A British artificial intelligence startup has made headlines by outperforming many human experts in a global forecasting competition.
Mantic AI Breaks Into Top 10
Mantic AI, a UK-based startup co-founded by a former Google DeepMind researcher, ranked eighth in the prestigious Metaculus Cup—a forecasting challenge where participants must predict the outcomes of real-world events.
The competition required forecasts on 60 future events over the summer. These included high-profile political situations, such as:
- A fallout between Donald Trump and Elon Musk
- Whether Kemi Badenoch would lose her position in the UK Conservative Party leadership
Despite not taking the top spot, Mantic AI’s performance surpassed many human participants, making it one of the highest-ranked AI teams in the contest.
AI’s Forecasting Power Surprises Experts
Experts were impressed with Mantic AI’s accuracy. Deger Turan, CEO of Metaculus, called the result “impressive.”
Professional forecaster Ben Shindel, who was outscored by several AI models, admitted:
“It’s certainly a weird feeling to be outdone by several bots… We’ve come a long way from last year when the best bot ranked around 30.”
Before the competition, it was assumed AI bots would only reach 40% of the top human scores. But Mantic achieved over 80%, closing the gap significantly.
How Mantic AI Works
Mantic doesn’t rely on a single AI model. Instead, it breaks down prediction tasks and assigns them to different machine learning models based on their individual strengths—including models from OpenAI, Google, and DeepSeek.
Co-founder Toby Shevlane explained:
“Some say LLMs just regurgitate their training data, but you can’t predict the future like that. It requires genuine reasoning.”
He added that Mantic’s AI often made more original predictions than human forecasters, who tend to cluster around average guesses.
Is Artificial General Intelligence (AGI) Next?
The rapid improvement in AI forecasting raises questions about how close we are to Artificial General Intelligence (AGI)—AI with the ability to think and reason like a human.
According to a Google DeepMind research paper, AGI could arrive as early as 2030.
💡 What is AGI?
- Artificial Intelligence (AI) today is narrow and task-specific.
- AGI, on the other hand, would possess general-purpose intelligence, enabling it to understand, learn, and solve problems across any domain, just like a human mind.
Mantic AI’s performance may be an early signal that such a future is not far off.
Key Takeaways
- Mantic AI ranked 8th in the global Metaculus Cup forecasting contest
- Outperformed many human experts with 80% of top human scores
- Experts call it a “major leap” toward human-level reasoning in machines
- Fuels speculation that AGI could arrive by 2030
