Apple Intelligence plays a word game and conceals the nature of AI

We once imagined the bright future brought about by the advancement of productivity: Because of the great abundance of materials, people no longer need to engage in too heavy specific labor, food, clothing, housing, and transportation have become readily available, and humans have more energy to engage in literary and artistic creations such as writing and painting. , become rich both materially and spiritually.

Then, after the emergence of this wave of AI, which represents the advancement of productivity, people suddenly discovered that they had thought wrong before: What if AI can help us with laundry and cleaning? How come I am still doing housework, but AI is doing creative work?

If you pay AI incorrectly, AI will respond with "illusion".

AI, an agent of power

Just imagine, if there was one phishing website in every ten Google search results, wouldn’t we be extremely angry?

The probability of generating nonsense represented by ChatGPT is really high, but we seem to be very tolerant of new things and take the burden of verification ourselves.

Logically speaking, the many AI tools that are emerging today, such as ChatGPT 4, Midjourney, and Sora, are destined to make many people unemployed, but it is not like the three-click warning "Don't answer, don't answer, don't answer." We The requirement for AI is to "answer early, answer early, answer early."

Progressives are like "Adventists" who look forward to more advanced intelligence coming to the earth, but they lack plans for the chain reactions that technology may cause.

In the movie "Oppenheimer", Oppenheimer, the father of the atomic bomb, confessed to himself after the atomic bomb exploded, thinking "I became the god of death, the destroyer of the world."

The inventor of technology is not the owner of technology. Technology like nuclear fission ultimately belongs to power and represents power.

Kate Crawford, a researcher at Microsoft Research and co-founder of the AI ​​Now Institute at New York University, believes in "Beyond Technology" that because AI needs to be calculated in Nvidia's graphics card, it also requires a lot of power capabilities. , it also requires massive knowledge and data training and learning. Otherwise, AI will be nothing. But precisely because AI relies on such resource-intensive entities, AI technology is also an agent of power.

By analogy, AI is the atomic bomb that has not yet exploded. Or to put it mildly, if AI is another nuclear fission technology, then it can be an atomic bomb or a nuclear power plant.

One data that market research organizations such as QuestMobile like to list is where do Chinese netizens spend their time online?

▲ Chart from QuestMobile

The answer is that in recent years, Chinese netizens have devoted about 60% of their mobile Internet time to mobile video and mobile social networking.

▲ Chart from QuestMobile

Considering the dominance of Douyin and WeChat in their respective fields, it can be inferred that Chinese netizens spend an estimated 50% of their time surfing the Internet on their mobile phones with Douyin and WeChat. It is also easy to predict where the remaining time goes, including shopping. Taobao, Pinduoduo and JD.com, Tencent Video, Bilibili, iQiyi in long-form videos, Honor of Kings and Genshin Impact in games, Toutiao in news, and the rest may be Meituan’s takeaways from Ele.me, Amap Maps, Baidu map navigation and so on. It can be said that major Internet companies such as Tencent, Bytedance, Alibaba, and Baidu have almost complete control over our mobile Internet time.

In the PC Internet era, there were large Internet companies, but there were no so-called Internet giants. People's time was diverted to various websites and games, and the degree of head concentration was not so high.

But don’t forget that WeChat was the winner in Tencent’s internal horse racing, as well as external competition such as Fetion and MiLiao. At that time, Youku, the number one in the industry, and Tudou, the number two in the industry, merged to become the third in the industry. Tencent Video and Bilibili were the ones catching up from behind. ; Meituan Waimai also fought fiercely with Ele.me and Baidu Waimai for many years to gain the leading position; Pinduoduo grew up in a corner that no one cared about, and finally went from being looked down upon to incomprehensible, and finally became unable to catch up; no one Giants are born.

