business and technology leaders in the region are focused on "strategies that reflect a pragmatic response to the region's complex economic, regulatory, and cultural landscape," moving beyond experimentation to scale AI in core functions. BCG's 2025 survey of over 4,500 employees across nine APAC markets found that adoption is running ahead of the global average, with employees using generative AI tools faster and more enthusiastically than their global peers. The question in Singapore isn't "when does AGI arrive?" It's "how do I use this to ship faster?"

This isn't anti-intellectual

It would be easy to frame the pragmatism as intellectual laziness, as if the builders just don't care about the big questions. That's wrong. This is a different theory of value. Philosophical speculation about machine consciousness and existential risk is genuinely important work. But it doesn't ship products, reduce logistics costs, or help an SME in Jurong automate their accounts receivable. For most of the world's economy, the relevant question isn't whether we'll achieve artificial general intelligence by 2027. It's whether current models can reliably extract line items from a PDF invoice in Bahasa Melayu. The answer, by the way, is yes. And that's worth billions in unlocked productivity.

Singapore's unique position

Singapore is a particularly sharp lens for this phenomenon, not because it's the only pragmatic market, but because its constraints force clarity. The country is small. With a population under six million, the domestic market alone can't sustain bloated AI ventures chasing moonshots. Efficiency isn't a nice-to-have, it's a survival trait. Every AI deployment needs to justify itself against a tight labor market and high operating costs. The government is AI-positive but regulation-pragmatic. Singapore launched its National AI Strategy 2.0 in 2023, with a clear focus on harnessing AI "for the public good." In early 2026, the government rolled out the National AI Impact Programme to strengthen AI capabilities across enterprises and workers. It also became the first country to publish a Model AI Governance Framework specifically for agentic AI. The approach isn't "move fast and break things," it's "move fast and govern things." The talent pool reflects this orientation. Singapore's tech workforce skews technical but business-minded. The startup ecosystem, ranked 4th globally in the 2025 Global Startup Ecosystem Index, is defined by what StartupBlink calls "bankable intellectual property and enterprise innovations," not speculative research labs. And the investment infrastructure backs it up. The government set aside S$150 million for the Enterprise Compute Initiative to help Singapore-based companies with AI transformation projects, and announced over S$1 billion for a National AI Research and Development Plan. DBS, one of Southeast Asia's largest banks, expanded its Spark GenAI programme to move SMEs from experimentation to deployment. Singtel launched AI.dea in April 2026 to accelerate AI adoption among small and medium enterprises. The pattern is consistent: identify a specific problem, apply current AI capabilities, measure the outcome, scale what works.

Intelligence is a commodity, so what matters is taste

Here's the thesis that connects all of this: intelligence is becoming cheap. Gartner predicts that by 2030, performing inference on a trillion-parameter model will cost LLM providers over 90% less than in 2025. Nvidia's Blackwell platform is already delivering 4x to 10x reductions in cost per token for production deployments. LLMs will become up to 100 times more cost efficient than the first models from 2022. When intelligence is a commodity, the differentiator isn't having access to it. Everyone will. The differentiator is knowing what to build with it. This is where the AGI discourse becomes not just irrelevant but actively distracting. Every hour spent debating whether GPT-7 will be "truly" intelligent is an hour not spent figuring out which specific workflow in your supply chain could be 40% faster with a well-designed agent. The builders understand this intuitively. They're not waiting for some theoretical threshold to be crossed. They're compounding, shipping iteration after iteration, learning what works in production, building institutional knowledge about where AI creates real value.

The counter-argument, and why it's half right

To be fair, AGI discourse isn't all hot air. The philosophical questions shape policy, and policy shapes funding. The EU AI Act, various national AI safety frameworks, and the entire alignment research field exist because people took the big questions seriously. If everyone adopted pure pragmatism and nobody thought about long-term risks, we'd probably end up in a worse place. But there's a difference between taking the questions seriously and obsessing over them. The discourse in its current form, the timeline predictions, the benchmark races, the breathless announcements about "achieving AGI internally," has become more performance than substance. When Sam Altman himself walks back the term, you know the hype cycle has eaten its own tail. The useful middle ground is simple: build with what exists today, stay informed about where the technology is heading, and don't let speculation about the future paralyze action in the present.

The builder's advantage

There's a compounding effect that the AGI-obsessed crowd misses entirely. Every company that deploys an AI-powered logistics optimizer today learns something. They learn about failure modes, about user expectations, about integration pain points, about which metrics actually move. That knowledge compounds. By the time the next generation of models arrives, these companies won't be starting from scratch. They'll be iterating from a position of deep operational understanding. The companies waiting for AGI, on the other hand, will still be waiting. Or they'll suddenly try to adopt AI at scale with zero institutional knowledge, which is roughly as effective as trying to run a marathon without having trained. This is the real divide. It's not between those who believe in AGI and those who don't. It's between those who are building now and those who are philosophizing about building later. The Asia-Pacific pattern, and Singapore's embodiment of it, isn't just a cultural quirk. It's a competitive strategy. And it's working. While everyone debates, the people shipping are compounding.

References