News from the AI & ML world
DeeperML - #aicompetition
|
@techcrunch.com
//
DeepSeek is rapidly becoming a major player in the AI field, attracting attention and concern from both US officials and established companies like OpenAI. There are allegations of DeepSeek circumventing US restrictions on advanced AI chip purchases. Reports indicate that the company obtained Nvidia chips through third-party transactions in Singapore, potentially violating export regulations. DeepSeek's growing influence is also evident in its AI model performance, which is now being used as a benchmark against which other models are being measured.
The competitive landscape is further complicated by the emergence of new AI models like the Allen Institute for AI's Tulu 3 405B, an open-source model that claims to surpass DeepSeek V3 and even OpenAI’s GPT-4o on specific benchmarks. In addition to the increased competition for AI superiority, there is discussion about protecting OpenAI from other competitors like DeepSeek including the use of watermarks and other methods to protect their IP. The European AI contender, Mistral AI, is reportedly losing ground to its US counterparts and facing significant challenges from DeepSeek's rise and may be losing market share and ARR to these other companies.
Share:
References :
Classification:
Vlad Savov@Bloomberg Technology
//
DeepSeek has emerged as a significant challenger in the AI industry, gaining rapid popularity with its R1 model, which is a low-cost platform. The company's app has topped download charts in 140 markets, prompting discussions about US companies sharing their AI advancements given DeepSeek's success. This has even led to Sam Altman, CEO of OpenAI, admitting that DeepSeek has lessened their lead in the field. DeepSeek's rise has fueled debate over AI knowledge sharing, also causing OpenAI to re-evaluate its own closed source approach.
However, DeepSeek's technology has also faced scrutiny. Texas Governor Greg Abbott has banned DeepSeek on government-issued devices, citing concerns about Chinese-backed AI platforms. Furthermore, security researchers have discovered the DeepSeek R1 model failed to block any of 50 malicious prompts, demonstrating the AI's vulnerabilities to jailbreaking tactics, raising questions about the platform's safety measures. This has led to concerns that the company may have sacrificed safety for a cheaper to build model, which did not prioritize safety.
Share:
References :
Classification:
|
BenchmarksBlogsResearch Tools |