DeepSeek’s rise raises concerns over Chinese AI dominance in the West ...Middle East

You may be surprised to see a new name atop the charts on Apple’s app store this week. DeepSeek, an AI app developed by a Chinese hedge fund manager named Liang Wenfeng, has beaten out ChatGPT to become the top free app on Apple’s store.

The stellar performance of DeepSeek, and the AI model called R1 that underpins its chatbot, has captivated users worldwide, helping propel it to the top of download charts.

“They were super-efficient in building the model. Instead of using some 16,000 chips to train the chatbot, they used 2,000. That’s impressive. Silicon Valley must be in a panic,” said Carissa Véliz, an AI philosopher at the University of Oxford.

The success of DeepSeek’s model – which is smaller and nimbler than many of the American competitors dominating the early AI race – has surprised many. This is largely because the prevailing assumption has been that bigger is better when it comes to technology, and the more hardware muscle behind an AI system, the better it would perform.

“Since the release of ChatGPT, American AI labs have been betting the house that bigger models means better models,” said Harry Law, an AI researcher at the University of Cambridge. “This is still true, but DeepSeek-R1 confirms that there is another option open to developers: instead of using more compute to make the underlying model bigger, they can spend it on improving the quality of outputs as they happen. This is the difference between what researchers call ‘training’ compute and ‘inference’ compute.”

The secret sauce behind DeepSeek’s R1 model is that it mimics how humans think out loud – a method called “chain of thought” within the industry. “At each stage it picks the best option before proceeding on to the next, which results in a much stronger final output,” said Law. “It’s the difference between writing the first thing that comes into your head during an exam, and planning your essay before you begin.”

Chain of thought reasoning isn’t new and is used in OpenAI’s models. But there’s a difference between the US and Chinese approach, said Law. “DeepSeek found a way to make the process of building the underlying model much cheaper,” he said. “Although the $5m price tag dramatically understates the true cost by ignoring spending on labour and hardware, the model still represents a major improvement in efficiency.”

While efficiency may be seen as a way to lower energy use, it could actually push it up, explained Law. “Because efficiency tends to increase rather than decrease consumption, it’s likely that more spending on chips is coming — not less,” he said.

For that reason, Law isn’t convinced that Chinese supremacy in the AI space will last long – not least because the US has the advantage of access to the computer chips, called GPUs, that power AI models. One of Joe Biden’s last actions before handing over the US presidency to Donald Trump was to institute export limits on those GPUs to scores of countries, while outright limiting access to the most powerful chips to a handful of countries, including China.

CIA joins growing support for Covid 'lab leak' theory – but the real story is in Germany

“It shows how limiting situations, no access to US chips, in this case, can catalyse innovation – which has implications for regulation,” said Véliz.

That has implications for the wider world of AI, too. “Factoring in improvements in efficiency, the next generation of US models may actually be more capable than many expect,” he said. “Compute is after all king, and America has the most chips. But as R1 shows, America may lead – but China will not be far behind.”

Concerns have been raised about the Chinese origin of DeepSeek, particularly by anti-China hawks in the United States who have just spent months fighting to get TikTok, a comparatively innocuous social media app, banned. DeepSeek’s Chinese links are exposed in interesting ways: ask the model to explain what happened at Tiananmen Square and it refuses to answer. “While data and censorship are concerning – for US models too, in some ways – R1 is only an issue in so far as what it says about the future,” said Law.

The key question vexing AI watchers is whether we should be concerned that a Chinese-based model has toppled ChatGPT from the app store crown. “I think at this stage that seems overblown,” said Law. He points out that the ability to run the model locally – meaning you don’t need to access it through DeepSeek servers – can limit the data you give over to a Chinese company.

Read More Details
Finally We wish PressBee provided you with enough information of ( DeepSeek’s rise raises concerns over Chinese AI dominance in the West )

Also on site :

• 5 U.S. Towns With Major European Vibes
• Powerball winner who took home $167m is arrested after ‘kicking’ a cop the next day
• Rubin: Trump’s Gaza diplomacy has failed. Will he break with Netanyahu and reverse course?