Nvidia ((NVDA) Earlier this year, stocks were hit after Chinese startup DeepSeek launched its low-cost AI chatbot, threatening its dominance. Kim Forrest, chief investment officer at Bokeh Capital Partners, joined the streets to discuss why this sell-off might be needed.
Related: NVIDIA’s 78% sales growth is impressive, but not sustainable
The following complete video transcript:
Conway Gittens: But now we have Nvidia ((NVDA) The results and the healthy appearance of this quarter are very healthy, how did it hit the DeepSeek sell-off in the tech industry? In hindsight, especially considering the idea that you may have too many AI models, you know, do you think the sell-off is necessary or overdoing?
Kim Forrest: I think it is necessary. For this main reason, Openai has achieved incredible success in building large language models. I won’t go out with nerds and tell you all the complexity I found so interesting, but they did a great job. But the way they do this is that they have trained a whole bunch of neural networks and then put them together, and they do it through what I think is a brute force method. They make it read the internet. OK, so the brute force method requires processors to read the Internet over and over for training. This is how neural networks do things. They look at things millions, billions, trillions of times and are creating data from them that they can then retrieve and use. OK
So DeepSeek shows people, investors that there may be other training methods. I think that’s the key, we probably don’t need the amount that OpenAI has been using and other technologies have to be created by computer scientists to better train these networks. I think this is indeed a difficult problem. How long does it take? The brute force method does not seem to return the benefits it has earned. Because if you look at Openai, they keep saying that they will have the next model that will be better. But they were about 18 months behind. And I think it’s because the old method of just digging out a lot of data and getting the neural network to chew on doesn’t work. So DeepSeek shows us other ways we should look at and doesn’t involve a lot of data centers.
Watch more videos:
- This common budget error is actually spent on you
- Senior economists say tariffs may not lead to higher prices.
- How automation changes professional sports
- Diversity is not what it used to be – How to invest in Legend
