How to make hardware products in the AI era？ The founders of Nothing mobile phone and Insta 360 think so.

1. AI is sweeping the entire technology industry, but there are also many hardware products that are just increasing the function of AI gimmicks. The integration of AI and hardware needs to be more compact and intelligent, rather than just helping users edit photo albums and organize documents.

2. Data security in the AI ​ ​ era is more important than at any time. In the AI era, the user information collected by hardware and applications is more and more detailed than in the past. In the future, human AI model analysis and data are likely to become the same as today’s mobile phone numbers, and once problems occur, they are irreversible.

3. In the long run, in the future hardware products, the big model may be deployed more in the cloud than on the device side. But it will also bring about data privacy problems.

AI is changing everything in the technology industry.

This change does not only happen at the software level, such as GPTs who can interact through web pages or apps. NVIDIA, which provides chip support for large-scale computing, is now the third largest company in the world by market value. Only a few years ago, this company was just a "game company" that sold graphics cards with the support of game enthusiasts.

The hardware side is also undergoing changes. Now the new products of mobile phone manufacturers are called "AI mobile phones", which can help you analyze photos and edit your messy photos in the photo album into a Vlog. Voice assistants who used to set the alarm clock can also help you edit a little red book or play a little idioms solitaire game after accessing the big model.

Personal computers, which are more desolate than the mobile phone market, are more actively embracing the arrival of AI hardware. Microsoft exerted its strength at the system level, while NVIDIA increased its strength at the chip level, and together with computer manufacturers, it launched a new era of "AI PC".

Hardware +AI is one of the biggest changes in the electronic consumer industry in the past decade, but it has also been questioned. Some hardware products only rely on the concept of AI and have not produced much innovation.

"AI has completely changed the whole industry." Pei Yu, co-founder of Nothing Technology, said in the roundtable forum of Phoenix Net Technology, but now the biggest volume is AI, and many unreliable projects have been born. "Many teams don’t have much passion in this field, but when they see opportunities, they come in and want to make a sum of money. This mentality is not very reliable."

Liu Jingkang, founder and CEO of Insta360, said that the Copilot+PC released by Microsoft may be the best product he has seen integrating AI and hardware. But more of them are "gimmicks". For example, some pet robots just integrate a big prediction model on the hardware, which is not much different from smart speakers. "Just using the concept of AI is not really for consumers. Solve practical problems."

During this year’s BEYOND International Science and Technology Innovation Expo, Phoenix Network Technology invited Pei Yu, co-founder of Nothing Technology, and Liu Jingkang, founder and CEO of Insta360, to participate in a round-table dialogue to discuss hardware innovation in the AI ​ ​ era.

01, AI era may not need new hardware.

Will brand-new hardware products be born in the AI era? Is it an iteration of existing personal electronic devices such as mobile phones and computers, or will innovative interactive products such as Apple Vision Pro be born?

As two founders who have been deeply involved in the field of consumer electronics for many years, Pei Yu and Liu Jingkang both believe that the demand of consumers is the first step.

"First of all, who is your target group? What’s wrong with them? How can I help them solve it? " Yu Yu said. But at present, it is certain that applications are separated from each other, and "applications are not so intelligent", which must be the direction of intelligent hardware in the post-application era.

Liu Jingkang pointed out that some companies will have the problem of looking for nails with hammers. A company has good technology and then looks for application scenarios. He believes that we should first find the current crowd and target scene, and then think about how to use technology to solve these problems.

The current mobile phone market has also begun to use AI to improve the innovation of some function points, such as summarizing documents in mobile phones through algorithms and intelligently cropping pictures. However, in Pei Yu’s view, the innovation of these function points is "actually boring", which is to create a solution that everyone does not need.

In the round-table dialogue, Liu Jingkang also said that many people will do some picture editing in the mobile phone, which is really boring. "Because everyone can do it, you can’t do anything particularly big (innovation)." This is a challenge for many hardware devices at present, but it is also an opportunity.

However, it seems unlikely that newer hardware products will be born in the industry. Pei Yu said that at present, the annual sales of mobile phones are about 1 billion, and TWS headphones are about 300 million. These are the products with the highest sales volume and closest to users, and they are also products that can really understand users and consumers. "Consumers have been taught how to use these products, and why do they need to make some new equipment?"

Today’s mobile devices still have a lot of room for iteration. Pei Yu pointed out that there is not much difference between the current mobile phone interaction and the Symbian S60 interaction. It is indeed smoother, faster, and some function points are better, but the basic logic has not changed. It is still to open an application, return to the desktop, and switch applications.

02. Data security in the AI ​ ​ era is more important than at any time.

Pei Yu believes that the core of security in the AI ​ ​ era lies in truth, that is, whether AI can give the most authentic and accurate answer and the most correct answer. At present, some large model products are still plagued by hallucinations when answering questions, and sometimes they give questions that are obviously inconsistent with the facts.

For example, Google’s Gemini previously launched the function of "Wen Sheng Tu", which portrayed the founding fathers of the United States as several blacks, and even appeared the "black version of George Washington". "This is not a true judgment. Although it is only a dazzling function now, if AI has to assume an increasingly important role in the future, he will not know what the truth is, and it will definitely be unsafe." Yu Yu said.

Data privacy in the AI era lies in whether it is transparent to users. Pei Yu said that enterprises need to fully communicate with users, what data they want to collect and what services they can provide based on these data.

Liu Jingkang pointed out that in the AI era, enterprises should be more restrained, and they can provide solutions close to the previous experience even if customers don’t provide personal information to enterprises. He is more worried about the privacy of users in the AI era. Liu Jingkang said that in the AI era, with more and more personal information collected by hardware and applications, and the ability of agents is getting higher and higher, AI may know itself better than human beings.

"When you are doing anything, there is another person who is receiving the same information and having similar thoughts with you. It has as much information as you get, thinks according to your preferences, and has stronger intelligence, better memory and more knowledge of public information than you. "

He suggested that regulators need to strengthen management and legislation. In the future, human AI model analysis and data are likely to become the same as today’s mobile phone numbers, and once problems occur, they are irreversible.

03. In the future intelligent hardware, will the big model be deployed on the end side or the cloud side?

When the big model enters the field of consumer electronics, one of the focuses of debate is whether the model should be deployed on the end side or the cloud side.

The computing power of the hardware itself is limited, and it can’t support the huge computation of the large model. Some companies tend to deploy in the cloud, but this will bring the risk of user privacy data leakage. Some companies tend to make the model small and deploy it only on the end side, but at the same time it will bring the problems of low model ability and slow processing speed.

Pei Yu believes that in the long run, the deployment of the model will definitely switch to the cloud slowly. First, the ability of the cloud is stronger, and it will become stronger and stronger later; The second is to save costs more. Most of the functions that the device can handle are not very important. "Users don’t really care whether it is the end side or the cloud side."

Liu Jingkang said that unlike the past Internet era, the computing cost in the AI era is fixed, and once AI provides a service, it will increase the rigid cost. Therefore, from the perspective of economics, the future computing will definitely move in the direction of distribution, that is, the cloud will participate more.

The only problem is data security. In the past, both domestic and overseas manufacturers have experienced cloud data leakage, and the security of users’ personal privacy cannot be guaranteed.

"From an economic point of view, the calculation of the future model will definitely go to the cloud, but from a security point of view, it can only be said that each has risks and challenges." Liu Jingkang said.