As mentioned before, AI at the current stage relies on resource-intensive entities, and small players have no opportunity to participate. For example, there are six leading large-scale model startups in China: Dark Side of the Moon, Minimax, Zhipu AI, Baichuan Intelligence, Stepping Stars and Zero One Thousand Things, originally founded by Meituan co-founder Wang Huiwen, should have been the top players in Light Years Beyond, but they were eventually acquired by Meituan. Interestingly, Alibaba has invested in 5 of these 6 companies; Tencent has invested in 3 of them, and also invested in Light Years Beyond; Meituan, which owns Light Years Beyond, also invested in Dark Side of the Moon and Smart Spectrum AI; then Lei Jun Capital (Kingsoft, Xiaomi and Shunwei) also invested in Lightyear Beyond, Baichuan Intelligence, Minimax, Wisdom AI and Dark Side of the Moon; top funds such as Sequoia and Hillhouse are also long-term Place your bet.

It seems that Baidu, Tencent, Alibaba and Meituan all have their own big models, but they still make large-scale external investments to avoid risks and increase the probability of winning, because they all believe that they cannot afford to lose.

The situation overseas is roughly similar, either the model of Microsoft + OpenAI, or the model of large models such as Meta's self-developed Llama. From the beginning, this has been a game of giants, a game of power, and another centralization of technology and the Internet. sports.

Recently, a picture has been widely circulated on social networks at home and abroad, comparing the different understandings and contributions of Zhou Hongyi and Huang Renxun to AI.

For example, Zhou Hongyi promoted 360AI by saying that it can turn normal pictures into color pictures, and some people reported that the normal picture was stolen by 360AI. Huang Renxun won’t go into detail about AI. NVIDIA, currently the second largest company in terms of market capitalization in the world, has become the biggest winner in this AI wave.

The comparison between the two sides clearly illustrates that the power to master the ability of AI can determine which direction AI will go. It is true that technology neutralists will say that technology is unconscious and has no position, but deepfake, AI onomatopoeia and other technologies have already become assistants for crimes such as fraud, and many science fiction movies have also begun to set up a kind of omnipotence in the online world. No, the "super AI program" that can launch nuclear bombs and cause satellites to fall can be used as movie clues. In contrast, decades ago, such movie clues were usually a nuclear bomb, a can of super poisonous gas, or a tube of world-destroying virus.

The paradox of AI does not stop there. When PC manufacturers and mobile phone manufacturers have given AI a name and started marketing again as AI PCs or AI mobile phones, consumers have remained indifferent. The shipments of computers and mobile phones have The volume has not changed due to the addition of AI functions.

On the contrary, ordinary people will even feel disgusted with products that are very High Tech. It seems that High Tech and Low Life are always together. So many AI incidents last year were labeled as "the world will change when you wake up" and "AI will make mankind completely defeated". "Human beings in the future will be divided into two types, those who can use AI and those who cannot." After that, it is difficult to say that the word AI is very offensive in the minds of the public. At the very least, it is like a fierce horse. It takes courage to tame.

Embracing change and accepting innovation is a difficult proposition for most people. More people are forced to follow the technology wave. Product managers, psychologists and actuaries carefully calculate every psychological activity of users, allowing users to indulge in endless scrolling of short videos and failure and victory in games. And in the dopamine cycle of krypton gold. Going forward, who can guarantee that they are not the Theodore who fell in love with AI in the movie "Her"?

Apple released Vision Pro at WWDC last year, which almost coincided with the release of GPT-4. I wrote an article at that time, " Apple Vision Pro and GPT-4, ushering in the era of "intellectual surplus ." My judgment at that time was that Vision Pro This kind of portable device is the best hardware carrier for AI such as GPT-4. The two will definitely merge and produce transformative power. Humans will have powerful external intelligence, just like humans driving vehicles with powerful horsepower. Same.

Even though I am not a technological optimist, I do still believe that "human beings in the future will be divided into two types, those who can use AI and those who cannot use AI." But what I am conservative about is that new technologies truly create a watershed. It will take longer, rather than a sudden gust of rain.

It's like there was manual intervention behind the stunning launch of Sora, and a year later, everyone's interest in Vision Pro has waned. The future is always swinging between certainty and uncertainty, and the only answer is Nvidia stock.

It was Nvidia founder Jensen Huang who held the most powerful power in the AI ​​​​circle and gave the most threatening speech:

Whether it is running for food, or running not to be regarded as food by others. You often don't know which situation you are in, but no matter what, keep running.

Apple Intelligence makes sharp technology gentle

Before talking about Apple's homophonic jokes and changing Artificial Intelligence to Apple Intelligence, I was actually thinking of a marketing overturn incident by Apple not long ago.

