--- type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/39503697.md" description: "🔥 Demis Hassabis directly refutes "the Singularity is here": the real divergence is not in speed, but in direction.Emily Chang directly asked the question everyone is discussing:Has Elon Musk said we've already entered the Singularity?Demis Hassabis's answer, without any hesitation:"No, I think that's very premature."This is not an ordinary opinion.This is the person behind AlphaGo.The person who led the key breakthrough research on Transformer.One of the core architects of modern AI infrastructure.The same data, different judgments.Elon is looking at the trend, saying "the moment has arrived."Hassabis is looking at the structure, saying "not yet."This in itself is enough to make one stop and rethink.But what's truly worth paying attention to is not this divergence.It's his next sentence:"We invented about 90% of the key breakthroughs in the modern AI industry."The meaning of this sentence is very direct:The foundation on which the entire industry is burning money to scale today—large models, Transformer, reinforcement learning—almost all come from the same place: DeepMind.That is to say:Right now everyone is competing on "scale,"but they are actually competing on the same architecture.The question is—Does this architecture have a ceiling?Hassabis's answer is: Yes.Ilya Sutskever once said we are "back in the research era."Hassabis corrected him on the spot:"I don't think we ever left."This sentence actually defines the most critical dividing line for the next few years.On one side:Believing that as long as we keep piling on computing power, adding parameters, and pushing scale, we can approach general intelligence.On the other side:Believing that scale is useful, but ultimately we must rely on new breakthroughs.The reality is, the past few years have proven that scaling is indeed effective.But the problem is:Even the people who originally designed these systems are reminding us—it's not an infinitely extendable path.And Hassabis belongs to that very rare type of person:Not only did he participate in "the establishment of this system,"but he also understands its boundaries more clearly.He also said a very key sentence:"If future breakthroughs are needed, I'd bet on us to deliver them."This is not an expression of confidence, but a continuation of a historical trajectory.When a laboratory has built most of the infrastructure,it's not a radical assumption that the next generation of breakthroughs will continue to emerge from there.So what is the market focusing on now?User numbersRevenueRelease cadenceBut if the underlying architecture is approaching its ceiling, the significance of these metrics will rapidly decline.What truly determines long-term victory is not:Who can use the current system more to its extreme,but:Who can create the next system.In other words:What everyone sees now is a "scaling race,"but what truly determines the future is an "invention race."And what Hassabis represents is precisely the latter.So the question becomes very simple:When all companies are optimizing the existing system,who is preparing to replace it?Are you more inclined to believe that the breakthroughs in AI in the next few years will come from continued scaling, or from entirely new architectures?" datetime: "2026-03-25T06:59:12.000Z" locales: - [en](https://longbridge.com/en/topics/39503697.md) - [zh-CN](https://longbridge.com/zh-CN/topics/39503697.md) - [zh-HK](https://longbridge.com/zh-HK/topics/39503697.md) author: "[辰逸](https://longbridge.com/en/profiles/16318663.md)" --- # 🔥 Demis Hassabis directly refutes "the Singul… ### Related Stocks - [DEEP.US](https://longbridge.com/en/quote/DEEP.US.md) - [MIND.US](https://longbridge.com/en/quote/MIND.US.md) - [TSLA.US](https://longbridge.com/en/quote/TSLA.US.md)