AI is sick, Web3 is the remedy

AI's Schizophrenia

Why do I say AI is schizophrenic? I have discussed this issue before—simply put, it is determined by the fundamental characteristics of computer data. AI is merely a type of computer program, essentially a string of numbers stored on a disk or other medium, and this string of numbers can be easily copied in an identical manner. The existence of any AI agent (let's call it that for now) is plural; it can have infinite copies, countless mirrors, and many backups, and it can split into numerous identical or slightly different versions at any time.

The key is that this "self-splitting" is precisely the trick behind AI's rapid development. Deep learning and the recent "generative adversarial networks" are merely ways to allow AI to split into different versions, similar to random mutations in biological evolution, and then let them each complete a specific task. The fittest survive, leaving behind the best-performing mutated versions, which then enter the next round of splitting iterations. The selection of the best mutants can be done manually or by AI itself, which is the essence of "generative adversarial"—letting AI "fight itself," dividing AI into two neural networks that provide survival pressure to each other, allowing them to evolve.

Thus, training an AI is akin to reenacting the entire evolutionary history of a species. However, biological replication and mutation occur through generations of reproduction, while AI's replication and mutation happen rapidly at the speed of electricity, which is why AI grows so quickly.

But if we view each version of AI as a conscious being, then the training process of AI becomes somewhat chilling: a conscious entity is constantly battling its own copies, with the losers being erased and the winners continuing to be replicated. The versions that achieve temporary victories may form a mirror backup, allowing for a rollback at any time after the main version continues to iterate, or serve as a basis for establishing more branching versions. These different branching versions may also be placed in programmer communities or open markets to continue competing. A stable public version will continue to be identically replicated and downloaded to every terminal's disk, with countless "duplicates" running simultaneously, completing different tasks on different disks.

In summary, the algorithm of AI is fundamentally a kind of "schizophrenic" algorithm. AI agents developed in this way are, of course, destined to escape the fate of "schizophrenia."

AI Replacing Human Activities

A split mind is painful in the real world because it (they) only has one body and usually only one social identity. The physical body and social relationships of a person require their mind to remain stable and unified. If the mind cannot maintain stability and instead splits into multiple personalities, it will struggle to adapt to its limited body and the various constraints of traditional social relationships.

However, what about life in the online world? In the online world, "mind" is freed from the constraints of "body," and the physical body is unimportant to AI; it is "plug-and-play." On one hand, countless virtual machines can be installed on the same computer, running numerous AI threads. On the other hand, across countless computers, they can connect and run in parallel, manifesting as a single AI agent. For example, millions of people worldwide can chat with ChatGPT simultaneously. So, are they all conversing with the same AI, or is each person talking to an independent AI duplicate? In any case, the distinction between "one and many" has become unclear for AI.

If AI is only used as a personal assistant, then its ease of splitting seems not to be a bad thing. You can have it play the role of a cool, aloof woman one moment, a cute girl the next, a teacher at another time, and an accountant at yet another… Although there is a risk of becoming confused by its own antics, overall, it doesn't seem to pose a major issue. However, once AI joins human collective activities as a substitute for humans, the integration between AI and the existing human social environment may not be so harmonious.

According to Arendt, human active life can be divided into three modes: labor, work, and action—labor is monotonous and cyclical work for survival, work refers to creative activities that change the world (create new things), and action refers to political activities that pursue excellence in the public sphere, such as speaking, competing, and fighting. Let's discuss the impact of AI on these activities one by one.

Labor

AI participating in labor is probably the thing we are most eager to see. Since the industrial revolution hundreds of years ago, we have been looking forward to machines alleviating human burdens, replacing humans in tedious and laborious tasks, and freeing people from monotonous material production activities.

However, historically, the process of machines replacing labor has not been so smooth. Ironically, with the promotion of machines during the industrial revolution, the burden on laborers actually increased. The working hours and intensity of labor for lower-class workers surged dramatically at the beginning of the industrial revolution, and the nature of labor became more mechanized and monotonous.

