news

tencent technology author han qing
editor: zheng kejun

on september 10, apple held a new product launch conference for the fall of 2024 and officially released the first ai iphone in history. since its birth, the iphone has become one of the most iconic smart hardware in the world and continues to lead the development of the smartphone industry. however, with the rise of large-scale model technology, apple's leading position seems to be facing unprecedented challenges.

in this conference, apple demonstrated its progress in hardware and software, but compared with the breakthroughs in large model technology by companies such as openai, many people began to question: is apple already lagging behind in this large model competition? at the same time, apple has always been known for its strict privacy protection mechanism, but in the ai ​​era, the acquisition and use of data has become crucial. can apple continue to protect user privacy while promoting the advancement of ai technology? this balance is becoming increasingly difficult to grasp.

in addition, will mobile phones still be the carrier of ai hardware in the future? with the rapid development of smart glasses, watches and other devices, the status of mobile phones as mainstream hardware seems to be challenged. can apple continue to rely on the strong market share of the iphone, or does it need to explore other forms of hardware devices to adapt to the needs of the ai ​​era?

more importantly, apple faces huge technical and strategic challenges in realizing its vision of "everything is ai". from the integration of terminal products to breakthroughs in chip design, what difficulties does apple need to solve in the process of achieving this goal? these issues are not only related to apple's future, but also to the evolution direction of the entire technology industry.

in order to explore these core issues, we invited two industry veterans to give us an in-depth analysis of apple's current situation, challenges and future development direction. next, we will listen to their professional insights and interpret apple's role and prospects in this intelligent revolution.

丨guests

wang yuquan, founding partner of haiyin capital, chief consultant of frost & sullivan china

as a venture capitalist, senior marketing and strategic investment expert, he has the latest observations and thoughts on global technological innovation.

li nan, founder and ceo of numiao technology; former cmo and senior vice president of meizu technology, and president of meizu blue

as a senior practitioner in the mobile phone industry, he has a unique understanding of innovation in hardware and consumer electronics products.

zheng kejun, chief editor, tencent technology

丨key points

● is apple lagging behind in the field of large models?li nan believes that apple has completely missed the mark in the field of ai, and its traditional advantages in chips and operating systems are no longer its core competitiveness; wang yuquan believes that apple has a "gold mine" and it still has opportunities in this revolution. its core advantage lies in its control over user behavior data. even if it moves slowly in the current intelligent revolution, it still has the potential to lead mobile intelligence in the future with this control.

● apple’s ai data privacy issues:li nan believes that musk has made it very clear on twitter that when calling a third-party large model, the data must still be given to gpt. i can trust apple, but i don't trust openai; wang yuquan believes that in the final analysis, there are only two choices for privacy and security, trust or not. no matter who you trust, there will be loopholes. the important thing is to "weigh the pros and cons and choose the lesser of the two."

● will the hardware product of ai in the future be a mobile phone?li nan believes that predicting the future is extremely difficult, especially in the field of technology, and even steve jobs could not see it clearly; wang yuquan believes that in the short term, mobile phones will serve as management centers to connect various devices. the core of artificial intelligence is to provide high-end services through scale. the value of hardware and software lies in the effective delivery of these services, rather than just existing as simple tools.

● what else does apple need to achieve ai for everything:li nan believes that apple has lost the ability to inspire users' emotions. as a consumer brand, it has lost the ability to "one more thing" and "think different", which may be a bigger challenge it faces. wang yuquan believes that the core of technological revolution lies in the breakthrough of concepts, and apple's current apple intelligence is only an improvement on existing services and lacks real innovation.

apple’s “everything can be ai”?

the following is a condensed version of the live broadcast:

did apple miss out on ai, or does it still have a "gold mine"?

summary: apple's core challenge lies in the balance between hardware and software cycles. it needs to choose between taking into account the compatibility of old devices and seizing the new technological revolution.

tencent technology: how do you evaluate this press conference? what do you think after watching it?

li nan:over the years, apple's annual release cycle has been fixed, and its press conferences have always had the same problem - they are so certain that i can guess what they will do even without watching them, and there are no surprises.

wang yuquan:i don't think apple is that bad. its strategy in the past is different from what it is now.

in the past, apple was more in the stage of "evolution".hardware pre-embedded strategyit is very effective.

what it does is ensure that any new applications it launches are supported on the first three generations of phones, so some features are not immediately applied to the latest models when they are launched.

this is why we always feel that apple always launches apps later than other companies. it's not that apple is incapable of launching them immediately, but because it wants to ensure that older models are also compatible. this strategy was feasible in the past. because the cycle of changing phones is about 3 to 4 years, when three generations of phones can support an app, 80% of apple devices on the market can use the function. this strategy supports the development of apple's application ecosystem, especially the app store.

