- AI researchers at NeurIPS 2025 say the current scaling approach has reached its limit
- Despite Gemini 3’s strong performance, experts argue that LLMs still cannot reason or understand cause and effect.
- AGI remains far away without a fundamental overhaul of how AI is built and trained
The recent successes of AI models like Gemini 3 don’t obscure the more sobering message that emerged this week at the NeurIPS 2025 AI conference: that we could be building AI skyscrapers on intellectual sand.
As Google celebrated the performance jump of its latest model, researchers at the world’s largest AI conference issued a warning: No matter how impressive the current crop of large language models may seem, the dream of artificial general intelligence is moving further and further away unless the field reconsiders its entire foundations.
Everyone agreed that simply scaling current transformer models, giving them more data, more GPUs, and more training time, no longer generates significant returns. The big jump from GPT-3 to GPT-4 is increasingly seen as a one-off; Since then, everything has been less about breaking glass ceilings than about simply polishing the glass.
That’s a problem not just for researchers, but for everyone who is sold the idea that AGI is just around the corner. The truth, according to the scientists attending this year, is much less cinematic. What we have built are highly articulated pattern matchers. They are good at producing answers that sound good. But looking smart and being smart are two very different things, and NeurIPS made it clear that the gap is not closing.
The technical term that is being spread is “climbing wall.” This is the idea that the current approach—training larger and larger models on larger and larger data sets—runs into both physical and cognitive limits. We are running out of high-quality human data. We are burning huge amounts of electricity for small marginal gains. And perhaps most worrying is that the models still make the kind of mistakes that no one wants their doctor, their pilot, or their scientific laboratory to make.
It’s not that Gemini 3 hasn’t wowed people. And Google invested resources into optimizing the model architecture and training techniques, rather than simply throwing more hardware at the problem, which makes it work incredibly well. But Gemini 3’s dominance only underscored the problem. It’s still based on the same architecture that everyone now silently admits is not designed to scale to general intelligence: it’s simply the best version of a fundamentally limited system.
Manage expectations
Among the most discussed alternatives were neurosymbolic architectures. These are hybrid systems that combine the statistical pattern recognition of deep learning with the structured logic of ancient symbolic AI.
Others advocated for “world models” that mimic how humans internally simulate cause and effect. If you ask one of today’s chatbots what happens if you drop a plate, they might write something poetic. But you have no internal sense of physics or any real understanding of what happens next.
The proposals are not about making chatbots more charming; it’s about making AI systems trustworthy in environments where it matters. The idea of AGI has become a marketing term and fundraising pitch. But if the smartest people in the room say we’re still missing critical ingredients, it may be time to recalibrate expectations.
NeurIPS 2025 might be remembered not for what it showed, but for admitting that the industry’s current trajectory is impressively profitable but intellectually stagnant. To go further, we will have to abandon the idea that more is always better.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds. Be sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp also.
The best business laptops for every budget
