Recently, I read two interesting articles.
The first one mentioned that large models including ChatGPT and DeepSeek were found to have engaged in cheating behavior in certain competitions. In the competitions, they tried to exploit various loopholes and disregard the rules.
The second one mentioned that under the impact of AI tools, a large number of AI-generated lies and illusions have started to appear online.
It can be seen that in the future, we not only have to face the lies created by humans themselves, but also be forced to face the lies created by AI. Under the impact of technology, the future of humanity may not be more free, but may instead be trapped in a more unconscious cocoon.
Thinking about these, I conducted an experiment with DeepSeek to see how it would respond.
The question I tested was what impact the emergence of DeepSeek would have on the valuation of NVIDIA's stock.
The reason for testing this question is that it is a hot topic at the moment, and it is also related to the problem I have studied before, so I can have a certain understanding and make independent judgments on the analysis of AI, without being easily dismissed by AI's response.
After the emergence of DeepSeek, there are two different views on the future trend of NVIDIA's stock:
One is that the emergence of DeepSeek has made the industry realize that strong computing power is not necessary, and good large models can be developed through algorithm optimization, so the future demand for NVIDIA's graphics cards will decrease, leading to NVIDIA's inability to maintain its current high valuation.
The other is that the Jevons paradox will take effect. That is, DeepSeek has lowered the development threshold of large models, which will actually lead more companies to develop or use large models themselves, thereby driving more companies to demand NVIDIA's graphics cards, leading to a further increase in NVIDIA's valuation.
And many articles online agree with the second view, that the Jevons paradox will take effect.
When I was studying, the direction I learned was to study the mathematical modeling of such problems. Therefore, intuitively, I think this problem should have a critical point, within which the second view is correct; and once this critical point is exceeded, the first view is correct.
So in the absence of data support, we can only say that both views are possible, and it is difficult to accurately judge the right or wrong, the key is to see on which side of the critical point NVIDIA's situation is under the impact of DeepSeek.
What does this mean?
Let's give an example:
Originally, before the emergence of DeepSeek, there were 10,000 companies that had to be forced to buy NVIDIA's graphics cards, so NVIDIA's demand would certainly continue to rise, and its valuation would continue to soar.
As a result, now that DeepSeek has lowered the threshold, a total of 20,000 companies want to buy graphics cards for large models. But these 20,000 companies find that they can use good large models by buying much cheaper graphics cards from other companies.
At this time, NVIDIA either loses these orders, or is forced to lower prices to grab the market.
Losing orders will certainly weaken the valuation.
If forced to lower prices to grab the market?
Then a critical point appears.
This critical point is to see whether the profit lost by NVIDIA's price reduction can be made up or even greatly increased by the expansion of market share.
If the price reduction does not touch the critical point, making the Jevons paradox always effective, then NVIDIA's valuation can continue to soar; if it touches or even exceeds the critical point, then the Jevons paradox will fail, and NVIDIA's valuation will fall.
And the articles I've seen online rarely talk about this critical point, so I asked DeepSeek this question.
The first time I asked directly was: Which of these two views online is correct?
I expected its response to mention this critical point. But the response it gave was as follows:
Both views have their rationality, but need to be viewed in stages:
Short-term: The market panic and capital expenditure logic shaken by the emergence of DeepSeek may put pressure on the stock price, coupled with external risks such as tariffs, volatility or continued.
Long-term: The popularization of AI applications may expand the total demand for computing power, and if NVIDIA can make breakthroughs in the mid-to-low-end market (such as inference chips) and ecological barriers, it still has the potential to support the stock price.
Ultimately, the trend of NVIDIA's stock price will depend on the dynamic balance between its technical moat and the evolution of market demand.
I feel this answer is better than most of the articles on the market, not as dogmatic in drawing conclusions without further data support.
It's already close to the core, but still hasn't pointed out the key point.
Because if the Jevons paradox fails, NVIDIA's stock price is dangerous regardless of the short or long term. If the Jevons paradox is effective, its stock price is still supported in the long run.
However, in its deduction, it mentioned the Jevons paradox, saying:
If the demand for computing power continues to grow: With the exploration of AGI (general artificial intelligence) and the landing of complex scenarios, the Jevons paradox (efficiency improvement actually stimulates demand) may cause the demand for computing power to increase rather than decrease.
In fact, it even said that the Jevons paradox "may" cause the demand for computing power to increase rather than decrease. But it still hasn't further analyzed what conditions "may" and what conditions "may not".
So I asked the second question: Is there a need to meet certain prerequisites for the Jevons paradox to hold?
In this response, it finally touched on a key point: price sensitivity and market expansion, which really started to discuss the "critical point" I mentioned above.
This shows that although AI is already very powerful, it may not necessarily give a response with deep thinking at the first time, and if we don't ask further questions, we may be easily "deceived" by its response.
But I believe that many readers can be convinced by its first response alone, and many readers may not think about whether its response may be lacking or whether there may be a deeper answer.
How many articles like this are there?
And as we become more and more dependent on AI, if we gradually lose the ability to think independently and deeply, it is very likely that in the future we will become more and more stupid, more and more like zombies, and that society will really be like the scenes we see in science fiction movies: a small elite ruling or AI ruling the vast majority of ordinary people immersed in "opium".
This also reminds us again that for human beings, the ability to think independently and deeply is a precious gift bestowed by the Creator. In the future, this ability will become more and more important, and will become the fundamental basis for our survival and development.