however, the changes in today's market have brought new challenges. today's revolution is different from the gradual evolution in the past. this revolutionary change forces apple to think about whether to continue with the old strategy and whether it should still wait for three generations of mobile phones to support new technologies before launching new features.

for example, the a18 chip launched by apple this time is advertised as the first generation chip that supports apple intelligence. so does apple have to wait until three generations of mobile phones on the market are equipped with the a18 chip before launching comprehensive artificial intelligence applications? if so, by that time, the market may have already undergone tremendous changes, and apple will lose the opportunity for revolution.

therefore, apple faces a dilemma: should it continue to rely on its past user base and continue with its previous strategy, or should it abandon its old users, lighten its load, and quickly follow new technological trends?this is how nokia died, but many times, later generations are unable to learn from the experience of their predecessors and will still make the same choice.

summary: when faced with risks and opportunities, founders and professional managers may have very different starting points for decision-making. cook's stability and responsibility led him to make a more conservative choice.

wang yuquan:ultimately, it's a question of founders and professional managers. tim cook is a very typical manager. we have anticipated that this is a revolution. we also expect apple to have a great advantage in this revolution.but we still know that apple will continue to adopt a conservative strategy and may miss this revolution.

tencent technology:if he were the founder, he might choose to take a gamble and go all out after grabbing an opportunity. cook has led apple for 13 years, and he may prefer steady development, ensuring a stable stock price, and being accountable to investors, while also considering maintaining the livelihoods of a large number of employees. this is something he must prioritize. especially in the context of last year when the large language model technology was not yet fully understood, cook's choice may be more cautious.

summary: apple has completely missed the mark in the field of ai, and its traditional chip and operating system advantages are no longer its core competitiveness.

li nan:the fact is that apple has missed this revolution.we all know that apple has been acquiring companies over the years, but it has not acquired any company that is really involved in large-scale language models (llm). the ai ​​that apple is talking about is completely different from the large-scale neural networks we are talking about today.

first of all, we need to clarify the concept of ai. apple has no core capabilities in the field of language models for large-scale neural networks. it is not far behind some chinese companies that rely on open source technology. although apple has some patents, such as the technology of predicting user intentions by recognizing pictures, there is no significant gap compared with other chinese brands, which is why it needs to access gpt-4o.

therefore, apple did miss the large-scale language neural network revolution triggered by the release of gpt4 by openai in march 2023. before march 2023, apple had two core technologies: chip design and operating system. but when large-scale neural networks became the center of computing power, the way of computing shifted from von neumann architecture to large-scale neural networks, which made apple's two core technologies no longer have advantages.this is the first time in years that apple has failed to capture the highest-value computing power segment.

after the release of gpt-4o, which is trained on native multimodal data, the entire field has made rapid progress. as we continue to track the performance and progress of all models, the speed we feel is amazing, just like the feeling of writing programs in the early days of the internet when i just graduated - new things appear almost every day.

the iphone is a mass-produced product with 70 million units, driving the global supply chain, and its software functions are already very complex. in this case, it is impossible for apple to keep up with such rapid technological changes without any preparation.

this shows two points: first, it is really difficult for apple to turn around during its transformation; second, it has completely missed the mark in the field of ai and appears to be very passive.

summary: apple has a "gold mine" and it still has opportunities in this revolution. its core advantage lies in its control over user behavior data. even if it moves slowly in the current intelligent revolution, it still has the potential to lead mobile intelligence in the future with this control.

wang yuquan:mr. li mentioned earlier that this apple conference was mediocre, so everyone’s reaction was normal, but apple’s stock did not fall sharply. the reason is very simple. although there were no surprises in the conference, apple’s terminal strength still exists.

i don't agree with mr. li on this.i don’t think this round of revolution is dominated by big models. big models dominated the previous round of revolution. whether it is a big language model or other types, their core is content. for example, the core value of openai lies in the fact that it uses data from the entire internet for training.

we need to expand the topic. in the era of artificial intelligence, there are no real thinkers. in the previous internet era, thinkers like kevin kelly appeared. in the wireless internet era, such thinkers did not appear, which led to a huge misunderstanding. many people think that wireless internet is just the wireless version of the internet. in fact, there is an essential difference. the core of the internet lies in the aggregation of content, so a large model will be born, which we call content intelligence.

but wireless internet is about the collection of actions. it is the first time that human behavior is collected and analyzed on a large scale. so we call it terminal intelligence, which is action intelligence rather than content intelligence. in the field of action intelligence, mobile phone manufacturers have a lot of control because user behavior is recorded and analyzed by mobile phone manufacturers. i think as long as this action control is not broken, apple's advantage will still be obvious.

although apple has not yet made any plans for mobile intelligence, and may not even be fully aware of it, by the time it really wants to do it, other companies may not be able to surpass it. therefore, apple's advantage is still firmly in its own hands.