After releasing the thinnest iPad Pro 2024 ever, Apple took advantage of the situation and launched a commercial: a hydraulic press representing absolute power crushed interesting things such as books, musical instruments, paints and dolls to pieces. Finally, under the heavy pressure, These embellishments in daily life have become a thin and light iPad Pro 2024.

After watching the commercial, most viewers did not feel the charm of technology, but regretted that the hydraulic press, which represented the power of technology, destroyed those original, interesting, and colorful daily playthings, turning them into a cold and thin machine. tablet pc.

▲ Apple iPad Pro 2024 advertising screen

Apple later deleted the ad and apologized.

AI technology makes it easier to use technological effects that subvert cognition to create a similar sense of "authority." Microsoft Office Copilot demonstrates automatically generating PPT, OpenAI demonstrates Sora automatically generates videos, or GPT-4o performs flirting skills, all in The "sense of authority" that shapes this technology makes humans feel ashamed of themselves and submit to them.

The "Luddite Movement" is an inconspicuous part of history books. "Luddites" who are anti-advanced technology are often synonymous with hindering progress, but technological introspectors also have enough reasons to describe "technological traps."

For example, historian Yuval Harari proposed a hypothesis in "A Brief History of Humankind": Human beings moved from a gathering society to an agricultural society and domesticated plants and animals. This seems to be a kind of social progress, but it may also be a scam. , it was not humans that tamed wheat, but wheat that tamed humans.

A long time ago, wheat was just a weed, but after it was domesticated and optimized for large-scale cultivation, human food sources became single. Repeated mechanical labor caused humans to suffer from intervertebral disc herniation and arthritis that they did not have before, and they were forced to settle down. Near farmland, the ability to resist natural disasters becomes extremely weak, and famine can often kill millions of people. The real winners are wheat, rice, and potatoes. As living creatures, their populations have grown to unprecedented levels.

There are many examples of more recent "technological traps". For example, humans have invented various electrical appliances such as washing machines, dishwashers, and floor washers. However, few people say that because of these appliances, life has changed. Relaxed, no more 996. From the perspective of urban white-collar workers, these devices are more like saving time on housework and increasing work time.

The same is true for AI tools. In the past, when many Wen Sheng Wen, Wen Sheng Tu and other AI tools were promoted, they would say, "The work of the previous 8 hours can now be completed in 1 hour, and you can have fun fishing in the remaining 7 hours."

It’s hard not to think this copy wasn’t dreamed up when advertisers were exhausted working late into the night.

Anyone with no workplace experience will also know that if an AI tool can allow an employee to complete the original 8 hours of work in 1 hour, then the boss’s first thought after learning about it will definitely be to fire 7 people and then replace the 8 people. Leave the job to the remaining person.

So the liberal philosopher and political economist John Stuart Mill said:

All the labor-saving machines hitherto invented have not lessened a man's toil.

To most people, AI is of course a similar “machine”.

After laying the groundwork for so long and depicting AI as darkly as possible from an ordinary person’s perspective, it’s natural for the power of light to appear.

Apple's handling of AI at WWDC24 was to try to reduce the cold feeling brought by AI as a sharp technology and put the user at the center. Therefore, interpreting AI as "Apple Intelligence" can be said to be a homophone and a stroke of genius.

After all, Apple’s intelligence still sounds a little livelier and warmer than artificial intelligence.

The other thing is Apple’s privacy policy.

Although privacy and security is an area that users don’t care about, but manufacturers still need to pay attention to it, on this planet, it is true that only Apple has the most serious attitude towards privacy and security.

Many of Apple's technical updates are accompanied by privacy and security updates. For example, when TouchID and FaceID technologies appeared, Apple would repeatedly emphasize that the user's biometric password data is only stored in the "secure compartment" in the device chip. TouchID and FaceID also do not store image information, relying only on mathematical representations. It is impossible for anyone to reverse engineer data from this storage, and fingerprint and facial image information are not accessible to the operating system, on-device applications, or cloud servers.

Another example is that Apple devices collect some user data, such as health information, etc., but Apple also uses "differential privacy" technology to add interference to the data, so that the server obtains the data but does not know who the user corresponding to the data is.

