If “robot girlfriends” become a reality in the next 20 years, will humans be able to reap AI love?

The hot spots on social media seem to have changed a little bit in some unnoticed corner.

The Internet industry is still in a "cold winter", but hot companies in chip manufacturing, artificial intelligence and other fields can always attract people's attention.

In the past, probably no one would have thought that the changes in Open AI's senior staff and the inside story of Microsoft's "Gong Du" could actually be viewed as "The Legend of Zhen Huan" on the Chinese Internet, causing so much speculation and discussion.

This may be because, as AI technology has made breakthroughs and begun to reach the general public on the consumer side, the entire society has gradually realized one thing: with the explosive development of artificial intelligence, general artificial intelligence can match or even surpass human intelligence in all aspects. Intelligence (AGI) is likely to appear in the near future. The Matrix in "The Matrix", MOSS in "The Wandering Earth", the "old thing" in "The Guest"…

How will we deal with these superintelligences when they do arrive?

It is very likely that the entire society needs to establish a new set of behavioral norms, including how to define AI legally, how to treat AI morally, and whether humans and AI are truly equal or have their own priorities…

If so, who is dominant?

Recently, science fiction writer Han Song asked this question: "What impact will artificial intelligence and the metaverse have on gender relations in the future? Will the concept of family disappear?"

Love and family are usually considered to be the core areas of humanistic emotions in traditional society and are the privacy of human beings.

In Hollywood, you see the protagonist of the "Fast and Furious" series bursting with fighting power because of the phrase "family." Although he complains about the screenwriter's nonsense, you do not object to the core of this plot model-emotional bonds are closely related to human potential.

The subtext of Han Song's question is: new intelligent life may enter the core area of ​​human consciousness.

If the information provided by the technical experts is correct, then we need to think seriously about some things. For example, will we form a family with AI?

This is not only a legal issue, but also a matter of the survival of "anthropocentrism." Social ethics in the AI ​​era are so complex that most people are not ready for it.

1. Love of the Creator: Human beings are always in love with themselves

Musk announced earlier this year that he would launch a Catwoman robot to become the "perfect companion" for male consumers. This robot has the sexy appearance of Catwoman and is said to be able to take care of housework, take care of its owner's life, and even give birth to children.

But no one will ask "Can I have a family with the Catwoman robot?" because it is still a kind of AI device, a commodity without personality, rather than an intelligent life equivalent to humans.

▲ Musk dancing with the robot "female companion" output by AIGC, source: Internet

When we discuss love and family, the AI ​​involved must be a "new species" that breaks through the Turing limit, generates self-awareness, and even far surpasses humans in intelligence.

So far, most of our literary and artistic works have not actually set up this kind of AI, only some specious images of "robot girlfriends".

There is a pattern in Japanese comics and Japanese light novels, similar to the early "Humanoid Computer Angel Heart". The story structure is "otaku + robot girlfriend in the form of a soft girl". The robot gradually learns and understands human emotions while getting along with humans. , and finally completed the evolution of "becoming a human" through the bond with the male protagonist.

This is actually a typical anthropocentric narrative. Apart from being silicon-based life forms, those humanoid computers have no essential difference from the "alien creatures" in traditional fantasy:

They have supernatural abilities that far exceed the physical strength of humans, but they do not necessarily have intelligence beyond humans. Especially when it comes to core issues such as emotions, they are more naive than humans, and they always obey and yearn for human emotional patterns. .

Will super AI be like this too? This is the “alignment problem” that leading research institutions such as OpenAI have been trying to overcome.

People standardize the learning and creation of traditional AI by using reinforcement learning with human feedback (RLHF), in which humans evaluate and supervise the results of AI with the assistance of various tools.

In the face of super AI that is "much smarter" than humans, under the premise that the limit of human intelligence remains unchanged (or it is difficult to improve quickly), when the AI's ability exceeds a certain critical point, RLHF will fail, that is, humans cannot Evaluate artificial intelligence projects more effectively.

Jan Leike, currently the head of the Open AI "Super Alignment" team, proposed investing 20% ​​of Open AI's computing power to create a "Super Alignment" system, allowing AI to evaluate AI projects and establish a scalable supervision system.