now most of our mobile phone behaviors are realized through apps, and the underlying layers of these apps are actually dominated by apple. if apple wants to connect these apps, it has a chance, but if it doesn't want to, the possibility of third-party apps from the outside world wanting to connect is very small.

this means that apple has a "gold mine", and if it doesn't dig it, no one else can.i saw similar situations when i was a mobile consultant before. it wasn't until one day when wireless internet was truly opened that the fence fell down, giving others a chance. it may take some time for this fence to fall down.so apple’s logic is simple: if it doesn’t drive the revolution, it will be difficult for the outside world to start a revolution.

openai's "revolution" is more of a continuation of the logic of the internet. i don't think it is the essence of terminal intelligence. currently, multiple apps running on mobile phones are separated. we have all experienced this pain in the past, such as the increasingly inefficient search function because the content is scattered across various apps. when your behavior is also decentralized, any platform's understanding of you is incomplete and cannot be integrated.

however, apple's understanding of users is becoming more and more complete. through mobile phones, watches and other devices, apple has a very thorough understanding of user behavior. therefore, its confidence is not unfounded. although this confidence may not be a good thing for us users, it does exist.

summary: the transformation of computing power is changing the landscape of software and hardware. the core competitiveness lies in providing users with a unique and high-value experience. openai is currently ahead of apple in this regard.

li nan:i work in the product field, and i like to explain the problem in the simplest terms. when we make hardware products, we often say: "software determines hardware." in fact, there is another sentence that we usually don't mention, but it does exist: "computing power determines software." in the past few decades, computing power has been based on the von neumann architecture, so there is no need to say more about this sentence, but the complete logic is like this. today, computing power has changed, and software has changed with it. if software like photoshop dominated in the past, now it may be ai software like midjourney. therefore, when the software changes, the hardware will inevitably change as well.

i would like to emphasize two points:

first, computing power determines software. when computing power undergoes revolutionary changes, software will also change accordingly. there is a high probability that there will be more than one company providing these software.

second, software determines hardware. when software changes, hardware will also change accordingly, and there will most likely be more than one company providing hardware. at the same time, i want to emphasize the last point:the so-called "entrance theory" is not supported by user value. you don't become powerful because you master the entrance, but because you provide users with transcendent value. you can't reverse the cause and effect.the core is to provide unique and highest value capabilities. objectively speaking, it is openai, not apple, that can provide this capability today. i no longer use an iphone.

i also want to share a survey result: we have tens of thousands of real consumers in our discord group in the united states, and some of them have bought meta glasses. we surveyed their dissatisfaction with meta glasses, and one of the reasons was that meta glasses could not use gpt-4o. meta glasses sold 2 million pairs, but in fact, consumers are not buying meta glasses, but llama. however, consumers feel that llama3 is not good enough, so they want to use gpt-4o.

summary: mobile terminals have great potential in behavioral intelligence and can directly execute instructions, while openai's gpt-4o is currently limited to auxiliary functions at the content level.

tencent technology:in the ai ​​era, the combination of mobile terminals and large model platforms has gradually become a focus. do you think apple's openness in this field will affect its competitiveness? how will the potential of mobile terminals play a role in this competition?

wang yuquan:i think,when a revolution comes, it is dangerous to use inferences from the past to predict the future.the best tool right now is of course openai, and i agree with that, but the premise of thinking that openai is everything is that there is no foreseeing that there may be a new revolution in the future. and i think there is a revolution. let us not forget that a hot topic recently is data monopoly. i just wrote an article about "digital feudalism" a few days ago, talking about the monopoly that is being formed by various digital platforms.

for example, i hope that my phone can complete tasks by calling my app, such as i can tell siri to "order food for me" and it can do it directly. but to be honest, openai can't do this at present, because the underlying api is difficult to open to third parties, while mobile terminals can do it. apple has not yet fully opened its api to third parties.

however, apple's mobile phone has already achieved a lot of underlying technology, such as video processing. in fact, what i am most looking forward to is that openai's gpt-4 can directly open the camera, recognize video content in real time and interact with me. but at present, it only has the function of taking pictures, and the video recognition function is not open. but if we look back, who can do this more easily? the answer is obviously the mobile terminal. so,mobile phones have great potential in executing many of our behavioral instructions, and gpt-4o is currently more like a consultant or advisor. i think content and behavior are two completely different things.

li nan:regarding this issue, i talked to two young people from stanford, the founders of "octopus". they reduced the end-side model to about 1b through distillation and pruning, and are learning the android 12 api. they are indeed worried about whether they can learn the ios api and whether apple will allow it? i told them: "the person who doesn't let you learn is the one who may be defeated by you."

