The meeting lasted about forty minutes, and Chen Yu assigned tasks to each major development team.
The company's technology department also has many project teams, involving natural language processing, recommendation systems, image generation, deep reinforcement learning, computer vision, and robotics.
Quantitative Capital is regarded by the outside world as an investment management company, but it is actually a high-tech company disguised as an investment management company.
As the meeting ended, everyone else left the conference room, leaving only Fang Hong and Chen Yu.
"I don't think there is any problem with the development route you have set." Fang Hong said with a smile, "But we can't slow down. I have reliable news here. At Google, their artificial intelligence laboratory GP proposed an internal meeting at an internal meeting.
The idea is to enable machines to 'learn' language and grammar, and they are also working on language models just like you."
Chen Yu was surprised and said, "Google is also doing it? What level of technology has it reached?"
As the real boss behind Qunxing Capital, Chen Yu believes that he must have reliable information.
Fang Hong said calmly, "I'm not as advanced as you, I'm sure of this."
According to the memory of the past life, around October this year, Google should release a language model based on gpt. This model learns language by encoding the input. It is mainly a pre-trained language model, but Google has not developed it before.
This language model.
Hearing what Fang Hong said, Chen Yu immediately felt relieved. In this field, one step ahead is often one step ahead, and the development speed of AI often develops at an exponential rate.
After a moment, Chen Yu looked at Fang Hong and said, "Our large AI model, the first-generation iterative research and development task is to realize the generation of human text, supported by tens of billions of parameter scale, has now entered the second-generation research and development, compared with
The first generation is larger and more powerful, and can produce coherent and attractive responses to various prompts, with a parameter scale of fifty billion levels."
"As for the third-generation model, it should require 200 billion parameter support. The goal of the third generation is to be larger, more powerful, and more advanced, and to achieve amazing performance on multiple natural language processing tasks."
"The upgrade of large AI models requires the support of strong underlying capabilities to be able to play successfully. It requires large computing power and large-scale training data to break through the bottleneck of AI. The operating costs are also very huge. Now the operating costs in one month exceed
The scale is RMB 300 million.”
After hearing this, Fang Hongyan said concisely and comprehensively, "Money is not a problem. If it is not enough, I will take care of it."
Chen Yu smiled and said, "If it's not enough, I'll ask you for it."
Its underlying capabilities include computing resources, data storage, power supply, network bandwidth, etc.
These are very expensive things.
The training of AI requires the use of large-scale computing resources, including GPU, GPU, etc. Cloud computing can provide powerful and elastically expanded computing resources.
The underlying data center of cloud computing can provide high-performance computers and hardware equipment, and can be optimized to improve the utilization of computing resources, thus accelerating the training and generation speed of AI. It can also provide powerful parallel computing capabilities, multiple
Computing tasks are simultaneously allocated to different computing nodes for calculation, improving training efficiency and speed.
In addition, training AI also requires network requirements. One is that link load balancing must be perfect, and the other is that the network can recover quickly when a failure occurs.
Because of the "barrel effect", as long as one link has uneven load and causes network congestion, which becomes the shortcoming of the barrel, then even if the other two paths are unobstructed, the collective communication time will still increase significantly, thus affecting the training efficiency.
It is similar to the situation where one core is in trouble and many cores are watching.
As the size of the cluster increases, link failures are inevitable. A similar link failure will cause the entire communication to stagnate, which will stagnate the model training. Therefore, it is necessary to quickly repair the failure.
Fang Hong crossed his legs and leaned back on the chair. He took advantage of the armrest of the chair with one hand and said thoughtfully, "Your AI still lacks one element."
Fang Hongyan said "image" concisely and comprehensively.
Chen Yu was slightly stunned, "Image?"
Fang Hong immediately started and said, "How to make the product more quickly accepted by the public? How to form user stickiness, especially to capture the younger generation, the AI must be visible, it cannot be a string of 0s and 1s, so the AI should be given a
The character image, that is, personification, must capture the younger generation, especially the generation born in the 1990s and 2000s. Furthermore, AI should be given a two-dimensional character image.”
After hearing this, Chen Yu couldn't help but nodded, "That makes sense."
