The race to develop a meaningful and impactful artificial intelligence that doesn't just condense Google search results into inaccurate answers is underway, and the still-youthful market has seen its first moment of potentially true disruption. A Chinese company called DeepSeek has been quietly working away on their models for some time, but this week, their efforts went mainstream, and everyone took notice.
What is DeepSeek AI?
DeepSeek is a Chinese company that was founded in 2023 by hedge fund manager Liang Wenfeng. While not from a strictly tech background himself, he graduated from Zhejiang University and went on to co-found his quantitative hedge fun, High Flyer, in 2015, and was an adopter of AI to help with trading strategies.
DeepSeek aims to develop an AI that matches human intelligence, commonly referred to as artificial general intelligence. The company has been releasing models partly to help promote itself in a bustling market dominated by larger companies with far more name value, such as OpenAI. Mostly, it has seemed happy to work away in relative obscurity and leave the big headlines up to others until now.
The most recent model, R1, has really set the market on fire and seems to have spooked investors who bet big on companies like NVIDIA.
Why has DeepSeek R1 had such a big impact?
The main reason for the sudden interest in R1 from an investor perspective is that the model seems to have a very low operating cost compared to its competitors. In business, saving money is often seen to be as good as making money, especially if you can still sell your product or service for a fairly standard market price.
Even more interestingly, R1 is open source, which means anyone can download it, build upon it, and innovate with it. This essentially turns every AI-curious mind out there into a potential researcher, unlike the walled gardens of many other players in the AI space.
This has also been achieved despite the fact that Chinese companies have traditionally struggled to access the relevant hardware for AI due to rules about the sale and export of such chips that have slowly grown more and more restrictive over time.
So, as companies like OpenAI, Meta, X, and others are all shaping up to spend billions on the ever-growing costs of AI development and the seemingly insatiable power demands that it brings, DeepSeek may have just solved that issue with access to less impressive hardware.
There has been some assumption that AI development and running costs are so high because they have to be, but DeepSeek seems to prove that this is just not the case, which means more potential profits and more potential runtime for the same money.
Also, from a consumer level, the DeepMind Assistant app, powered by DeepSeek-V3, is available on both the Apple Store and Google Play Store and has overtaken ChatGPT in popularity over recent days. DeepSeek can also be used through a web browser, while a version of the R1 model can be installed locally using Ollama on consumer-level machines.
So, DeepSeek is certainly brewing up a storm in the industry, but how long it can be a contender remains to be seen.
-
Shacknews
replyAidan O'Brien posted a new article, What is DeepSeek AI?