At WWDC24, Apple also spent several minutes explaining Apple Intelligence's privacy policy.

▲ Apple Intelligence architecture

To put it simply, Apple Intelligence is a device-cloud hybrid intelligence. After the latest iPhone, iPad and Mac products upgrade the system, they will have two built-in device-side models with a small number of parameters. When users have AI needs, or Apple Intelligence finds that it needs to work , the local end-side model will be used first to handle tasks, such as simple tasks such as Vincent diagrams.

When users need more powerful intelligence, for example, users want to use Siri, but Siri has been taken over by GPT-4o. GPT-4o is a third-party OpenAI. Apple, which has always emphasized a closed ecosystem, is facing a very open situation. AI, as well as the ongoing pursuit of privacy, have private cloud servers as a transfer station for data desensitization.

The logic of the flow of user data at this time may be like this. The user uses Siri to ask a question, and the iPhone determines that the local end-side model cannot answer it, so it needs GPT-4o to handle it, so the data with the user information first goes to Apple's private cloud server. On the server, this server may also have a large model. At this time, if the large model in the cloud can answer, it will answer. If it cannot answer, it will encrypt the user information or remove the user information and send it to the OpenAI server. OpenAI will use GPT-4o to answer, and Apple's private cloud server will then send it to the OpenAI server. GPT-4o's answer is returned to the user.

Although the meme posted by Musk is very damaging, the general logical expression is still relatively vivid. Of course, regarding what model Apple Intelligence uses in what scenarios, whether the local model is self-developed or open source and modified, and whether the cloud model is one or multiple. We still know very little about whether it is only OpenAI or there are other helpers, as well as the specific workflow of this private cloud server.

After the naming of Apple Intelligence has eliminated the cold feeling of technology, and the privacy protection of Apple's private cloud servers has enhanced the sense of security, at the specific AI application level, what Apple wants to highlight is the "sense of reliability."

Apple gave us this example at WWDC24: Let's say one of my meetings is rescheduled to late afternoon, and I want to know if this will affect my ability to make it to my daughter's show on time.

At this point, Apple Intelligence can determine: who my daughter is, the details of the show she sent a few days ago, the time and location of my meeting, and the estimated travel time between my office and the theater, ultimately answering whether I can make it. On the show.

This is the specific manifestation of Apple Intelligence as a system-level capability: it truly has the ability to understand users, because it has opened up many different application data of users, and knows the logic of combining these data, and can also deduce the user's problems and needs, and finally provided a solution.

This typical ability is not intended to replace the user's ability at work, but to act like a real assistant, helping users organize information, make suggestions and make decisions.

It is this ability to sink into the bottom of the system and open up application data that also requires Apple to repeatedly emphasize its privacy policy. Users’ requirements for AI are very similar to leaders’ requirements for assistants. They must be reliable in terms of work ability and He is loyal and cannot threaten himself.

At this point, Apple has well restrained the sharpness of technology and maintained its sense of power. With system-level AI capabilities, it has formed a current crushing trend against various scattered AI functions in the Android camp. We can even vaguely see that Apple itself is struggling with, for example, how to integrate invisible AI capabilities with tangible graphical interfaces, and how to integrate powerful AI capabilities with the App Store ecosystem created by millions of Apple ecosystem developers. coexist?

From this point of view, WWDC24 without "humanity is completely defeated" and "the world changes overnight" may not be an undercurrent. It is just that Apple's first AI experience needs to be elegant and rigorous, rather than overbearing and rough, but the essence of AI remains unchanged. .

Whether it is an undercurrent or a current, many people advise you to learn how to swim so as not to sink in the current, but there may be another possibility.

In the movie "Interstellar", the protagonist team lands on a planet whose surface is entirely water. Some team members look at what seems to be a towering mountain in the distance, but the giant thing about 10,000 meters high is not a mountain, but a huge wave.

How can we guarantee that when we see something that seems to be as calm as a mountain coming slowly in the distance, it looks like a huge rushing wave when it gets closer?

In the plastic greenhouse of fate, every cabbage that has been sprayed with too many pesticides once had a dream of becoming a pollution-free organic vegetable.

# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo