In case you haven't heard, there's a brand new AI star on the town: DeepSeekthe subsidiary of Hong Kong-based quantitative analytics firm High-Flyer Capital Management, sent shockwaves across Silicon Valley and the world earlier this week with the discharge of a brand new open-source big reasoning model, DeepSeek R1, which is OpenAI's strongest model available o1 – at a fraction of the fee to users and the corporate itself (when it comes to training).
While the launch of DeepSeek R1 has already reshuffled a consistently turbulent, fast-moving and highly competitive market for brand spanking new AI models, in recent months OpenAI has been competing with Anthropic and Google for essentially the most powerful proprietary models available, while meta platforms often with “close enough” to open source competitors – the difference this time is that the corporate behind the new model is predicated in China, the geopolitical “Enemy” of the United States, and whose technology sector had until this point been widely viewed as inferior to that of Silicon Valley.
Therefore, there is no such thing as a shortage of hand-wringing and existentialism on the a part of US and Western Bloc techies who’re suddenly doubting OpenAI and the massive tech corporations' general strategy of throwing extra money and more computing power (graphics processors, GPUs, often the powerful gaming chips) on the training AI models) to the issue of inventing ever more powerful models.
Still, some Western tech leaders have received a largely positive public response to DeepSeek's rapid rise.
Marc Andreessen, co-inventor of the groundbreaking web browser “Mosaic”, co-founder of the browser company Netscape and current general partner of famous Andreessen Horowitz (a16z) enterprise capital company, posted on X today: “Deepseek R1 is one of the crucial amazing and impressive breakthroughs I've ever seen – and, as open source, a profound gift to the world (robot emoji, salute emoji).”
Yann LeCun, chief AI scientist at Meta's Fundamental AI Research (FAIR) division, posted on his LinkedIn account:
DeepSeek has benefited from open research and open source (e.g. PyTorch and Meta's Llama).
They developed latest ideas and built them on other people's work.
Because their work is published and open source, everyone can profit from it.
That is the ability of open research and open source.”
And even Mark “Zuck” Zuckerberg, founder and CEO of Meta AI, appeared to wish to counter the rise of DeepSeek together with his own means Own post on Facebook guarantees that a new edition of Facebook's Llama open source AI model family can be “the leading state-of-the-art model” when released sometime this yr. As he put it:
“”
He even shared a graphic showing the two gigawatt data center above Manhattan mentioned in his post:
While he’s an advocate for open source AI, Zuck is clearly not convinced that DeepSeek's approach of optimizing efficiency while using far fewer GPUs than large labs is the suitable one for meta or for the long run of AI is.
But as US corporations raise and/or spend record amounts on latest AI infrastructure, which many experts say is rapidly losing value (because of hardware/chip and software advances), the query stays which vision of the long run will ultimately prevail dominant AI provider for the world. Or will there perhaps all the time be a mess of models, each with smaller market shares? Stay tuned as this competition gets tighter and tougher than ever before.
