I'm beyond excited to announce that technology expert and London-based founder of the
AnyModel.co
multi-model AI subscription service, Jamie Munro, has agreed to jump on board with me here at The Customer & The Constituent, to bring my readers the latest insights and commentary on the lightning-speed-moving freight train that is Artificial Intelligence.
He'll also offer a far wider perspective than anyone here in New Zealand, being at the helm of developments in numerous AI models globally. Read his full bio here.
In his first column, he provides us with some insights into the phenomenon that has been the new Chinese AI model, DeepSeek, since it took the global market by storm this past couple of weeks.
Over to Jamie:
China's DeepSeek AI: What's the Fuss All About?
"DeepSeek R1 is the Chinese AI model that's been causing a meltdown in Artificial Intelligence circles over the last week or so.
"The chatbot has only been available for literally, just days, but it's already triggered mass panic among American AI companies and sent a shockwave through the stock market.
"The reason? A Chinese start-up with less than 10 employees and very limited funding was able to rival the performance of the best models from tech giants like OpenAI and Google - who are now scrambling to catch up.
"And the kicker? DeepSeek is only charging five percent of the cost that the big players are demanding.
"As well as clever computer programming, development of modern AI models requires thousands of hours of 'training' using expensive computer chips known as GPUs. Thanks to the ongoing AI arms race in Silicon Valley, these chips are in very short supply.
"Whilst that's great for the manufacturer, Nvidia's stock price, it makes it very difficult for less-established players to get their hands on this all-important component.
"In fact, so critical is the supply of these chips that tech giants are poaching employees from rivals by boasting about the number of GPUs they have access to. And the U.S. Government has imposed export controls on the most advanced chips, putting them in the same category as advanced weapons.
"Meantime, OpenAI, the current market leader, is estimated to have spent over $10 billion training their latest 'reasoning' model. By comparison, DeepSeek spent just $6 million, and due to aforementioned export controls, they only had access only to older, less-advanced GPUs. Despite this, the DeepSeek team was able to develop an innovative training approach, with the results speaking for themselves.
"Even OpenAI CEO, Sam Altman was impressed, posting:
"Deepseek's R1 is an impressive model, particularly around what they are offering for the price. It's invigorating to have a new competitor!"
"Despite the present doom and gloom from the analysts of the AI giants, the AI wars are far from over and most of DeepSeek's innovations have been published under an open-source licence. This means that anyone is free to take DeepSeek's work and use or modify it in any way they see fit.
"I'm hopeful that these developments will continue to advance AI performance and drive down the prices across the board, making AI more accessible to more businesses and individuals."
The Ed: More cutting-edge insights into the fast-moving tech world from our new AI expert columnist coming soon. Starting with, next issue . . . (DRUM ROLL): A Special Feature interview on Artificial Intelligence in government.
