top of page

DeepSeek: Game changer or flash in the pan?

I’m sure you saw last week’s headline about the global stock market shock, caused by a Chinese AI company that didn’t even exist 18 months ago. Well, who are they – and what really happened? This week, Nerd News brings you up to speed with all things DeepSeek R1.


DeepSeek, a Chinese artificial intelligence company, recently unveiled its R1 model, an open-source AI system that hopes to rival leading models from companies like OpenAI and Meta.


DeepSeek R1 ‘thinks’ deeply about a question before answering and in early performances, seems to be comparable to these longstanding incumbents.


Notably, DeepSeek claimed to have achieved this with significantly lower costs by utilising less advanced GPU chips – challenging the prevailing belief that cutting-edge technology and substantial financial resources are essential for superior AI development.


The release of DeepSeek’s R1 model had an immediate and significant impact on the stock market, with many major US tech stocks experiencing a sharp decline.


Market darling Nvidia – the producer of computer chips until now considered essential to building massive large language models – saw its value drop by approximately $600 billion, the largest single-day loss in history for any company.


Other tech giants, including Microsoft and Alphabet, also experienced stock downturns.


This market reaction reflects investor concerns about the potential disruption that DeepSeek’s cost-effective AI model poses to established industry leaders.


In response to R1, tech leaders and policymakers have emphasised the need for accelerated AI development in the United States.


President Donald Trump described it as a wake-up call for Silicon Valley, underscoring the urgency for the US to maintain its competitive edge in AI technology.


The White House is currently assessing the national security implications of DeepSeek’s rapid proliferation, with some policymakers likening the event to a Sputnik moment, highlighting the critical importance of sustaining US leadership in AI innovation.


This follows closely on the heels of President Trump’s executive order titled “Removing Barriers to American Leadership in Artificial Intelligence” signed on 23 January 2025.

The order mandates the creation of an AI action plan within 180 days to sustain US AI leadership, focusing on human flourishing, economic competitiveness, and national security.


Concurrently, Trump announced a private-sector investment of up to $500 billion to fund some of the world’s largest AI infrastructure, aiming to outpace rival nations in this critical technology.


His announcement was made alongside key industry leaders, including Sam Altman (CEO of OpenAI), Larry Ellison (Chairman of Oracle), and Masayoshi Son (CEO of SoftBank) – underscoring a significant public-private partnership in advancing AI infrastructure.


Users testing DeepSeek R1 against OpenAI models quickly noticed a couple of stark differences.


When asked about Tiananmen Square, OpenAI provided historical context, while DeepSeek returned nothing or deflected the question. This is not surprising. AI models reflect the biases of their training data and the constraints of their creators, whether in China, the US, or elsewhere.


With bias identified as a major concern for many AI commentators, the question remains ‘Who gets to decide what an AI will and will not say, and what happens when those decisions are made by a government rather than by a company’?


Another survey suggested that DeepSeek seems to contain far less stringent guardrails than the larger American models when it comes to requests to design bombs or other antisocial prompts.


These aren’t the only questions set in train by the drop of DeepSeek R1...


Grab your popcorn: 5 things we are salivating to find out going forward



1. Did DeepSeek really build R1 for a fraction of the cost?


Reports suggest R1 was trained on less powerful hardware. However, does that translate to real cost savings? If so, what does that mean for the future economics of AI development? Or is it simply some great PR spin?


2. How does DeepSeek R1 compare to OpenAI, Anthropic and Meta models?


Open-source is great, but does R1 truly rival GPT-4 and Claude? We have seen flashy AI launches before, but how does it hold up under real-world use? Time will tell.


3. Just how safe is content uploaded to a China-based LLM?


Many feel the privacy and security concerns voiced around TikTok would apply equally to DeepSeek.


4. Will this trigger an AI arms race between China and the West?


Is DeepSeek’s success a Sputnik moment for AI, pushing the US and its allies to accelerate their own LLM development and framing AI as a political battleground for this century?


5. How will Silicon Valley and the stock market react next?


Nvidia, Microsoft and Google took a short-term hit, but will they counter with even bigger AI announcements? Was the market panic a precursor to a future shakeout of what many see as a highly overvalued tech sector?


Australia's new Chief Scientist

Also, congratulations to Professor Tony Haymet on his appointment as Australia’s Chief Scientist.


Haymet, an oceanographer will be Australia’s 10th Chief Scientist, following on from Cathy Foley.


To find out more about the H-Dog, click here



Solution to last week's birthday brainbuster


Last week’s birthday quiz proved very popular and quite the challenge - thanks in no small part to my omitting a vital hint from the original email!


To revisit the problem and see a full solution click here.


To our randomly selected winners Pat and Lindsay - well done, your prizes are in the mail.


And thanks for the many birthday wishes I got last week!


That’s all from me for now. If you'd like more geeky fun, please check out my other newsletters below, or connect with me on LinkedIn and/or X.


Yours in numbers,

Adam

0 comments

Recent Posts

See All

Comments

Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page