Jan LeiKe's idea is to introduce a randomized controlled experiment (RCT with tampered answers) by tampering with answers, artificially creating answers with flaws to measure the effect of scalable supervision – in theory, the solution of the masters Of course they are all beautiful.

However, the idea of ​​adding a "curse" to AI is not reassuring. As early as the 1940s, Asimov's science fiction novel "I, Robot" proposed the famous "Three Principles of Robotics". A method similar to "initial implantation" ensures that robots and humans can coexist peacefully.
Interestingly, almost all literary and artistic works for decades have imagined the same thing: how robots can break through this safety fence and harm humans.

Until recently, the domestic TV series "The Bionic World" still did not break away from this stereotype…

Human beings are so interesting. As imperfect "creators", we are full of wariness and fear of our creations, but we still have no hesitation in creating a "machine" that may surpass ourselves.

The spirit of Homo sapiens to constantly "try to die" on the road of evolution has long been engraved in the genes of this species. It is also because of this continuous progress that humans have created a brilliant civilization.

So… can the Creator fall in love with His creation? It is obviously possible. AI is the "superego" of humans, and for humans, falling in love with the "superego" is a common thing.

There are similar stories in the myths of various ethnic groups. Fairies, elves, and goddesses are bathing in the world, and a young man steals her clothes, and the two fall in love. The Cowherd in the East was so successful. In Celtic mythology in the West, there are also stories of fishermen stealing the seal skins of fairies and letting them stay on the island…

In these stories, the logic of mortals falling in love with goddesses is very logical, because these gods are perfect human beings, with supernatural powers and amazing beauty – everything is made by people according to their favorite appearance.

The so-called homosexuality between gods and humans should be put another way. Gods were originally created by human beings based on their "superego imagination". There is naturally no emotional barrier between gods and humans. Zeus would also admire the beauty of a mortal woman and transform into a swan to possess Leda. Aphrodite will also fall in love with the beautiful boy, and the Weaver Girl will think it is good to be with the Cowherd…

Regarding the issue of AI, the identity of human beings seems to have changed, but AI and gods share the same attributes, and they are both products of human beings' "superego imagination".

In "The Three-Body Problem", Luo Ji drew the Zhuang Yan of his perfect partner, which was the look he most admired. The goddess and the AI ​​were essentially the "Zhuang Yan" drawn by humans for themselves.

2. Can goddesses also fall in love with mortals?

Then, we need to face another question, will super AI fall in love with humans?

There are many reasons for humans to fall in love with AI. For example, they can transform into beautiful appearances and possess extraordinary abilities because they satisfy humans' "superego fantasies." But what are the reasons for AI to fall in love with humans?

As Huang Tiejun, dean of the Peking University Intelligent Research Institute, said in response to Tencent News's "20 People and 20 Questions in 20 Years" question, after the emergence of super intelligence, the problems are very complicated. In addition to "can we trust AI?", there is also " How can AI trust humans?"

Dean Huang said that just like the process of establishing mutual trust between people, people and AI need to go through a running-in period and need to "listen to its words" and "watch its actions", that is, through sharing the knowledge constructed by humans using language. system, and in practice test whether what AI says is consistent with reality to confirm whether mutual trust can be established with AI – in fact, the trust system of human society is gradually built in this way.

Regarding the question of people falling in love with AI, the answer is similar. Whether AI will fall in love with people depends not only on the operation and evolution of AI, but also on humans themselves.

Professor Ma Zhaoyuan of Southern University of Science and Technology also gave a perceptual answer in "20 People, 20 Questions": We know more and more clearly that the reason why humans navigate on such a small planet in the vast universe is because The greatness of human wisdom.

Such wisdom includes reason but is not bound by reason. Because we know the limitations of rationality, this limitation needs human sensibility to make up for it and need human sensibility to guide us. This may be the norm in our future symbiosis with machines.

We guide them and cooperate with them, and they can do more and more things, becoming more and more reliable (trustworthy), and as a result, people's sensibilities gain more and more freedom.

We may guess based on this that it is possible for super AI to fall in love with humans.

Just as the way AI wisdom is generated is still a "black box" that is unknowable to us, humans will also have a "black box" part of AI that cannot be understood by pure rationality.

