No, do you need to use ChatGPT to build a car now?

The popularity of ChatGPT has caused global concept stocks to start a surge.

In the domestic capital market alone, a number of AI companies such as Baidu, iFlytek, and Yuncong can be said to be rising all the way.

Even Hanwang Technology, which is engaged in handwriting recognition and OCR, can hit the daily limit 5 times in a row, which has attracted many netizens to say "this can all be related"...

But family members , as the saying goes, "There is no most unexpected thing, only more unexpected ones." There are even more unexpected companies that are going to join ChatGPT -

Make cars!

Ah this... a conversational robot, is it now powerful enough to build a car? ? ?

Building a car, how to AIGC?

The first thing to note is that AI car manufacturing does not refer to AI controlling the production line, but the use of AI algorithms to complete the design of auto parts.

The equivalent of ChatGPT in the automotive industry is the automotive algorithm design model AAM™ (Automotive Algorithm Modeling), proposed by PIX Moving, an autonomous skateboard chassis and vehicle development and manufacturing company.

Very similar to AI painting and AI chatting, AAM™ is also used by humans giving parameters or sketches, including areas, functional requirements and constraints that allow AI to generate.

Different from the premise of "based on a huge amount of data and rich training resources" for text and painting, in the industrial field, the amount of data is often difficult to reach the amount of data in the field of painting and text.

Therefore, based on the characteristics of industrial scenarios, PIX proposes a strategy of combining rule-based small models and large models, using the large model to generate approximate designs by learning massive vehicle data, and then optimizing the approximate designs through classic algorithms It is a precise engineering document to meet the manufacturing needs of industrial scenarios.

AAM™ can be seen as a designer with “experience” and “mature techniques”.

Experience comes from a large amount of training data. Mature techniques refer to the completion of subsequent simulation verification while giving the design.

According to the traditional process, after the preliminary design plan is given, it must be verified by simulation, and areas that fail to pass verification must be returned for modification.

In the past, if human engineers wanted to modify a plan, the workload was actually not much smaller than designing a new one from scratch. It could take as little as a few days or as much as a week.

As for AI to modify the plan, it only requires changing a few parameters, which can be done in a few minutes.

You can also generate batches at one time according to different parameters, and then filter out the best ones. While saving time, the quality of the final selected solution can also be improved to a higher level.

How much can the entire process be accelerated with the participation of AAM™?

The figures given by the PIX Moving team are: the time for the entire product is reduced by 60%, and the time for a single part is shortened from days to a few hours.

Compared with consumer-level AIGC algorithms such as AI painting and AI chat robots, AIGC is much more difficult to use in industry.

For example, when AI is used to modify photos or videos, there is no need to consider the physical properties of optics and materials. It only needs to "fool the human eye" at the pixel level.

In industry, it is not that simple. The designed parts must eventually enter the production process. They cannot fool lathes or 3D printers. They must be real and usable and meet manufacturing requirements.

△3D printer is producing an AI-designed skateboard chassis structure

##△3D printed self-driving skateboard chassis (actual vehicle) designed by AAM™

As we all know, AI algorithms based on deep learning are uninterpretable and are really not good at this aspect.

In order to overcome this problem, the AAM™ algorithm is completed by the cooperation of AI algorithm and classic geometry and physics algorithms. This is also the innovation of AAM™ compared with other AIGC applications.

With the cooperation of AI and traditional methods, the designed parts have not only entered production, but have even been installed on cars. It has been used in the self-driving skateboard chassis developed by PIX, as well as the chassis and frame design of shared mobility space (Robobus) and personal mobility space (NEV) products.

△Shared mobile space (Robobus)

△Personal mobile space (NEV)

In addition, PIX’s partner Fulongma Group has applied AAM™ to the design and manufacturing of cleaning robots.

△Fulongma Group Unmanned Sanitation Robot

According to the PIX Moving team, experienced engineers are actually easily fooled Due to limitations of past experience and path dependence, it is not easy to produce works that are both creative and meet production needs.

AI is not subject to this restriction. The more training data, the stronger its capabilities.

Generally speaking, in addition to speeding up the process and saving costs, AI car manufacturing also has room for creativity and design quality.

It is understood that AAM™ has been in use within PIX Moving for more than a year. It was precisely because of such remarkable effects that the team gained the confidence to open it to the outside world.

Productization is a key step in bringing AI algorithms to the market. Just as GPT-3’s influence has been limited to the technology circle for many years since its birth, ChatGPT became a global hit when it was made easy for everyone to use.

PIX Moving team said that the final product will be a Web-based cloud service.

It does not require the cumbersome configuration and deployment of industrial software in the past. It can be used as long as you open a browser, and it supports multi-person online collaboration.

Through intelligent workflow innovation, AAM™’s ultimate goal is to bring disruptive changes to the industry.

AIGC is changing the production model

It is not difficult to see that the so-called "building a car with ChatGPT" is not what it means literally.

Instead, the AIGC model represented by ChatGPT is used to change the design, engineering and manufacturing links in the traditional car-making process.

A major pain point inherent in traditional car design can be summarized as "one hair affects the whole body", because the subdivision processes it covers are not only complicated, but also interlocking.

For example, computer model drawing, engineering analysis, cost estimation, reliability experiments, etc. Once the customer proposes a change request for a certain link, it is tantamount to sending a "Start Over" to the design team...

The field of automobile design is not only fragmented in terms of demand, but also gradually developing in the direction of diversification and non-standardization. Different scenarios such as logistics, security, agriculture, etc. will make automobile design It is impossible to achieve "one template covers everything".

As for AIGC like ChatGPT, one of their major features is that they can use expert-level knowledge and ability reserves to present high-quality results in a very short time.

With the combination of the two, it is not difficult to figure out the goal of PIX Moving - to make AAM™ the ChatGPT of the industry.

Judging from the actual results after "on the job", the correctness of this model can also be confirmed.

It is understood that with the support of AAM™, the efficiency of the automobile design process can be increased by 10 times, which is 50-100 times higher than that of manual design, and it can also provide "better accuracy and stability". The kind of "optimal solution".

Moreover, PIX Moving not only exerts its strength on the design end, but also proposes RTM™ (Real-time Molding Model) on the manufacturing end. The method of this combination punch can be summarized as follows:

• AAM™: digitize and systematize vehicle design, quickly determine high-quality design solutions;
• RTM™: take over the determination of solutions, and quickly carry out digital manufacturing through a mold-free real-time molding system.

△RTM™, real-time molding technology demonstration

This model of improving automobile manufacturing efficiency with AIGC ideas has also gained global recognition Recognition from the industrial software giant Autodesk:

In October 2022, PIX Moving was awarded the Innovator of the Year Award for its innovation and contribution in AI algorithm design and new manufacturing processes.

But the game played by PIX Moving is not limited to the field of automobile manufacturing. It is understood that they will also expand this field in the future. This model has been replicated in architectural design, aerospace, shipbuilding and other industrial design fields.

As for the more far-reaching goal, perhaps just like Ford's invention of the assembly line a hundred years ago, which allowed cars to enter thousands of households and become mass consumer goods, PIX Moving's process innovation will make personalization the mainstream of the next generation of industrial production. As CEO Yu Chuan said:

The manufacturing process determines the boundaries of hardware product innovation.

......

So can ordinary developers experience ChatGPT in the industry? This can be done.

It is understood that in April 2023, PIX AAM™ will open some free function modules to the outside world in the form of SaaS or PaaS!

Friends who want to experience it can wait for the benefits.

The above is the detailed content of No, do you need to use ChatGPT to build a car now?. For more information, please follow other related articles on the PHP Chinese website!