AGI Still 30 Years Away, Claim AI Startup Founders
"Maybe for me, that would be around 2045," stated Ege Erdil when asked about artificial general intelligence (AGI) timelines, expressing a view that stands in stark contrast to the accelerated predictions dominating Silicon Valley's AI discourse.
Erdil and his co-founder Tamay Besiroglu, who recently launched Mechanize after leading AI research organization Epoch, represent a surprising counter-narrative in the AI space with their significantly longer AGI timelines, writes End of Miles.
Why the 30-year wait?
While many prominent AI organizations and researchers project AGI within 2-5 years, the Mechanize founders present compelling evidence for their longer forecast. They point to fundamental constraints in computing infrastructure and the diminishing returns of scaling up AI systems.
"We've gone through maybe 9 or 10 orders of magnitude of compute, and we got various capabilities that were unlocked. And we're getting to a level where it's becoming harder and harder to scale up compute." Tamay Besiroglu
The former Epoch AI researchers have analyzed specific constraints like energy availability and GPU production capabilities. Their calculations suggest only three to four orders of magnitude of scaling remain before the industry hits significant physical and economic limitations.
This ceiling would represent "a pretty sizable fraction or a non-trivial fraction of world output on just building up data centers, energy infrastructure, fabs, and so on," according to the Stanford-educated economist.
Reasoning doesn't equal intelligence
Erdil, a computer scientist by training, argues that current AI achievements are impressive but narrow when viewed in context. He dismisses claims that current reasoning models are close to general intelligence.
"If you compare it to a random job in the economy, I agree they are better at doing coding tasks that will be involved in R&D compared to a random job in the economy. But in absolute terms, I don't think they're that good." Ege Erdil
The entrepreneurs emphasize what they call "Moravec's paradox" — the observation that AI systems excel at tasks humans find difficult (like abstract reasoning) while struggling with capabilities that evolved over millions of years and appear simple to us.
"These things that we find very impressive when humans are able to do it, we should expect that AI systems are able to make a lot more progress on," notes Besiroglu. "But we shouldn't update too strongly about just their general competence."
Challenging Silicon Valley orthodoxy
The Mechanize founders' perspective runs counter to pronouncements from figures like OpenAI's leadership, who have suggested AGI could arrive within just a few years. Their assessment applies sophisticated economic and technological analysis to question the prevailing narrative.
When asked what would change their minds, the AI researchers set a high bar.
"Something that would reveal its ability to do very long context things, use multimodal capabilities in a meaningful way, and integrate that with reasoning and other types of systems. And also agency and being able to take action over a long horizon and accomplish some tasks that takes very long for humans to do." Tamay Besiroglu
Their skepticism comes despite significant advances in AI reasoning capabilities. The computational demands of creating truly agentic systems that can operate autonomously across domains remains their key concern.
The business paradox
Interestingly, despite their extended AGI timeline, the researchers remain bullish on AI's economic impact. Their company, Mechanize, is dedicated to "fully automating work" — suggesting they see tremendous value in narrow AI applications long before AGI arrives.
This nuanced position highlights a key distinction often lost in AI discussions: the difference between achieving true artificial general intelligence versus capturing economic value through more specialized AI systems.
"There's going to be this enormous increase in economic growth, which is going to mean enormous amounts of wealth, and incredible new products that you can't even imagine," explains Erdil, pointing to their vision for AI's impact even without full AGI.
Their company aims to accelerate automation broadly rather than pursue the more science-fiction concept of human-level AI that dominates public imagination.
Why it matters now
The timeline debate has profound implications for research priorities, investment strategies, and policy decisions. If the Mechanize founders are correct, the industry focus might better shift toward practical applications rather than AGI alignment concerns.
For policymakers wrestling with AI regulation, a longer timeline could provide breathing room to develop thoughtful frameworks rather than rushing emergency measures. For investors, it suggests caution around companies promising imminent AGI breakthroughs.
"The amount of influence you can have is still quite low," Erdil notes about attempts to shape AGI development decades in advance. "It might be orders of magnitude greater than it would have been 2000 years ago and still be quite low."
Whether their prediction proves accurate or not, the Mechanize founders' data-driven challenge to Silicon Valley's AGI optimism represents a significant counterpoint in the ongoing debate about humanity's AI future.