After a moment, he looked at Fang Hong and said, "I know how to do it. If you want to do it, let's do it big. Then I will set up an additional department to create a virtual image brand operation department for AI, create an exclusive IP image, and enrich
Its character design has a complete story background."
"Including hiring some professional animation production teams, screenwriters to produce some sophisticated story clips to enrich the character's image, and promoting it as the first aboriginal person in the 'virtual world', etc."
Fang Hong said, "It is not easy to build an IP from scratch until it is accepted by the public and deeply rooted in the hearts of the people. You have to put in the effort and be willing to spend money. It also requires time to settle. The sooner you do it, the better."
Chen Yu nodded. After being prompted by Fang Hong, he immediately understood the importance of visualization and personification of AI.
And it has great potential. When the app is launched in the future, it can become a must-install virtual assistant for users, including the Internet of Everything in smart homes. In the future, various home appliances and smart devices can be hosted by virtual AI assistants.
Now that this IP image is established and with the growth of the younger generation, there is a lot of room for imagination in smart home appliances alone, air conditioners, washing machines, computers, mobile phones and other equipment. These manufacturers are not capable of developing AI.
You only need to access the AI assistant of Quantitative Capital.
This chapter is not over yet, please click on the next page to continue reading! If access to this virtual AI assistant cannot be supported in the future, users will not even want to use products such as washing machines or air conditioners of this brand.
After all, there are many brands of washing machines and air conditioners, but there is only one such powerful AI assistant, and the IP virtual image created has accompanied the growth of a generation and formed user habits. As long as other air conditioners and washing machine products can support this AI
, users will have a great tendency.
At that time, those manufacturers will not be able to do anything without access, and it is impossible for quantitative capital to make air conditioners and washing machines on their own. This is not a bad thing for manufacturers. It can save themselves from being forced to enter unfamiliar fields and invest heavily in order to build a so-called moat.
To build AI and ecosystem, you only need to make your own products well and refined.
…
Fang Hong stayed at Quantitative Capital for an hour and a half before leaving, leaving these tracks to Chen Yu and having great confidence in him.
Returning to Jingxinju Villa, Fang Hong went to the study room and saw the fame system data panel in front of him. Then he directly spent 100,000 achievement points to redeem a prop card and checked Chen Yu's character attributes.
[Target Chen Yu]
[Excellent character]
【prestige】
【Healthy and Excellent】
[IQ 145, genius]
[Excellent coordination]
[Wealth 17.6 billion yuan]
…
"These hexagonal data are pretty good." Fang Hong was very satisfied with the six attribute data, especially the IQ value of 145. As long as it reaches 140 or above, it is considered a genius.
The most important thing is that these data of Chen Yu are all initial data and have not been enhanced or improved by Fang Hong.
Especially the IQ level, between 120 and 139 is considered an excellent level, while the level of most senior talents is around 130 to 139. Those who have been promoted to the leading talent level by Fang Hong have IQ levels above 140.
.
Chen Yu's initial data level exceeds those of promoted leading talents, which shows that it is very rare.
Finally, Fang Hong decided to raise Chen Yu's IQ level to 150. The further the level is raised by one unit, the greater the consumption of prop cards required.
Chen Yu's initial data was so good, which indirectly saved Fang Hong a lot of achievement points.
After a while, Fang Hong exchanged a large number of R&D prop cards, all of which were designated to improve Chen Yu's R&D capabilities, and the data was reflected in the increase in his IQ level.
After raising Chen Yu's IQ level to 150 before stopping, Fang Hong glanced at the consumed achievement points and couldn't help but muttered, "12.6 million achievement points were actually consumed..."
It takes 100,000 achievement points to redeem a prop card, which increases Chen Yu's IQ level from 145 to 150, which is an increase of 5 units, but consumes 126 prop cards.
However, Fang Hong doesn't feel bad. The more famous Chen Yu becomes in the future, his prestige and social influence will increase significantly, which will also indirectly accumulate reputation points and achievement points for Fang Hong. He has consumed more than 12 million achievement points here.
You will definitely get your money back and even make a lot of money.
