On DeepSeek and an Educational Red Flag
Written by: Christian Talbot (Posted Mon, Feb 3rd, 2025 | 9:03 am)
If you are moderately aware of what is happening in the world, you have probably heard of DeepSeek.
It recently rocketed to the #1 spot in the Apple App Store after news stories revealed that it met or exceeded results from the leading frontier AI models. Collectively, Nvidia and other AI chip manufacturers lost $1.2 trillion (yes, trillion) of market capitalization when it was revealed that DeepSeek had been trained at a fraction of the cost to develop other frontier models.
Besides DeepSeek’s cost-efficient superior performance, the other storyline has been the control exerted by the Chinese government over DeepSeek’s responses. If you ask about Taiwan, censorship in China, and other topics about which the CCP has strong feelings, you can imagine the results.
But this is not a post about politics. This is a post about epistemology.
In a world of increasingly abundant AI, how do we get clear on what we know, how we know it, and why we think we know it? After all, DeepSeek isn’t the only AI using algorithms to deliver answers. Every AI shapes what we know—or think we know—through algorithms that even their creators do not entirely understand.
Jeppe Stricker framed the issue this way: “The emergence of cost-effective but restricted AI models suggests a future where educational technology accessibility comes with embedded constraints on academic inquiry.”
In other words, when the technology we use is inherently biased and uses opaque algorithms to produce responses, we do not know what we do not know.
That’s a huge educational red flag.
Whether it is DeepSeek “extinguishing” facts or ChatGPT providing sycophantic replies or any LLM providing the most likely (ie, “average”) response to a query, educators must develop strong AI literacy.
The best resource for individual teachers to establish AI literacy is AI for Education’s course for educators. It is free, high quality, and self-paced.
And the only resource to ensure that an entire school can become literate, safe, and ethical in its adoption of AI is the RAIL endorsement from Middle States.
If I’m in the casino, I’m placing all of my chips on more, not less, AI.
If that’s correct, then how will you know what you don’t know?
