Three years in the past, ChatGPT was launched. Today, AI is not a brand new time period solely mentioned by younger folks in huge cities like Beijing, Shanghai, and Guangzhou. Even grandmas in rural areas could find out about DeepSeek and chat with Doubao. AI is changing into an infrastructure in our lives, virtually as important as water, electrical energy, and the Internet. What makes it particular is that AI is a expertise that straight produces intelligence, and intelligence has all the time been what people are most happy with. What will occur to human society if intelligence might be mass – produced at extraordinarily low price and excessive effectivity?
Since the discharge of ChatGPT, discussions in regards to the extent to which AI will substitute human jobs have by no means stopped. In truth, AI is reshaping human life and society in all features. To higher perceive AI and our future, we interviewed Zhang Xiaoyu, a science – historical past author and impartial scholar who has lengthy been involved about how expertise impacts human society. He lately printed a brand new e book known as The Pre – historical past of AI Civilization, on the duvet of which it’s written, “Compared with AI, we are pre – historic animals.”
Zhang Xiaoyu studied the historical past of political thought in school. Later, he went to Germany to pursue a doctorate. At a studying membership, science and engineering college students usually requested philosophy college students how they considered new applied sciences. Zhang Xiaoyu truthfully replied, “We can’t. The books we read are all from 2000 years ago, and we haven’t considered how to answer these questions under modern technological conditions.” This expertise led him to examine expertise, “because I really think technology is important.”
We talked about the 4 core ideas in his e book and plenty of of his views that subvert anthropocentrism. For instance, AI will not be a software of people however is probably going to be a brand new civilization; AI will substitute 99% of human jobs and utterly subvert the present social construction of people; how the 1% deal with the 99% is how AI will deal with us sooner or later… We additionally mentioned how, from the brand new perspective introduced by AI, we are able to re – perceive ourselves, human nature, and human values.
In addition to these summary ideas and views, we additionally talked about some helpful and provoking makes an attempt within the subject of AI that he has noticed. Zhang Xiaoyu usually communicates with AI entrepreneurs. “As we all know, these people are post – 2000s.” As a put up – 2000s myself, I additionally talked with him in regards to the decisions of this technology of younger folks, corresponding to falling in love with AI and the reflection on meritocracy caused by the event of AI.
Almost concurrently our interview, Elon Musk “blurted out” in an interview on the Tesla Gigafactory that humanity has entered the technological singularity, and 2026 would be the 12 months when AGI is realized. We cannot decide whether or not the longer term has “arrived,” however as Zhang Xiaoyu wrote in his e book, everybody, not simply AI specialists, in each self-discipline and subject, and even strange folks, ought to take into consideration a sequence of questions relating to AI and people as a lot as they’ll, as a result of this issues our future, and “if we do nothing, history will surely develop in the worst – case scenario.”
Below is the dialogue between Jingxiang Studio and Zhang Xiaoyu:
Sometimes, the working guidelines of society are decided by the only mathematical legal guidelines
Jingxiang Studio: This e book primarily focuses on 4 ideas: emergence, human equivalence, algorithmic judgment, and the civilization contract. Let’s begin with emergence. It’s an idea that helps us perceive how we are able to “create intelligence.” Why did you select such a technological perspective, and the way do you perceive emergence?
Zhang Xiaoyu: In easy phrases, emergence means “simple rules + large scale = system upgrade.” For a system, when its scale is massive sufficient, its guidelines are easy sufficient, and its variety reaches a sure threshold, given sufficient time, it would expertise a system leap from a easy system to a posh one.
Emergence is first a organic idea. For instance, the evolution from single – celled organisms to multi – celled organisms. From the attitude of a single cell, its finest technique might sound to be to grow to be a most cancers cell as a result of most cancers cells are immortal. However, within the historical past of organic evolution, single cells bear useful differentiation and grow to be particular cells, corresponding to neurons and platelets, and at last kind a multi – celled organism.
Another instance is ants. The technique of a single ant could be very easy: random strolling, making an attempt to cowl each path. But for the ant colony, that is the only technique to perceive the encompassing atmosphere. The intelligence degree of a single ant could be very low, unable to acknowledge people, and its cognitive vary is proscribed to a small space round it. But once they kind a colony, they appear to have the knowledge of division of labor and cooperation.
Take the evolution of the nervous system for instance. The perform of a single neuron cell could be very easy: to transmit info. But when there are sufficient of them and the construction is advanced sufficient, intelligence emerges. It’s very probably that synthetic intelligence, human intelligence, and the intelligence within the organic world all basically come from the ability of emergence.
Of course, this can be a perception, not confirmed scientific data. But I do know many chopping – edge AI researchers who imagine that the intelligence of AI comes from emergence. In the trade, emergence is manifested because the scaling regulation. It can also be an engineering guideline. If you enhance the dimensions of computing energy, the efficiency of the mannequin will enhance.
So, emergence will not be solely a scientific idea that helps us perceive how intelligence is generated but additionally a tenet within the trade. Naturally, it’s the first and most vital idea.