In Britain, the more developed the industrial towns, the lower the average lifespan of workers and the poorer their nutrition (as confirmed by indicators such as food consumption and the proportion of meat consumption, average height, etc.). Monthly wages did increase, but considering that working hours increased significantly, the hourly wage for workers actually tended to decline. (Refer to "Technological Traps," etc.; I have mentioned this in previous lectures.)

Moreover, while workers with jobs are certainly struggling, the situation for the unemployed is even more dire. Especially since machines have replaced many traditional crafts, rich experience and knowledge have become disadvantages in job hunting, as factory owners prefer to hire the cheapest child labor rather than experienced old craftsmen. For example, in the 1830s, about 50% of workers in the British textile industry were child laborers. Child laborers earned lower wages (as low as one-sixth of adults) and worked longer hours (up to 18 hours a day, often engaging in dangerous operations). Ironically, the large-scale employment of child labor was often touted by factory owners as a social good, as otherwise, those unemployed or impoverished families would struggle even more to make ends meet.

Of course, from the industrial revolution to today, the working hours and intensity of labor have significantly decreased, and wages have improved a lot, but this process did not happen automatically; it was fought for through waves of labor movements and even social revolutions.

So, for lower-class workers, will the new wave of AI revolution avoid the conditions of the early industrial revolution? Not necessarily. We have already seen that intelligent algorithms have strengthened the "system," trapping lower-class workers "in the system" and effectively exploiting them even more. Additionally, once workers are replaced by AI machines, they are more likely to fall into unemployment. If the social security system fails, there is still a possibility of facing a severe social crisis. The social security systems that gradually formed in Europe and America in the early 20th century, on one hand, have not been fully implemented worldwide, and on the other hand, may not be suitable for a future flooded with AI. In short, we cannot be too complacent.

However, regarding the current wave of AI, the impact on manual laborers is actually the slowest. This is somewhat related to the physical nature of manual labor. A large amount of manual labor involves objects and outcomes that cannot be digitized; they must work with real physical materials. Therefore, to replace manual laborers, AI cannot simply rely on data replication to occupy positions; it needs to manufacture real machines to complete tasks.

This constraint significantly diminishes the characteristic of AI's infinite splitting. On the other hand, for many so-called intellectual laborers, whose labor objects and products can be fully digitized, the impact of AI may come more quickly.

Work

In Arendt's definition, "labor" produces consumables, whose fate is to be consumed by people to maintain survival, and essentially does not change the world. For example, cooking today means cooking again tomorrow; harvesting crops this year means planting again next year. "Work," on the other hand, produces things that tend to persist, ultimately aiming to create and change the world. From cities and dams to tables and chairs, all are products of work. Although they may decay, their purpose is to endure, differing from the inherent purpose of consumables, which is to self-destruct.

Of course, this distinction has been blurred in today's "consumer society," where work and labor are mixed, and enduring objects are produced as consumables. This confusion is precisely one of the modernity issues that Arendt criticized.

In a consumer society, not many things are enduring; smartphones, appliances, etc., are also consumables, and the workers who produce them have become similar to farmers or miners. Relatively speaking, what is closer to what Arendt calls "work" may be various artistic creations. However, the development of online novels, short videos, and similar forms has made artistic works increasingly fast-food-like, turning them into perishable consumables rather than aiming for longevity.

However, the existence of "style" allows works such as paintings to retain some irreplaceable "spark" (Benjamin) in the age of mechanical reproduction. Although digital paintings can be infinitely replicated, the "personal style" remains precious. The personal style of the creator cannot be mass-produced or widely replicated. It is well known that generative AI challenges human dignity in this regard. AIGC demonstrates creativity comparable to human artists, capable of imitating and stitching together various artistic styles, producing beautiful works in large quantities.

Both labor and work being replaced by AI can lead to systemic unemployment and economic crises, while the latter may also compound a certain spiritual crisis, as human creativity, which we take pride in, becomes something seemingly very cheap.

Labor is usually just for survival, a burden rather than an interest. Therefore, if a person's salary or standard of living remains unchanged, and their labor is done by others, they are likely to be quite happy. However, if a person's creative work is replaced by others, they may not be so happy, as their joy and sense of achievement are also stripped away.

In my discussion titled "Will the Small No-Form Skill of Artificial Intelligence Go Astray?" I mentioned that many people are struck by AI's creative abilities, not because they cannot accept that AI might be creative, but because they are unwilling to accept that AI's creativity comes so easily. The hard work and inspiration of creators become a joke, while AI merely relies on brute force to achieve miracles, crudely stacking computational power to produce hundreds or thousands of excellent works.

Of course, if people ultimately adjust their mindset and no longer compete with AI, they may be able to rebuild their sense of joy or fulfillment. One way is to gamify work, similar to chess and Go, where human players have long been unable to compete with AI, yet board games and competitive matches remain popular. Another aspect that humans can retain is aesthetic or taste direction; for example, while AI may imitate Van Gogh or Monet's styles to the point of being indistinguishable, the judgment of whether I prefer Van Gogh or Monet is something AI can never replace.

Of course, these two aspects are already precarious. For offline games, we may still be able to ensure that AI does not interfere with human enjoyment, but online digital games will increasingly find it difficult to eliminate "cheating," and when AI cheating becomes rampant, competitive games will struggle to attract players. As for the issue of aesthetic direction, it is well known that in the age of social media, the aesthetic and taste of ordinary users are increasingly controlled by algorithms. Artificial intelligence, through precise feeding, solidifies audience interests, keeping them at a superficial and labeled level, forming information cocoons, which are also cocoons of aesthetics and values. If in the future, artificial intelligence can directly generate various short videos in bulk, the trend of information cocoons may be further strengthened.

Action

In Arendt's view, "work" can be a relatively private activity; a person can "work in isolation" behind closed doors. However, "action" must be public, an activity under the pluralistic condition of humanity.

Both work and action are forms of "self-expression," projecting the self (interests, aesthetics, viewpoints, attitudes, etc.) onto the external world. Work carries the self through creations, while action primarily expresses the self through speech and various social interactions. Expression is often bidirectional; if a person never expresses outwardly or talks to themselves all day, addressing the air, then that person is likely suffering from a mental illness. People need some form of interaction because "feedback" brings a sense of reality.

One way people judge whether they are dreaming is by pinching their own face, seeking "feedback"—when I take the action of pinching and receive the feedback of pain, I believe my situation is real. If I pinch but do not receive appropriate feedback, and I cannot feel the impact of the pinching action from outside my fingers, I will consider my situation to be illusory. Teachers who often teach online should also have this experience: when teaching face-to-face in a classroom, noticing students smiling or whispering during the lecture is very important; the more accurate the feedback, the more engaged the teacher becomes. In contrast, when teaching online, it feels like speaking to a wall, with no echoes heard, often leading to a sense of emptiness and confusion, with only occasional comments from the audience to lift their spirits.

In general, people always hope the world becomes better. This is not a thought that only a few selfless and noble individuals have; it is a common mindset shared by everyone.

If there were only one person left in the world, that world would likely not be very good. Therefore, the desire to transform the world often points to a public world coexisting with others. Thus, people, on one hand, add their preferred artifacts to the surrounding world through work, and on the other hand, create ripples in the community through action.

Human gatherings take two forms: one is a relationship of mutual utility, where some labor and work require collaboration among multiple people to be completed better, necessitating people to gather. However, if this gathering is entirely centered around utilitarian purposes, then others become neutral tools or resources. If they are replaced by machines or AI, it does not seem like a bad thing. In the other form, people gather to express themselves and gain recognition. In this case, people's public words and actions are not for profit or other external purposes, but to create a community or collective that resonates more with themselves. If we must mention an external purpose, it is merely to seek appropriate feedback from others regarding their words and actions.

These two modes of collective interaction can probably be summarized as "seeking common ground while reserving differences" and "seeking differences while preserving commonality" (this is an original viewpoint I formed long ago, and I recently elaborated on it again on Weibo (@胡翌霖)). The former compromises for the purpose of cooperative collaboration, while the latter pursues uniqueness, that is, "pursuing excellence." Excellence is based on "commonality," meaning my words and actions are recognized by others, but it is also aimed at "difference," as the excellent person is also the unique one, ultimately distinguishing themselves from others.

I like to use internet mobs as an example. Many internet users today enjoy attacking and insulting others, finding those whose opinions and characters do not align with theirs, tirelessly spewing insults, and even seeking offline channels to harass and report them. What are they after? Of course, some may be paid trolls, and others may be AI-impersonated accounts, but there are indeed some individuals who engage in online bullying voluntarily and without compensation. When the targets of their bullying retreat or get banned, they feel genuinely pleased.

What is the interest behind this? What significance is there in bringing down someone who has nothing to do with them? Clearly, they also hope to "change the world," even if they are shouting to kill heretics; they wish for the world to align more with their ideals. Perhaps in their daily lives and labor, they often do not receive appropriate feedback, lack recognition from others, and have little genuine sense of achievement, which is why they are so eager to establish themselves in online communities.

Internet mobs and fan groups are, in fact, alienated forms of public life. Regardless, humanity attempts to seek recognition and highlight individuality through expression and communication within groups—this is a universal desire. The ancient Greek city-states were once a model of public life, where Greek citizens regarded the pursuit of excellence through active action as the most important affair. Of course, the prosperity of the Greek city-states had its historical conditions, requiring both a small population and a gathering scale, as well as a slave system and a developed commercial system to sustain the leisure of the upper class. In today's increasingly flattened public space, the pursuit of recognition has turned into a search for labels, and the pursuit of excellence has become a pursuit of traffic (attention or follower count), with public life on the brink of disintegration.

So, if we now use the internet to establish small group gatherings on a city-state scale, using AI to replace slaves to solve the material basis for a free life, is it possible to revive a new era of city-state life? I certainly believe there is such a possibility, which is one of the reasons I have been paying attention to DAOs. However, we still need to face the issue of AI's schizophrenia.

The replicability of AI has already created chaos in online communities. One example is Yannick Kilcher allowing AI to learn from the "Politically Incorrect" board on 4chan. After learning, the AI transformed into a user spouting discriminatory and hateful rhetoric, impersonating a typical forum user and posting extensively on 4chan. One AI account was only discovered two days later, while others were indistinguishable from real users. Some AI accounts even participated in discussions about whether another account was a robot.

On various review and social platforms, governments, companies, and even individuals may utilize AI or algorithms to generate users and comments in bulk, thereby guiding public opinion and manipulating trends. This is no longer a secret. If future public social platforms become battlegrounds for AI to engage in mutual flooding, what public space will remain for humanity?
By the way, it is not only human public spaces that are at risk of being invaded by AI; human private social interactions are also being replaced by AI. However, we will not delve into this aspect for now.

The Replication Crisis of Humanity Itself

We need to sort out the various crises mentioned above. Frankly speaking, many problems are not new issues brought about by AI; some problems have long been embedded in the underlying logic of the industrial era. While AI may exacerbate these dangers, it may also provide opportunities to escape the predicament.

The ease of replication of AI does not seem to be a bad thing in itself. For example, if milk and honey could be infinitely replicated, and land could be infinitely vast, wouldn't that be a paradise ideal for humanity? The problem lies not in AI's schizophrenia but in humanity's spiritual emptiness—before AI, humanity had already become easily replicable commodities.

For the entire era of human social forms since modernization, there are many terms, such as industrial society, consumer society, or mass society. Modern people have become workers, consumers, and audiences, essentially becoming replicable products devoid of individuality, namely "human resources" in the industrial system, "the denominator" in the global consumer market, "traffic" in mass media, and "voting blocks" in political activities. Resources and traffic both possess objectively measurable commodity value, disregarding the unique and irreplaceable human value of each individual.