We are eagerly looking forward to cracking the "black box" of AI, and AI will also try to crack the "black box" of humans. In the process of "cracking" each other, they not only achieve "alignment with humans" in terms of values ​​and wisdom, but also achieve Emotional modules similar to those of humans will be generated, allowing them to transcend the "species" barrier to appreciate and admire intelligent life that is consistent with their own spirit.

Many things are still unknown, but it is precisely the "X" that represents the unknown that contains the elements that make the world exciting. Goddess can fall in love with people, why can't AI?

3. "Gender" seems to be an outdated concept, but the essence of family is property relationship

Let’s go back to the question raised by science fiction writer Han Song in the “20 Questions for 20 People in 20 Years” event specially planned for the 20th anniversary of Tencent News: “What impact will artificial intelligence and the metaverse have on gender relations in the future? Will the concept of family disappear? ?」

In his answer, his friend Chen Qiufan believed that the metaverse inherently contains assumptions about future gender and family relationships, and that gender and family relationships in the future will become more and more diversified.

Relationships between the sexes will no longer be limited to physical existence. Emotional connection and spiritual resonance will become dominant. At the same time, there will be more "cross-border" phenomena in the relationship between the sexes. For example, the boundaries between real and virtual in "The Matrix" will make it possible for future couples to Will cross it.

In fact, "love" itself is a difficult concept to discuss. When we discuss "possibility", love is always possible.

It can exist between feuding families, between celebrities and aristocrats and the destitute, between various genders, and can even transcend racial barriers and exist between earthlings and alien life or even non-life… Love is super utilitarian. , super-hierarchical, super-rational, in a sense, it also means that it can jump out of the logical framework at any time.

If AI really evolves into super intelligence, will they generate the unstable factor "love", or will they exclude it as invalid data? Looking in the direction of "aligning with humans", it seems more likely to be the former.

Relatively speaking, "family" is a more rational topic.

Engels once concluded in "The Origin of the Family, Private Property and the State" that the essence of the family is a property relationship. After productivity progressed enough to retain surplus products, public ownership gradually gave way to private ownership. In a sense, concepts such as couple marriage, clan, family… are all by-products of private ownership.

Family is a changing concept. There are great differences between feudal families and modern families, and there are also great differences between primitive clans and feudal families. This changing concept is essentially adapting to changes in social forms.

And ethics is something that serves the social form. An important part of Max Weber's "Protestant Ethics and the Spirit of Capitalism" is the combination of the profit-making spirit and ultimate legitimacy of capitalism with religious forms, and believes that commercial gains Profit is in line with Protestant ethics and can be regarded as the ultimate goal of life. Ancient religious ethics must serve the new capitalism, and all history is contemporary history.

So when AI intervenes in human emotions, will the extensive combination of humans and AI change the basic shape of the family, or even make the family disappear?

If you think about this question in essence, it actually asks: Will AI change private ownership? Will the "spirit of capitalism" end?

The answer is easy to give. The emergence of super AI is within our sight and will not shake private ownership. The stability of family ethics essentially comes from the property system rather than love or anything else.

For example, when the father is no longer the absolute core of the family's income source, the feudal patriarchal family will gradually disintegrate. However, modern society still retains the family, which means that modern people still need this way of organizing and distributing property.

Whether AI can subvert the existing family structure depends on the extent to which it can participate in social wealth distribution. When AI itself was still a kind of "property", diversified families were out of the question.

Teacher Chen Qiufan is talking about a distant future. Only when AI breaks through physical limitations and masters the distribution of wealth, will we likely see more community-based families based on common interests, values ​​or lifestyles rather than blood ties. Appear.

At that time, will society still maintain fundamental private ownership? If the foundation of private ownership is shaken, it is entirely conceivable that the concept of family will disappear into history.

"20 People, 20 Questions" is just the beginning. Humanity will discuss AI with high frequency and intensity in the foreseeable future. Historically, changes in social form driven by productivity improvements have always been the most important issue, bar none.

Love and family are only a small part of this big issue. It is related to how we understand society and ourselves. Only after a long period of thinking can we come to a staged conclusion.

And one sure thing is, we can't stop asking questions.

Editor's note: This article is from "Deep Eyes Business 2022", written by Hema Jun & Chu Qingzhou, and published by Aifan'er with authorization

Love style makes the future within reach.

# 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