Jingxiang Studio: I used to be most impressed by your writing that each Chinese individual ought to perceive emergence as a result of the reform and opening – up was an instance of emergence.
Zhang Xiaoyu: Indeed, that is one thing that each Chinese individual can really feel. The market economic system can also be an emergence within the human social system. Just take into consideration what our lives have been like earlier than 1970. It was unimaginable that there can be monetary firms, Internet firms, and so on. at this time.
The essence of the market economic system is to use cash as the only sign for the emergence of this advanced system. Its guidelines are easy sufficient. Everyone would not want to focus on concepts, ideologies, religions, or languages. As lengthy as you acknowledge cash, you may function throughout the similar system, and this method will generate numerous types of knowledge and creativeness.
Jingxiang Studio: You wrote in regards to the idea of “human equivalence” proposed by Leopold Aschbrenner, a former worker of OpenAI. It refers to what number of people an AI is equal to when it comes to effectivity and price in producing intelligence, measured in tokens. You additionally wrote that “sometimes, the operating rules of society are determined by the simplest mathematical laws,” and “human equivalence” is the mathematical regulation that performs a decisive function within the AI period. Can you clarify the underlying logic?
Zhang Xiaoyu: First, why are the working guidelines of society decided by the only mathematical legal guidelines? This is kind of easy. For instance, in historic information, behind the rise and fall of emperors and dynasties, the only mathematical regulation is the Malthusian lure. In historic instances, the meals manufacturing of the land elevated linearly, whereas the inhabitants elevated exponentially. An individual within the Qing Dynasty wrote that the inhabitants might enhance by 128 instances in three generations. The exponential development of the inhabitants and the linear development of meals manufacturing inevitably led to conflicts. When meals manufacturing couldn’t meet the inhabitants development, there would naturally be wars, famines, plagues, and uprisings, adopted by the change of dynasties.
Behind the assorted twists and turns in historical past, it’s truly the fundamental mathematical legal guidelines which might be at work, and different components are much less vital variables. I usually say that if you happen to have been to journey again in time, you do not want to know which emperor is in energy or what insurance policies are being carried out. Just calculate the mathematical relationship, and you may inform whether or not you’re within the early, center, or late stage of a dynasty.
Similarly, within the AI period, the only mathematical relationship is human equivalence. Whether we’re chatting or I’m writing, in essence, we’re all outputting intelligence. When chatting, we are able to produce about 200 tokens per minute, and at most 200,000 tokens per day. However, a big – language mannequin can generate 1 million tokens in only one second, and what’s extra, it solely prices one yuan. Of course, in actuality, if the content material is significant, it would positively take multiple second and should price a number of {dollars}, nevertheless it’s nonetheless less expensive than human labor. In Beijing, if you happen to pay me 100 yuan a day, it is arduous for me to survive, however a big – language mannequin can do my work for a number of days for only one yuan.
So, so long as your boss is calculating ROI and we’re working throughout the capitalist system, you may predict what is going to occur. All speculations in regards to the AI period are primarily based on this mathematical relationship: AI is extremely low cost. In phrases of skill, in accordance to some check outcomes, all massive – language fashions will attain the extent of a doctoral pupil by 2025. Only about 1% of educated people have a doctoral diploma. So, numerically, it is already a incontrovertible fact that AI can produce extra intelligence than 99% of people at 1% to 1‰ of the price.
Will we stay higher after 99% of human jobs are changed by AI?
Jingxiang Studio: In phrases of each price and effectivity, people are like “pre – historic creatures” in contrast to AI. However, we do not say that we’re pre – historic creatures in contrast to steam engines or meeting traces. Compared with earlier industrial revolutions, why does AI deliver such a powerful sense of disaster to people?
Zhang Xiaoyu: During earlier industrial revolutions, there have been additionally discussions amongst technological philosophers. But for 1000’s of years in human society, most individuals have thought of psychological labor superior to bodily labor. Those who use their minds rule, and people who use their power are dominated.
AI straight produces intelligence, which makes it completely different from all earlier applied sciences in historical past. Technologies just like the printing press, the Internet, and software program engineering in historical past have been all about packaging and transmitting intelligence, not producing it. Our species has all the time thought of itself probably the most clever on Earth, and that is now being challenged.
Secondly, most individuals overlook a truth. 100 years in the past, the tempo of technological progress was coordinated with the pure generational substitute fee of people. For instance, if you happen to grew to become a coachman round 1860 and have been in your thirties when Benz invented the car, though it was sluggish at first, if you happen to have been a rational center – aged individual, you’ll know that it could ultimately substitute you. But it did not matter as a result of if the prototype was made in 1880, it would not be mass – produced and on the streets till round 1890 or 1900, whenever you can be shut to retirement. And your baby, rising up, would know that this factor would substitute coachmen, so he would not select to study this occupation.
That is to say, the cycle of technological progress at the moment was about 20 years, roughly one technology. By the time the brand new expertise matured, the previous technology would retire and the affect can be naturally absorbed. However, at this time, ChatGPT has reached the extent of a doctoral pupil in 4 years, and it takes greater than 4 years to get a doctorate. The fee of technological progress is totally completely different, and we won’t stick to the previous methods of considering.