we have seen this countless times in the era of open source software. you can help huawei beat apple, or help android beat ios, because you can provide a better end-to-end ai experience than siri and help perform tasks.whoever is closed will be attacked by the open source world.the final result is uncertain, but fifty-fifty is a common result, and you may be defeated.just like oracle today, oracle no longer has the influence it once had.

summary: the future of ai technology is full of uncertainty, with success and failure intertwined. the form of terminals and ai is still evolving rapidly, and the future is full of unknowns and possibilities.

tencent technology: currently, many new ai applications are emerging, but the success rates seem to be minimal. what do you think of this phenomenon?

li nan:the gpt4 revolution started in march 2023, and it is now september 2024, which is just over a year. in this period of more than a year, the possibilities are endless. we know that in the late stage of the highly mature internet industry, the winning rate of investment is very low, and capital believes that there is only a 1% or less chance, which may be even lower today. therefore, 100 failed ai application cases are entirely possible, but it does not mean that the one in a thousand successful cases does not exist. we have seen some examples, such as meta glasses sold 2 million pairs in less than 12 months, and it really provides real value to users.

wang yuquan:it is very difficult to use history to predict the future. before the emergence of bat (baidu, alibaba, tencent) in the internet, bbs (forum) was very popular and everyone was discussing it. later, many people said that bbs would die out, and the facts proved to be so. the key question is, what will bat look like in the future? i think there is little exploration and discussion in this regard.

we often say that any new technology is like a tank, but the key is how to fight a "blitzkrieg". openai's chatgpt is indeed a "tank", but what is its "blitzkrieg"? i don't think i have found it yet.

the only application that has a certain scale may be ai search, but its form may be very different from that in the future, and it has not yet been finalized, which is what makes it interesting. therefore, many forms of terminals may be very different from the direction we discuss today.

summary: the success of future smart terminals depends not only on performance improvement, but more importantly on the actual user experience. the difference between mobile phone applications and internet applications will determine the future development direction.

tencent technology: how do you view the key factors of future "blitzkrieg"?

li nan:speaking of "blitzkrieg", the rapid rise of the iphone is a famous blitzkrieg in history, and it has a "golden line" for sales. we believe that the actual sales of the iphone should exceed 700,000 units, and the supply chain orders exceed 1.4 million units, so any product that does not reach this line, we believe, is not a new revolutionary prototype. however, meta glasses have exceeded 2 million pairs in sales. although i dare not assert whether it is really successful, at least it has reached a basic standard line and demonstrated an effect similar to "blitzkrieg". i think,in future competition among mobile phone manufacturers, the intelligence level of the model will be decisive.

wang yuquan:i think what we are seeing now may not be the same in the future. as someone who has experienced the wireless internet era, i also worked as a mobile energy consultant at that time.at the beginning of that era, everyone paid great attention to technical indicators such as performance, but later we learned that performance was no longer a limiting factor at a certain stage. therefore, this is also a prediction and it is impossible to be completely accurate, but we should remain confident in the continued improvement of performance.

back to the core of the problem, i think the future "blitzkrieg" still depends on the actual user experience. when performance is no longer a bottleneck, the real obstacle is the user's willingness to use it. from this perspective, mobile terminals are the key. in fact, we have just taken the first step. we are mainly discussing the transfer of existing openai business from the internet to mobile phones, but we ignore that mobile phones and the internet are completely different. applications on mobile phones and internet applications may be very different, and research in this area is not deep enough.

data privacy: a game between value and risk

tencent technology:next, let's talk about the issue of privacy and security. previously, the two of us had different views on this topic: yuquan recognized apple's security mechanism, while li nan was skeptical about apple's privacy protection. today we can talk about it in detail.

summary: privacy and security are unavoidable risks in technological development, but when the value of technology far outweighs the risks, people tend to continue using it. platforms that control data, especially companies like apple, may take the lead in future competition.

li nan:i can trust apple, but i don't trust openai. i think musk is telling the truth. the issue of user privacy is actually a false proposition. apple emphasizes privacy protection, which is actually more of a way of showing its attitude to users, "i care about you and i'm here to serve you." so i think it's better to just listen to it.

wang yuquan:i don't think it's important at all.ultimately, there are only two choices: trust or not. no matter who you trust, there will be loopholes.so far, i have not seen a system that is completely free of vulnerabilities. perhaps in the future, a centralized computing platform built on decentralization will be able to solve this problem. now, many people are studying technologies such as federated learning, but there is still a long way to go before practical applications. this is not only a technical issue, but also involves whether the platform can be established and whether the application can succeed. before that, you can only choose to use the services provided by these centralized platforms.

strictly speaking, these platforms all have vulnerabilities, and even in some aspects, westerners think that apple's vulnerabilities may be greater, especially in china, where apple also relies on cloud guizhou. all of these are hidden dangers. if you think there are problems, you can choose not to use them, but the problem is that if you don't use them, you will fall behind. so most people still choose to use these technologies. we will find thatmany times the technologies we use are not that safe, but when the technology develops to a certain level, the security issues are gradually diluted. so i think that as long as a certain security standard is met, these technologies should be used, because the benefits and value they bring far outweigh the risks brought by security.

