"Is the detection algorithm announced by Baidu the technical output of your cooperation?" Hinton's attention is more focused on technology, and he is a relatively pure scholar.
"It is not so much the technical output of the cooperation, but the technical services provided unilaterally by me." Meng Fanqi cooperated with Baidu at the time, and to a certain extent, he also hoped that he could take the initiative in the negotiations at this time.
"Unilateral technical services?" Hinton's eyes trembled, "What about the innovation on the detection side...?"
"The detection end has made a single-stage direct regression, which circumvents the traditional two-stage area proposal mechanism. Combined with my backbone network, the speed and performance have been significantly improved."
As Meng Fanqi spoke, he directly took out the same experimental results form that he showed to Robin Li at the time, as well as the final version of DreamNet's paper.
Meng Fanqi will publish this paper tomorrow, so that he can omit part of it when he attends the conference the day after tomorrow.
So there is nothing wrong with taking it out directly to the two people at Google to take a look at it first.
"This speed seems a little more reasonable." Hinton roughly estimated the number of parameters and calculations of the model, "It's a very shocking idea, it's true that a hero comes from a boy!"
Hinton's life experience was mainly in school. This time he came to Google mainly to use his fame and prestige to have a supporting effect.
Seeing these experimental results and papers, Hinton was very satisfied and his joy was beyond words.
I was thinking, this guy is still studying for an undergraduate degree. I'll see if there is any way in the future to get him to study for a doctorate under my name.
Jeff was on the side, looking at Hinton standing next to him, smiling from ear to ear, thinking that he really couldn't count on Hinton too much in the negotiations.
Although he has a lot of bargaining power on the successful cooperation with Baidu, Jeff still fights for Google's initiative.
"But since this technology has been cooperated with Baidu, it shouldn't be handed over to Google, right?" This is a very reasonable argument, but he never expected that Robin Li would actually offer a very good discount this time in order to make friends with Meng Fanqi. technical cooperation contract.
"According to the contract signed with Baidu, I still have the right to provide this technology to the company I joined. However, the company needs to guarantee two things: it will not continue to provide this technology to the outside world, and it will not get involved in China's testing-related markets. and industry.”
Jeff can completely accept this situation. Google has already withdrawn from the Chinese market, and there is not much direct competition between the two parties.
But he didn't understand what Li Yanhong was thinking. He only signed a technical service contract instead of buying out the technology at a higher price.
This also reflects the difference in approach between Robin Li and Jeff.
Robin Li believes that the key to the matter lies in whether he can get along well with Meng Fanqi. Compared with the subsequent development, this time's technology is just a drop in the bucket.
Jeff, on the other hand, prefers to thoroughly control existing technologies and discuss matters as they arise.
"We initially took a fancy to your creativity and ideas on DreamNet, so we contacted you through your previous email exchanges with Alex. But in just three or four weeks, you have already been involved twice in a row. Achieving great results again, and in a completely different field than backbone network design."
Pressure to pressure, Jeff was also quite shocked and appreciative of Meng Fanqi's achievements and abilities.
"From this point of view, the two-year signing fee of US$5 million we offered last time, and the total annual salary package of one million (~30 salary + ~60 stock + ~10 bonus) price may be outdated, right? "
Jeff turned his head and made eye contact with Hinton, and then asked tentatively, "How about doubling your signing fee on this basis?"
"Jeff, you told me at the very beginning of the negotiation to go straight to the point and not make any detours. Why are you now starting to take steps and test things a little bit?" Meng Fanqi frowned slightly. Jeff's operation was a bit of a double standard.
"Thirty million US dollars for three-year signing fee, you have to sign for at least three years." Jeff changed his style directly, as if he had revealed his trump card, but in fact he still left some leeway for himself.
From 10 million to 30 million, it sounds like a threefold increase, but it also changed from two years to three years, so in reality it is just another doubling.
The annual salary in this model is very small, and most of it is the signing fee buyout.
Meng Fanqi thought for a while and found that two years was more acceptable to him, but three years was a little too long.
If I had joined the company in early 2014, two to two and a half years would have been enough around the beginning of 2016 to end the early period of frantically publishing papers.
At that time, Meng Fanqi plans to quickly launch and implement some technical products by relying on his academic status and capital accumulation, as well as his prediction and grasp of the general trend and successful models.
At that time, if you still sign with Google, you will inevitably have a lot of inconveniences.
The reason why Ilya left Google to go to OpenAI at the end of 2015 was because the procedures for implementing these specific projects in a large company were very cumbersome.
Even for top technology companies like Google, it is inevitable that laymen will guide experts and management will override technology.
It's okay when you're just doing research. It's not easy to involve other companies and put it into production and deployment.
"I still hope to formulate a sharing contract in the form of one signing per year. If it is a sharing contract, I can waive the signing fee at all, and you can also decide whether to renew the contract based on my output." Meng Fanqi still adheres to his contract model, once The buyout operation is really a loss for myself.
"After you join Google, you will inevitably have technical collaborators, as well as multi-department collaboration, the sharing model, and how the specific contributions will be divided. We still need to carefully discuss and confirm the specific terms."
Jeff is not opposed to this model. After all, the proportion of net profit can be agreed upon. If he really gets a larger share, it is not a bad thing for the company.
But at this moment, they really didn't expect that this undergraduate with little social experience would directly reject the two-year contract of tens of millions of dollars offered by them, and the subsequent three-year contract of thirty million dollars.
This was completely unexpected by the two of them, and they had not prepared a particularly perfect sharing plan.
Regarding Meng Fanqi's request, Jeff also needs to discuss it with other senior executives. If you don't get this ratio right, it's easy to give out too much.
You must know that when Hinton and the other two heard about the $44 million last year, their legs were a little weak and their steps were a little unsteady.
When Jeff comes this time, the real bottom line is actually $50 million for four years, which is an average annual signing fee of $12.5 million.
But at present, it seems that the pure signing fee cannot satisfy this young man's appetite.
"It doesn't matter. We can discuss the specific terms of the sharing model tomorrow or after the IMAGENET seminar the day after tomorrow."
Meng Fanqi was not in a hurry to finalize the contract today.
The highlight of Baidu's side, the technology conference is still to come.
When the details there are announced and the specific situation of the cooperation is revealed, the space for operation will only become larger and larger.
