Currently, global tech giants are vying to invest heavily in building massive infrastructure, believing that "great power can create miracles," and that increased investment will allow them to advance in artificial intelligence. To be at the forefront of the (AI) competition.
Against this backdrop, a debate is brewing in Silicon Valley: where exactly will the "scaling laws" lead AI technology?
Demis Hassabis, CEO of DeepMind, Google's AI research company, recently stated his position on this issue . DeepMind recently released the Gemini 3 AI model, which has received widespread praise.
Last week, Hassabis said at an AI summit in San Francisco: "We must push the current system to its limits in terms of scaling up, because it will eventually become a key component of the Artificial General Intelligence (AGI) system, and may even constitute the entire AI system."
Artificial general intelligence (AGI) is currently still in the theoretical stage, and it refers to artificial intelligence with reasoning abilities equivalent to humans. AGI is the ultimate goal pursued by all leading artificial intelligence companies, which is why they invest huge sums of money in infrastructure and talent.
The core logic of the AI scale law is that the more data you feed to an AI model and the more computing power you invest, the higher the intelligence level of the model will be.
Hassabis believes that the law of scale alone may be enough for the entire industry to achieve AGI, but he also speculates that one or two other technological breakthroughs may be needed to achieve this goal .
Simply relying on scale has obvious limitations. On the one hand, the amount of publicly available data is limited; on the other hand, increasing computing power means building more data centers. This is not only costly, but also puts enormous pressure on the environment.
Some AI observers also worry that the huge investments made by AI companies behind large models in scaling up will show a trend of diminishing returns.
Some tech leaders, including Meta's chief AI scientist, Yang Likun, believe that the AI industry cannot rely solely on the law of scale and needs to explore an alternative development path. Yang Likun recently announced that he will be leaving Meta to found an AI company.
In April of this year, Yang Likun stated at the National University of Singapore (NUS): "Most of the really interesting problems perform extremely poorly under the law of scale. You can't take it for granted that more data and more computing power mean smarter artificial intelligence."
Ilya Sutskever, co-founder and former chief scientist of OpenAI and a key creator of ChatGPT, recently challenged the conventional wisdom that scaling may be a key roadmap for AI progress.
He stated that the AI industry should move beyond simple "scale" accumulation and return to a "research era" that emphasizes fundamental innovation.
(Article source: CLS)
