He compared this to theft.
According to him, the Chinese company applied the outputs from OpenAI’s models to train its own.
”I don’t think OpenAI is very happy about this,” Sacks stated.
Microsoft and OpenAI have begun an investigation into this matter, media reports indicate.
Microsoft security experts noticed in the fall of 2024 that individuals associated with DeepSeek were ”exfiltrating a large amount of data” using the OpenAI API.
In the official documentation for the DeepSeek R1 model, the OpenAI package is utilized.
DeepSeek’s Success
Over the weekend of January 25-26, the DeepSeek-R1 neural network caused a stir and led to sell-offs in stock and cryptocurrency markets. The company’s chatbot surged in ranking among free apps on the App Store in the U.S., surpassing ChatGPT.
The reason lies in offering top performance at significantly lower costs compared to American models. OpenAI CEO Sam Altman responded to the emergence of this competitor, calling the much-discussed DeepSeek-R1 ”impressive.” He also promised to outdo the Chinese rival
”Clearly, we will provide much better models. Moreover, it’s great to have a new competitor! We will roll out several releases,” he added.
In a comment, Credentia CEO noted that DeepSeek could shake OpenAI’s dominance in the AI sector.
”This has already happened. […] But leaders in AI change about every month. One month it’s Google, then Anthropic, then OpenAI, then LLaMA [Meta], and now DeepSeek. Their model today is not the best. It is unlikely to remain a leader. I am sure we will see much more competition and many more players. But nothing terrible has happened to American competitors so far,” he remarked.
Gershuni also agreed that OpenAI could mount a worthy response to DeepSeek, as Altman promised.
”Yes, of course. We will see this happen many times again. Moreover, both companies will continue to attract attention repeatedly. This is just the beginning,” he noted.
According to him, increased competition will lead to better products on the market.
”[…] In AI, there is also a political aspect, some global dominance at play. Whether it will be China, the USA, or a decentralized internet remains unknown […],” he believes.
Meanwhile, the U.S. Navy has already prohibited its personnel from using DeepSeek’s AI technology due to ”potential security and ethical issues,” CNBC learned.
Currently, solutions from OpenAI and DeepSeek offer their own advantages and disadvantages, according to Gershuni. Among ChatGPT’s strengths, he cited a more refined product that includes:
- Better application
- Advanced voice mode
- Operation of the AI agent ”Operator”
- Online store GPT Store
Among its drawbacks are a closed and paid system.
DeepSeek, on the other hand, is free with open-source code. The model can be used locally on one’s device without internet access.
Impact on Stock Market
The emergence of DeepSeek has raised questions about the viability of multi-billion dollar investments in graphics cards since its model offers comparable performance to top solutions without requiring many expensive chips. Consequently, Nvidia’s stock fell by 17% during trading on January 27.
Opinions arose online regarding a repeat of Cisco’s history when its shares plummeted significantly after a substantial rise during the dot-com crash.
Gershuni believes Nvidia will not face Cisco’s fate.
”[…] There are no reasons for Nvidia’s shares to crash seriously today. They might be worth several hundred billion less, but that doesn’t impact anything. Nvidia remains an absolute leader, effectively a monopoly, and will likely stay that way for some time. I would rather buy Nvidia shares; however, this is certainly not financial advice,” he emphasized.
After a double-digit decline, the chipmaker’s value rebounded by 9% during trading on January 28.
Niles Investment Management fund manager Dan Niles believes that the correction in shares is not yet complete.
”I think investors should be very cautious in thinking this is the bottom,” he stressed.
It should be noted that experts have called the excitement surrounding DeepSeek premature.
Previously, the startup introduced a set of new multimodal AI models that it claims surpass OpenAI’s DALL-E 3.
This article was originally Posted on Coinpaper.com