the question now is "weigh the pros and cons and choose the lesser of the two." on the one hand, there is the so-called security and privacy, which may result in the leakage of some user privacy; on the other hand, all privacy is controlled by a certain platform, such as apple.which side will you choose? the "technological feudalism" we discussed earlier is more harmful because it leaves you no chance of a comeback. this is why i think apple may be ahead in the next round of competition. it has too much control over user data. apple can choose not to share data with others, but can let its own applications use this data, just like how internet explorer defeated netscape.

summary: although apple's conservative attitude towards privacy protection and data supervision has maintained user trust, it may also limit its development potential in the field of terminal intelligence, especially when it comes to key technologies such as ai agents and api openness.

tencent technology:so at present, if we want to find a technology giant that can protect user privacy better than apple, who else can it be? in addition, apple is indeed a consumer brand, and many times its actions are for communication effects. as for how it specifically processes and uses this data, we actually don't know.

wang yuquan:the key is,the boy who slays a dragon may become a dragon.when i protect myself particularly well and all users flock to me, i have enough say. guess what i will do?

of course, to be honest, i don't think apple is completely engaged in dissemination. i think that now, especially in europe, data supervision is becoming more and more stringent, and apple may have delayed the launch of many new technologies because of its cautious attitude towards data monitoring.

for example, in the field of artificial intelligence, everyone is talking about big models, but i am more concerned about ai agents, which can help me complete tasks. a core question is whether i can call other apis or even other apps on my phone. in the internet field, the core of api is to be able to call other applications. i think,apple's decision not to open these features is likely due to concerns about privacy, regulation, and possible fines. since no one else has opened these features, apple may not open them for a long time, but this also means that the functionality of the terminal's intelligence will be greatly limited.

summary: li nan believes that android open source and market diversity will promote the implementation of technology, and even if apple is not open, it will not hinder innovation. wang yuquan pointed out that whether it is apple or android, the underlying functions are not open, and cross-app calls still face technical and regulatory barriers, which limits the real intelligent behavior analysis and application.

li nan:no, the android camp is open source. even if apple does not open certain functions, we can demonstrate the results through technical demos, and eventually some brands will implement them and achieve market success. apple's global mobile phone shipments only account for 14%, and the remaining 86% are mobile phones of other brands. therefore, technological progress will not be hindered by apple's failure to make certain functions.

there will always be radical brands that gain advantages through open source ecosystems in different countries, markets and brands. i believe that this diversity will promote the realization of functions. for example, octopus is using the api of android open source code for training. although some apps do not open interfaces, we can still realize functions through other means, such as simulating gui operations through apple's patents, or using web urls to call functions. although the app ecosystem has limitations, it does not mean that needs cannot be met.

wang yuquan:the open source camp has the same problem. even if it is open source android, if it does not allow you to call other apps from the bottom layer, you cannot implement cross-app calls. it is not open to that extent. if apple, as a mobile terminal manufacturer, does not open many underlying functions, third parties cannot call them. not only apple, but other brands will also face the same regulatory issues. therefore, even if apple phones only account for 14% of the market share, other brands will be subject to similar restrictions. the core issue is whether mobile terminals can now implement smart operations such as "order food for me." ordering food requires third-party apps to open data to analyze user behavior, and many apps are unwilling to open these data.

although there are other solutions, such as simulating the operation interface or calling the web url, these cannot really solve the core problem of calling the internal data of the app. even if some restrictions can be technically bypassed, the platform does not open the data interface and cannot achieve in-depth behavioral analysis. the greatest value of the mobile phone to me is to analyze my behavior and make decisions, and these behavioral data are not open now. if the terminal manufacturer does not open the underlying functions, even if there are technical means, it will be difficult to achieve the intelligent operation i expect. simply put, can you design an application to monitor my behavior every time i call the app on my mobile phone, and know what i did in wechat?

summarize:in the future, demand mining will go beyond the limitations of apps, and obtaining user behavior data through technologies such as natural language processing will become a new direction. apps are not the only solution, and natural language and new data collection methods will bring deeper understanding of users.

li nan:in the past 20 years, people’s needs have indeed been mainly reflected in the behaviors in apps, but the case of rewind has given us some new ideas. the concept of rewind is to record all your sounds in a day through a pendant, and plaud provides an ai recorder to record meetings when needed. these applications tell us thatalthough we are used to expressing our behaviors through apps, human needs and desires are not limited to apps. we have more natural and fundamental ways of expressing ourselves, and we cannot assume that apps are the only solution just because we have used them for 20 years.

