Is DeepSeek the future, or a mere stepping stone? AI technologists react

So far, 2025 has been quite a year for AI.

Just days after the founders of OpenAI Oracle and Softbank hyped a $500 Billion-with-a-B AI Infrastructure project called The Stargate Project at a post-inauguration White House press conference, a Chinese open-source AI model called DeepSeek R1 started to overshadow OpenAI’s ChatGPT. Before the month was out, DeepSeek’s “thinking model” app beat out OpenAI’s ChatGPT in US app stores, taking the No. 1 spot.

It caused just a bit of pandemonium as OpenAI chipmaker NVIDIA’s market value took a $600 billion nosedive in response. 

It wasn’t the app store downloads that disrupted OpenAI and its eye-wateringly expensive $500 billion announcement. It was DeepSeek’s claims that their model had developed and trained for the shockingly low cost of $5.6 million, using far fewer chips than anyone in the West thought possible. To put it into perspective, ChatGPT o1 costs between $100 to 200 million and requires about 30,000 chips to train. 

“If DeepSeek’s breakthrough proves valid, it could mark a pivotal moment: one of the key constraints on model development may have been removed,” Patrick Callahan, head of data analytics at Labware, an enterprise laboratory platform, and a member of the Delaware Artificial Intelligence Commission, told Technical.ly. 

The $5.6 million claim changes everything in terms of AI accessibility if true, even taking into account that DeepSink’s parent company, High Flyer, invested hundreds of millions in the company’s hardware spend since it launched in 2023. Even if the costs are higher than initially reported, it uses fewer GPU chips and requires less energy than ChatGPT, making it far more efficient and environmentally friendly. 

It’s also open source, unlike Western models like ChatGPT, so developers can study, use and modify it for free, making it far more accessible. Removing the barrier of paying for a license alone is potentially a big change, as competitors realize they need to go open source to stay in the game.

So DeepSeek appears to have advantages in the global AI race, but is the DeepSeek hype real? Is the West overreacting or underreacting? We asked a couple of mid-Atlantic region AI experts for their thoughts.

‘AI is already shaping society’

Callahan works closely with AI and follows its lightning-fast evolution. He says that, rather than jumping on the hype train as soon as news drops, people should take a breath.

“I don’t remember who said it first — I firmly believe people often overreact in the short term and underreact in the long term,” Callahan said. “This principle has served me well: it’s usually wise to pause, gather information, and discern what’s actually true. Complex situations often contain many nuances, so taking the time to understand all the details is essential. This is one of those moments.”

There’s no doubt, however, that the DeepSeek news made a big impact, he says; it made international headlines, moved stock markets, and sparked global concern. 

“[It] shows how profoundly AI is already shaping society — especially considering that AI barely made the news a few years ago,” Callahan said.

‘DeepSeek is a stepping stone’

Matthew Saponaro is the founder of the AI Whoo School, which teaches elementary and middle school-aged students technology skills using AI, as well as soft skills. In the early days of AI Whoo, one of the services it offered was helping facilities and parks ethically capture data to help them better serve their guests. He’s experienced the speed of AI’s evolution in his work firsthand.

“This reminds me a lot of when I offered our park assessment services,” Saponaro told Technical.ly.

He recalls specifically purchasing an expensive machine equipped with high-end GPUs to process the video footage during park assessments around 2019. Just two years later he was able to adapt the algorithms, with the same level of accuracy, to run on the phone hardware — leaving the expensive machine to collect dust. 

“Similarly, DeepSeek is only a stepping stone,” Saponaro said. “We, as a society, will eventually have embeddable generative AI with the same capabilities accessible on our phones.” 

Who will ultimately deliver that embeddable generative AI to your pocket is the question. The United States wants to be the global leader in AI technology. So does China, and the West is quickly learning that China is potentially more advanced than many realized, as massive “smart cities” like Chongqing and Chengdu change the landscape.

“When I went to China last year, I saw the smart cities first hand, from robot food delivery drivers to robot receptionists in the hotels,” said Saponaro. “The technology will just keep advancing and will become more practical.”

We need more critical and creative thinkers

When the Stargate announcement happened in January, Callahan was already bracing for imminent incoming changes. “While I couldn’t be more excited about the possibilities [The Stargate Project] brings, I’m also concerned about the rapid pace of change and its implications for society,” Callahan said. “It’s a transformative moment that I don’t think many people fully grasp yet and I don’t think there is any slowing down.”

Now with global players demonstrating that they’re just as capable as US developers, despite ongoing efforts to block China from accessing US chips, the race is on.

Fears of an AI-dominated dystopia may be exacerbated by the speed of AI evolution, but Saponaro says that, like in China’s increasingly automated smart cities, much of the technology will be there to meet mundane everyday needs — and that requires the input of humans who understand more than lines of code.

“This change is why we need more critical and creative thinkers,” Saponaro said. “AI, by definition, is unable to achieve these human aspects.”