Another situation is that I believe we’d like to re – perceive the historical past of your entire industrial revolution. In the primary two industrial revolutions, the expertise was that the extra expertise superior, the upper the per – capita GDP development fee, and it was usually a helpful factor for everybody.
But there was a turning level after 1970. There is a really well-known saying in financial historical past: “Where is the computer?” In the Nineties, computer systems could possibly be seen in every single place in day by day life, however why wasn’t it mirrored within the per – capita GDP development?
This is a really apparent contradiction. When folks studied this phenomenon, they discovered that on the micro – degree, completely different applied sciences within the industrial revolution had qualitative variations. The applied sciences within the first two industrial revolutions usually prolonged the economic chain. For instance, the steam engine, the electrical gentle, the phone, the fridge, and the car. When these new merchandise have been invented, there have been a whole lot or 1000’s of business chains behind them, and every industrial chain concerned 1000’s of firms and staff. The industrial chain was continuously increasing, creating new jobs at each step, so it was a inclusive development.
However, for the reason that Seventies, applied sciences like automated CNC machines have been shortening the economic chain, and staff have been displaced from the meeting line. From this attitude, we are going to discover that lots of our frequent – sense beliefs are unsuitable. I keep in mind in center faculty, the textbook stated that within the industrial construction of developed international locations, the service trade accounted for the next proportion. But if you consider it, it isn’t {that a} increased proportion of the service trade means extra superior. It’s as a result of automated machines have displaced staff from the meeting line, and these staff have to flip to the service trade.
In the AI period, this phenomenon is much more apparent. What the machines within the Seventies did to blue – collar staff, AI is doing to white – collar staff at this time. I believe we are going to quickly see a really unusual phenomenon: the decoupling of technological progress from the inclusive development of people and per – capita GDP.
● Image supply: Modern Times
Jingxiang Studio: We usually say that expertise liberates folks, nevertheless it would not appear to be the case?
Zhang Xiaoyu: Compared with the Middle Ages earlier than the economic revolution, our present residing requirements are positively significantly better. But we are able to solely say that expertise needs to liberate folks, however folks could not need to liberate themselves. That’s why there are “bullshit jobs.” Many instances, bosses do not actually suppose that the work brings a lot worth, however moderately that it offers them a way of energy. The incapability of individuals to liberate themselves is a higher tragedy.
Jingxiang Studio: Your view of human nature can also be mirrored within the idea of “algorithmic judgment.” Different from many people who find themselves important of algorithms, you suppose that algorithmic governance is far fairer than human governance. There is a passage within the e book:
The advice algorithm offers every individual what they deserve… If we imagine that we should always exploit and oppress our friends, then it would suppose that people ought to be exploited and oppressed; if we imagine that we should always struggle and kill our friends, then it would suppose that people ought to be handled violently; after all, if we imagine that everybody deserves to be liked, then it would suppose that people are worthy of affection.
Zhang Xiaoyu: When we speak about advice algorithms at this time, there’s a idea of being trapped within the algorithm. But to put it bluntly, the roles supplied by the advice algorithm are literally fairly good. I usually say that the highest 3% of earners in China are delivering takeaways downstairs in first – tier cities. How can we perceive this?
Among China’s 880 million labor power, the tax – paying threshold is 5000 yuan, and the variety of folks paying particular person revenue tax is about 60 – 70 million, accounting for about 7% of the overall labor power. If the tax – paying threshold is raised to 8000 yuan, it is solely 3%. By registering as a Le Pao rider on Meituan, the typical month-to-month revenue final 12 months was about 8000 yuan. This signifies that these “trapped in the algorithm” are already within the high 3% of the system. Intellectuals suppose that being a rider is difficult, however that is as a result of they have not labored in a manufacturing facility. If they did, they’d discover that human – to – human remedy is much more merciless than algorithm – to – human remedy. The advice algorithm gives primary employment for a lot of folks and should even be fairer than native state – owned enterprises within the Seventies.
The core of algorithmic judgment is that this. We should not suppose that people deal with one another effectively and that algorithms have ruined the scenario. In truth, people deal with one another badly, whereas algorithms are impartial. They simply deal with everybody in accordance to their guidelines.
Jingxiang Studio: Although 99% of human jobs can truly get replaced by AI, is it doable that we should always artificially create some jobs?
Zhang Xiaoyu: It’s doable. I additionally advocate for this. There has to be one or two generations to naturally transition and see what the following – technology expertise can be like. So, I’ve an idea known as UBJ (common primary jobs) on high of UBI (common primary revenue). For instance, a number of years in the past, when Marco Rubio, the present US Secretary of State, was a senator, the small – and – medium – sized enterprise committee beneath him issued a report. Concerned about massive – scale unemployment attributable to automation and synthetic intelligence, the proposed answer was for the US to arrange various state – owned enterprises. Each state within the US has its personal public