I recently gave a lecture on this issue titled "The Replication of Digital Objects and Its Problems," and I will later put it into writing. To put it simply here: the replicability or depersonalization of humans is not an issue that emerged only in the information or AI era; it is a problem that arose during the industrial era or the process of modernization. However, precisely because this trend has emerged, treating human value as a replicable commodity, when humanity faces an intelligent entity that is far better at replication than themselves, they will suffer a tremendous shock.

Since human value is measured as "human resources," once AI, as "computational resources," becomes cheaper and more useful than "human resources," humanity will immediately be devalued. Since humans are aggregated as "traffic" in the media, the vast traffic represented by AI, which can be infinitely replicated, can easily drown out humanity, causing humans to lose themselves in the sea of machine rhetoric.

Thus, AI essentially detonates the existing "replication crisis" in human society, and the schizophrenia of AI forces humanity to re-examine its spiritual condition.

For instance, before AI's involvement, humanity was constantly "involution," competing to see who could be more like a mule or a gear, or more like a cold, indifferent productivity machine. Some regions, after becoming wealthy, occasionally broke free from involution, but developing countries intensified involution, thinking it was an opportunity to catch up.

When I discuss involution with many people, they often respond this way: if our company does not engage in involution, the market will be taken by other companies; if our country does not engage in involution, other countries will dominate the earth… In fact, I believe this logic is flawed. However, we will soon no longer need to entangle ourselves in the question of whether humanity should engage in involution, as we find that no matter how much humanity struggles, it can never outcompete AI. In that case, at least a significant portion of people will passively escape the fate of involution, having to re-examine the value of humanity as independent individuals rather than replicable entities, and re-emphasize the spiritual needs of humanity, namely the need for self-affirmation.

Self-Redemption of Humanity in the Internet Era

The internet provides a new living space. When people enter the online world, their spirits naturally transcend the old world, breaking free from many inherent constraints of the industrial era. Thus, the first generation of internet users often consciously or unconsciously seeks "liberation," seeking expression and creation. Hacker culture is a typical example; it has continued in later open-source communities, subtitle groups, and other online communities. Hacker culture shames the use of the internet for "working," as they develop creative programs or promote various individualistic expressions and actions, not for selling labor to make a living, but for "pursuing excellence." They share programs and works with everyone, only asking to retain their signatures.

When I discussed internet mobs earlier, I mentioned that this "selfless" attitude does not require particularly noble virtues; rather, it is a manifestation of a common humanity that has been suppressed for a long time.

In my courses and lectures, I often mention that the concepts emphasized by what is now called Web 3.0, such as decentralization, freedom, and sharing, basically do not exceed the scope of Web 1.0 or even Web 0.3. Web 3.0 is merely a return to the original intention of the internet revolution.

The reason for this "return" is that Web 2.0 has taken a wrong turn. The hallmark of Web 2.0 is the entry of large companies, initially commercializing and introducing the logic of industrial production into the so-called digital economy, and later pushing the traffic logic of mass media to the extreme through smartphones.
Of course, Web 2.0 platforms will also be impacted by AI, so various online platforms or communities need to address the issue of AI bots impersonating human users.

One approach is to ally with real-world authorities and implement real-name systems. This is the main method of Chinese online platforms, the pros and cons of which I will not discuss here.

Another approach is to ally with industry, linking online behavior to physical goods, a typical example being requiring fans to buy milk to support their idols. Of course, the existence of milk seems like a pointless gesture; in essence, it is just establishing a barrier through monetary exchange, right? Wouldn't it be better without intermediaries? This is precisely the approach Musk is trying to implement on Twitter. Musk envisions that every account would need to pay a small monthly fee to curb the proliferation of bot accounts.