apps are not the best solution, and they can’t be. if an app is not open, we can choose not to use it, but that doesn’t mean we can’t discover needs or solve problems.plaud can record your meetings, rewind aims to record your day's audio, and in the future there may be technology to record visual information. when we have this data, we will be able to understand users more deeply, even if we don't know exactly what they do in the app.

natural language processing capabilities can proactively initiate conversations and even directly ask users about their needs, which is a capability that apps do not have. therefore, there will be new ways to tap into demand in the future. these data will be extremely valuable, but they do not necessarily come from the operating behavior of apps. although the data in apps is valuable, i am skeptical about whether all apps can truly utilize this data and implement ai.

however, this also brings another possibility. for example, we are working with dingtalk, whose founder is very familiar to me. they are trying a new direction - training app behavior data through ai, and then using this ai to serve users. when an app can understand natural language and call its internal functions, it can communicate with other apps. therefore, in the future, even if we don’t know the internal behavior data of all apps, we can obtain information through natural language, such as "what time does this user have a meeting every day?"

and within 8 months, i believe you will see a scene like this.

summary: in the future, mobile phones will go beyond their current functions and become tools like smart consultants, helping users quickly solve problems and conduct analysis. although small companies have technical capabilities, users trust large companies such as google and apple more because they can ensure data security and provide stable services. the combination of mobile phones and artificial intelligence will completely change the way we interact with devices.

wang yuquan:everything i just mentioned is correct. for example, you can use ai tools to record videos, analyze apps and perform actions. these technologies can be achieved regardless of whether they come from large or small companies.but the question is, would you trust a small company?think back to how android rose to prominence. there were many people working on mobile operating systems at the time, but android was the one that ultimately succeeded because it had google behind it. google promised to maintain the system for free, and everyone trusted it. small companies, however good their technology was, could hardly win the trust of users. mobile phones are very private devices, and users want their phones to be secure.

secondly, if the operating system opens its api, there will definitely be supervision. large companies can ensure security, but it is actually very difficult to let a third party continuously record your behavior. therefore, our discussion is based on several basic points.first of all, i think mobile phones are going beyond their current basic functions and are no longer just used for communication and entertainment. in the future, they will take on more tasks.for example, i often need to do research. in the past, i would open my computer, search on google, and organize the information. now, i am used to using gpt-4o. there is a voice interaction function on my phone, and i can ask gpt-4o questions at any time, such as "what do you think about this question?" or "help me organize the information." it can give answers immediately, which greatly speeds up my research efficiency.

the only problem at present is that the results sorted on the phone need to be imported into the computer, but i believe that in the future, companies such as apple will soon realize this and may implement this function directly on the computer. the addition of artificial intelligence has made my research process much faster. it feels like i have a powerful consultant to assist me at any time. it has a powerful database behind it, which can quickly provide the required analysis and conclusions.

therefore, with the development of artificial intelligence, the way mobile phones are used will be completely different from the past. if we only solve problems from the perspective of existing mobile phones, we will not truly understand the potential and use of future mobile phones.

summary: it is impossible for super apps to connect data in the short term, but with the development of ai and big models of apps, data interoperability may be achieved through communication between apps through natural language in the future.

however, in the short term, competition is still mainly concentrated among giants, and the true realization of decentralization may still be a long way off.

zheng kejun from tencent technology:another question is, will apps be willing to share data with each other in the future? or will they be more like talking to each other through a security door, but keeping their data locked inside? i find it hard to imagine what this scenario would be like.

li nan:no need to think about it, it definitely won’t work.apple's data belongs to apple, and dingtalk's data belongs to dingtalk. the app ecosystem itself is a problem, and the app format will inevitably face challenges in the future. however, many apps are moving towards ai, and applications like notion are already moving in this direction.

in the future, apps may be able to communicate with each other through ai. for example, midjourney and notion can communicate through natural language, so users don't need to input complicated prompts. if all apps can understand natural language and generate language, why can't they communicate with each other? once the conversation is realized, the data is equivalent to being connected. i think we may see similar applications appear within 8 months.

wang yuquan:in fact, it is not difficult to imagine that all major platforms are based on large models, and there is already a unified communication language between large models - tokens. one platform inputs tokens and outputs tokens, and the other platform does the same. in theory, they can communicate with each other. but the problem is that the standards must be unified enough.

openai's platform itself has great potential, such as the previous gpt store, but its management experience is obviously not as good as the app store. the gpt store simply displays applications, while the app store opens many ports. although decentralized app interoperability may be realized in the future, i think it will still be a competition between giants in the short term. large models may bring new unified standards at the application level, which may be the direction of the future.

the future of ai hardware:

mobile phones are still the core, but the variables are still unpredictable

summary: in the short term, mobile phones will serve as management centers to connect various devices. the core of artificial intelligence is to provide high-end services through scale, similar to ford's mass production of cars. the value of hardware and software lies in the effective delivery of these services, rather than just existing as simple tools.

tencent technology:will mobile phones become the best hardware for ai to interact with the physical world? what will be the next generation of product paradigms?

wang yuquan:the discussion about the hardware form after the mobile phone has actually started a long time ago. although the intelligent era has just arrived, the functions of mobile phones have been widely studied long before. at that time, there was a popular concept called pda (personal digital assistant), but we proposed that the positioning of mobile phones should be higher, called "personal intelligent assistant" (pia), because its computing power far exceeds that of traditional digital assistants. even today, this definition is still applicable, and mobile phones are still very powerful devices.

we also mentioned the concept of "body area network", which refers to the network formed by devices such as headphones, watches, and laptops through mobile phones. although the computing power of headphones and watches is weak, they process and transmit data through the mobile phone as a central device. in the future, the mobile phone may no longer be the main interactive interface, but a background center for managing and processing other devices. the interactive interface may be a vr device or other peripherals.although the devices we carry are becoming more and more diverse, the mobile phone as a "personal intelligent assistant" remains at the core.

as we further explore the future,the core capabilities brought by artificial intelligence are very similar to those of the industrial revolution. the core capability of the industrial revolution was the large-scale production of complex products, such as cars, while the core capability of the intelligent revolution is the large-scale provision of services.in the past, many services could only be provided by humans, but now artificial intelligence can already achieve this. the value of hardware and software lies in the fact that they deliver services and meet user needs. if hardware and software cannot effectively deliver these services, their existence will lose its meaning.

take ford motor as an example. in 1908, ford began mass production of cars, which were sold for only $950, which is less than $10,000 today. tesla does not have a car that can reach this price today. at that time, ford reduced manufacturing costs through large-scale production, making cars more popular. similarly, our goal today is to provide high-end and scarce services on a large scale through artificial intelligence to meet the needs of the public.services that were once only available to humans can now be provided by artificial intelligence. hardware and software are just the carriers of these services, and technological progress is to better deliver these services.。

summary: predicting the future is extremely difficult, especially in the technology field. even steve jobs couldn't see it clearly. the key is to act, keep trying, and be flexible in the short term, rather than trying to control the long-term future.

li nan:i have seen many people in the hardware industry who try to predict the future and gamble with their products. usually, they lose 100 million rmb in one failure, and this is just a few cases.

it turns out that probably no one on earth can see more than 12 months into the future, because i have seen too many chinese bigwigs fall.

i once saw an interesting interview where the boss of an early pda and smartphone manufacturer talked about a meeting he had with jobs. at the time, they couldn't make a high-quality smartphone, so they went to apple for negotiation. in the middle of the meeting, jobs drew a picture of a macbook, and then surrounded it with various devices, such as speakers and mobile phones. jobs said, "look, your mobile phone can be added to the periphery of this circle." but another person said, "you are wrong, the mobile phone should be the center of this circle." this story tells us an important truth: even a genius like jobs may not be able to see the future clearly in the early stages of the mobile internet revolution.

so, if someone says today that he can be sure that the mobile phone will still be the center of the future, this prediction is a bit beyond human ability. however, i believe that there is indeed an opportunity to shake the central position of the mobile phone.

therefore, the most important thing is not to predict the future, but to try more without failing. maybe a project can make 160 million rmb in 6 months, just like meta glasses have broken the sales gold line of iphone. we should focus on action and product release, rather than over-predicting the future.those who can see clearly for 12 months are almost gods. if you can avoid killing yourself within 6 months, you are already very good.

summary: from the perspective of industrial design,the iphone 16 has enhanced its photography capabilities through features such as 4k 120hz video shooting, which meets the needs of today's video era. the addition of a shooting button reflects "software defines hardware" and improves the user experience.

tencent technology:at this apple conference, a new control button was added, which can directly use the visual search function. do you think this is a reasonable interaction design? why did apple add this seemingly traditional hardware design when siri can already be quickly awakened?

li nan: judging from steve jobs' design style, he would rather remove all the buttons.i was involved in a product that won the if gold award, and we also tried to reduce the number of buttons in the design.

however, this time apple went the other way and added a button. but i don't think this design is completely unreasonable. from an industrial design perspective, the most important upgrade of iphone 16 is 4k 120hz video shooting, real-time adjustment and mixing functions. these enhancements are obviously intended to strengthen the iphone's capabilities in photography and high-quality video output, which is in line with the needs of today's tvc and vlog era.therefore, since the software has so powerfully enhanced these functions, it is reasonable to add a shooting button to allow users to use these functions more comfortably. this is the embodiment of "software defines hardware" and provides real value to users.

as for whether it is related to vision pro, i personally don't think it has much to do with it. maybe apple wants to link them together, but in my opinion, this is just wishful thinking on apple's part.i don't think vision pro has any substantial success. no matter from which dimension, this product has failed. there is no evidence to prove its success. as an apple user for 20 years, i can't prove its success.

tencent technology:perhaps vision pro is still in the process of iteration, and its battery life and weight may be improved in the future. if these problems are solved, it will have wider applications.

li nan:what you said makes sense, but it requires apple to show its ability to "do subtraction". i agree that there may be improvements in the future,but now it seems that apple has not demonstrated enough subtraction capabilities in many of its products.

summarize:the significance of the new button depends on whether apple has deeper plans, but if it can enable smart assistant features, the button will have value.

wang yuquan：i think the key lies in apple's next move, and its true intentions are still unclear.

while it talks a lot about apple intelligence, the applications it shows seem very childish. so i doubt whether there is a follow-up plan. if there is no follow-up, this button is meaningless; but if there is a follow-up, this button may be useful.

in theory, adding a button means that this function will be used frequently. if this button is just for camera, i don’t think it should be designed this way, but if it is for visual intelligence, for example, if you press this button, the camera will not only turn on, but also recognize objects in real time and start conversation support, then this button makes sense. especially if it can start a powerful smart assistant function, then this button becomes an assistant button, and such a design is valuable.

tencent technology:its multimodal function may not be limited to shooting. the multimodal function i imagine should be more comprehensive. it is not just about pressing a button to identify an object in the frame. it is like a friend who is always by your side. you don’t even need a wake-up word. you can interact at any time as long as you carry your mobile phone.

wang yuquan:i think it still needs a wake-up mechanism. recording you anytime and anywhere is scary. there must be a wake-up signal to let users know clearly that they have started the interaction and acquiesced to the device recording status. although users may not actually have real control over the information, they must appear to regulators as if they have control.

what problems does apple face in realizing ai for everything?

tencent technology:if apple wants to achieve "ai for everything", from terminal products, chips to the ecosystem combining software and hardware, what difficulties still need to be overcome? which ones need to be solved now and which ones need to be solved in the future?

summary: apple has lost the ability to inspire users’ emotions. as a consumer brand, it has lost the ability to “one more thing” and “think different”, which may be the bigger challenge it faces.

li nan:one clear challenge apple faces in the future is its competition with qualcomm. apple has been developing its own communication baseband and trying to integrate it into the soc. although it is unclear whether this goal will change in the ai ​​era, it has been working hard in this field for many years. in china, many users often complain about the signal problem of apple phones. this baseband problem is indeed a problem that apple has had to overcome for many years.

in addition to the communication baseband, apple has great potential in the technology of the end-side model. apple has strong chip design capabilities and can provide more support for the end-side model at the hardware level. we have already seen its improvements in this regard on the a18 chip, and the end-side model capabilities of the iphone will become stronger and stronger in the future.

but i think the bigger problem apple faces is that it has lost its ability to excite consumers. this problem is bigger than just the technical difficulties.today's apple is no longer able to drive consumers' emotions like in the past, and even the release of new products cannot excite users. for example, i saw huawei's three-fold mobile phone today, and i felt more excited than apple's press conference.

apple has lost its ability to be “one more thing” and “think different”, and as a consumer brand, it no longer resonates with users emotionally. this is a more serious problem than a technical challenge, especially in the consumer field, where the act of adding items to a shopping cart is driven by rationality, but users are emotionally driven when it comes to final payment.

when the real computing revolution happens, what apple should do most is to regain its original “think different”.

summarize:the core of technological revolution lies in the breakthrough of concepts, but apple’s current apple intelligence is only an improvement on existing services and lacks real innovation.

wang yuquan: to be honest, the biggest challenge is the concept. the real leaders of every technological revolution are often the revolutionaries of concepts.apple completely revolutionized the music industry with the ipod because they didn’t consider themselves a hardware company. i was particularly impressed by the launch of the iphone. in his first speech, steve jobs said that it was a combination of the ipod, internet navigator, and phone. this conceptual upgrade fully demonstrated apple’s technological advantages.

but now, apple intelligence is more about upgrading apple's existing services rather than fundamentally promoting a technological revolution. i think this is a huge limitation. it still revolves around how to improve apple's existing services through intelligent technology, rather than thinking about how to use apple's hardware to support a wider range of intelligent needs. the key question should be how artificial intelligence can really help people solve problems, rather than simply improving functions such as cameras.

both software and hardware should serve user needs. the problem now is that apple has not demonstrated any real new breakthroughs.take the camera as an example. if the camera can become an opportunity to expand the functions of the phone, then it is the right direction. but if apple intelligence is only used to help the camera, then it is the wrong direction. so,at present, the "revolution" that apple has been calling for for a long time has not really shown revolutionary performance.