Intel roars Nvidia, is it because of the automatic driving safety model plagiarism?
Recently, Intel and Nvidia once again stood on the same starting line, this time the two giants will be in the autopilot safety model. What is a “security model”? Simply put, it is a strategy based on a mathematical validation model that is designed for autonomous driving.
Not long ago, Nvidia announced its own security model “Secure Force Field” (SFF) on GTC 2019 and began to define hardware in software. Intel/Mobileye, which announced the Responsible Sensitive Security Model (RSS) two years ago, couldn’t sit still.
On the 25th, Intel also published an article in a tit-for-tat manner .
Intel/Mobileye is blunt to say that the opponent’s SFF model is nothing more than a “replica” of the RSS model. Intel’s anger can be understood, Nvidia CEO Huang Renxun described SFF as the “first” (same type) security model of the autonomous driving industry last week, and did not put the RSS that was born early.
Mobileye CEO Amnon Shashua pointed out that SFF is very much like RSS. “You can imagine that we were surprised when we heard Nvidia SFF.”
In fact, last year Nvidia and Intel had several negotiations, the topic is to cooperate in the automatic driving safety. However, the two giant summits ended up in disappointment, and the two sides who shared each other’s ills failed to work together toward the future.
Intel claims that Nvidia expressed its willingness to cooperate in the RSS security model, but the two sides ended halfway at the end of last year.
“Nvidia chose to quit cooperation, we are very surprised,” Shashua wrote. In Intel’s view, everything is Nvidia’s fault.
Shashua stressed, “In my opinion, SFF is just a bad imitation of RSS. Of course, it is covered in black and green (Nvidia color) coat. To say that this security model has any innovation, I am afraid I can only find it in the language description. “
Nvidia certainly does not agree with Shashua.
David Nister, vice president of Nvidia’s Autopilot Software Division, responded with an interview. “Our SFF security model was established a few years ago, with a focus on the basic anti-collision core principles, rather than setting up a lot of rules and exceptions like RSS. SFF uses Both the algorithm and the mathematical validation model were developed by Nvidia and did not copy Intel.”
Nvidia’s exit from Intel’s cooperation is mainly due to too many frame frames set by the other party. Nvidia’s head of automotive business, Danny Shapiro, said the other side insisted that Nvidia (and other partners) accurately implement the RSS equation. As the industry security model continues to evolve, Nvidia finds it difficult to continue to work with Intel.
Nister also admits, “It is true that the industry has only one security model with the best standards, but in the real world, different drivers and different vehicles share the same path, so the security model must learn to live in peace.”
Intel/Mobileye has set foot on the RSS path, but in Nvidia’s view, “Intel can’t arbitrarily stipulate that others must follow the RSS rules.”
SFF vs. RSS
However, Shashua’s article is expected to be very good, he even directly compares RSS and SFF point-to-point, and thus the two security models are almost identical. Perhaps the security models of the two companies are slightly different in some respects, but Shashua feels that small places can be ignored.
In his opinion, Nvidia lacks transparency.
An Intel spokesperson said that from Shashua’s article, “SFF is indistinguishable from RSS in its main concepts,” and the downside of Nvidia’s imitation is that the main components of the model are missing and not clearly defined. In her view, there is still a big gap between the so-called “safety procedures” of SFF and the “appropriate response” highlighted by RSS.
Shashua emphasized that in the terminology of RSS, “appropriate response” is the absolute core concept of the entire architecture. “In my opinion, if you want to really apply the autopilot security model, you have to clearly define this appropriate response, which Mobileye has been doing.”
It’s boring to listen to competitors, or listen to the opinions of onlookers.
Ian Riches, senior director of global automotive business at Strategy Analytics, said: “From the news we get, Nvidia SFF does not differ significantly from Intel RSS,” he explained. “Nvidia claims that the SFF model has frame-by-frame analysis capabilities in GTC 2019, and it can better solve both vertical and horizontal detection tasks in a better way. Nvidia is only verbally speaking and does not actually prove how powerful SFF is.”
Although Nvidia SFF is somewhat familiar, is Intel really worthy of this incident?
Intel also answered, the spokesman said: “The cottage version of RSS will only be confusing, and even cause many companies to abandon the open security model. At present, some companies still insist on using opaque security models for their own interests, but Intel believes this is for market development. There is no boost.”
The spokesperson also stressed, “Our security model is completely transparent and open to all companies. Intel welcomes everyone to apply RSS to their own solutions, just like Baidu.”
Riches expressed similar concerns to Intel that he feared that the proliferation of security models would be affected. “When RSS came out, Intel spent a lot of resources on the front screen and even offered all the materials for the practical application of third-party companies. After the announcement, Intel also upgraded RSS several times.”
Intel’s concern is that if a large number of security models are “flooding”, people will not open their minds, because protecting their own IP will become more important than promoting innovation.
Of course, Nvidia also gives SFF “open and transparent” positioning, and Nvidia even claims that “SFF is an open platform that can be used in conjunction with any driving software.”
At GTC 2019, Nvidia said: “The security decision strategy in the motion planning stack, SFF monitors and blocks dangerous behavior. It has been able to distinguish between obstacles and complex road rules.”
Security model selection station?
So, has the autopilot safety model begun to stand on the sidelines? This is really the case.
Riches revealed that Mobileye has been secured by Valeo, Baidu and China Intelligent Transportation Systems (ITS). In addition, Flowserve, Israel’s Champion Motors and Beijing Public Transport will also adopt the RSS model.
On Nvidia, Shapiro acknowledged that there is no company that supports SFF.
If RSS can dominate the country, will Intel receive an authorization fee in the future? In Riches’ view, Intel really wants to go with the licensing model, but it won’t charge.
However, on the issue of charging fees, Intel spokespersons are obviously “shaken” – “free licensing is indeed our original intention, but Nvidia is a question of doubt. Is there any good news to keep it transparent?”
At the end of the article, Shashua concluded, “We hope that Huida can be more transparent in the details of the “safety procedures” so that the industry can evaluate the safety and effectiveness of SFF.” At the same time, he also hopes that “SFF’s “safety program” is based on independent innovation, and Non-imitation of Mobileye’s leading technology.”