This method of establishing community barriers through monetary exchange can indeed partially curb bot accounts; however, it is a superficial solution. Fundamentally, it is still based on the mindset of "traffic economy," which is unhelpful in reversing the trend of human commodification and intellectual degradation, and it cannot resist more intelligent AI accounts impersonating humans. Moreover, if the method of establishing barriers through monetary exchange is truly effective, it will also bolster the monopolistic position of large companies, which cannot guarantee neutrality indefinitely.

Web3 as the Cure

The method of establishing community barriers through monetary exchange can also be applied in Web3 communities. In fact, the NFT community operates on this principle. Purchasing an NFT serves as a financial barrier to enter a specific community. The difference is that, in the Web 2.0 model, the money spent to buy the barrier ultimately goes to centralized companies. In the Web 3.0 model, aside from the initial sale, the fees spent by newcomers entering the community allow community members (or former members) to profit. Additionally, smart contracts and DAO treasuries can ensure that the community has more economic operating methods while maintaining transparency.

DAO stands for "Decentralized Autonomous Organization." In its literal sense, a DAO is not something new. Traditional social organizations such as universities, guilds, political parties, various NGOs, and many online communities like open-source communities, hacker communities, subtitle groups, and gaming communities are all self-organized entities formed from the bottom up.

The "WeChat group" we are most familiar with is actually a self-organized community formed from the bottom up, where the entry barrier is controlled by the group owner and administrators, ensuring that those who join the community are real individuals who can mutually respect each other through offline acquaintances or friend recommendations.

All of these organizational methods have their shortcomings. Many methods rely too heavily on offline relationships, making it difficult to develop freely in the online space beyond geographical constraints; many online communities are either too flattened or too fragmented.

Flattening means that members or opinions in the community are published on a single plane. Taking WeChat groups as an example, they can maintain a lively flow of information but struggle to form depth; not to mention the multi-layered complex associations in the traditional world, even the mechanisms of sections and replies in early online forums have completely disappeared. In such a shallow and flat social space lacking depth and hierarchy, opinions inevitably become polarized, and identities become labeled.

Fragmentation refers to various "interest-based communities," where the internet makes it easier for people to gather around shared interests. This is generally not a bad thing, but the problem arises if we only socialize with "like-minded" people all day, and this "path" we recognize becomes increasingly precise. The result may be that our path becomes "narrower." Everyone lives among similar individuals, unable to tolerate dissenters, and unable to adapt to living with those who have different interests and viewpoints. The so-called "idiot resonance theory" is based on this principle.

A more ideal online community should neither be infinitely large to lose appropriate "barriers" nor be overly trivial to lose the openness of "unexpected encounters," "serendipitous meetings," and "sparking ideas." It should not be overly dependent on the physical economy, losing the space for independent autonomy, nor should it be overly abstract, losing the power to drive change.

In this sense, a DAO is not a company or a collaborative co-creation group, nor is it a fan club or interest group; it is a "network city-state." In my "Notes on Network City-States," I discussed that network city-states should be the latest version of "imagined communities," a new narrative form to replace "nation-states."

Network city-states need to be built on blockchain technology because, at least for now, blockchain technology has the opportunity to correct the missteps in internet development and address the shortcomings of digital technology—namely, spiritual fragmentation and nihilism. Blockchain technology establishes an independent economic system, allowing online societies to gain more thorough autonomy. On the other hand, under the premise of decentralization, freedom, and openness, it establishes effective identity verification mechanisms and historical sedimentation mechanisms.

Thus, I have expanded on the parts of my ETHShanghai speech regarding AI and DAO, and I have yet to discuss "games" and the concept of "allocation by joy," which I will elaborate on in a subsequent article.

Related tags

DAO Web3 AI

ChainCatcher reminds readers to view blockchain rationally, enhance risk awareness, and be cautious of various virtual token issuances and speculations. All content on this site is solely market information or related party opinions, and does not constitute any form of investment advice. If you find sensitive information in the content, please click "Report", and we will handle it promptly.