AI Doomer vs Techno-Optimist: Social Fabric, AI Girlfriends, Mass Media
China Demographics, Character AI, Meta, Influencers, Incels, and Insurrectionists
As the world gets increasingly connected, the social fabric of society seemingly is more disconnected due to the negative effects of social media. The human mind becomes increasingly vulnerable to the adversarial designed reward systems of firms like Meta, Bytedance (TikTok), and Google. There are two ways to look at the new issues that are arising.
One viewpoint is that of an AI technology doomer. The end state of the unchecked optimization pressure brought about by deep learning recommendation systems and generative AI will result in a subpopulation of terminally addicted individuals that are isolated from each other by the design of the system.
These people will be locked into an extractive relationship with the machines. It’s easy to get depressed seeing the declines in rates of people hitting various developmental milestones. This includes rates of forming friendships, having sex, and marriage. In the same vein, there are also clear increases in rates of obesity, mental health issues, drug use, and suicide.
On the flip side, there is a techno-optimistic capitalistic way of viewing the modern problems of society. Humans’ needs have been met on most levels of Maslow’s Hierarchy, and now we are facing a new class of problems that we have previously not had to previously deal with on this scale. Currently our machines are engineered to give us a constant stream of dopamine, but they can also be engineered for longer time biased reward systems. For techno-optimists, the ingenuity of humanity will be the driving force in innovating away many of humanities’ newest problems.
SemiAnalysis is an ad-free reader-supported publication. To receive new posts and support, consider becoming a subscriber.
While most of the focus is on productivity and automation, there is a huge market for AI being implemented into other use cases including healthcare. Furthermore, even beyond those tangibles, the market for AI in the human elements of socializing, counseling, therapy, support, and personal development are massive. Character AI is one of the most innovative and unique startups in the world because they are tackling a problem many seem to ignore.
We will analyze the current raise and their future, given Meta/Chai/Minemax attempts at aiming for similar markets, but first let’s talk about a societal and demographic problem we feel the entire world is ignoring. This has huge implications for the demand of AI, semiconductors, and more broadly geopolitical stability. Stick with us for this ride, and don’t worry, it will loop back to the topics we usually cover.
We want to note the below is focused on population level historical and biological statistics. No harm is meant to any marginalized communities.
The One-Child Policy
In 1979, the CCP introduced the “One-Child Policy”, which sought to curb the country's exponential population growth to alleviate pressures on various vital resources, facilitate economic expansion, and reduce strain on welfare systems and the state-planned economy. This drastic measure, according to China’s estimates, prevented approximately 400 million births, substantially slowing down population growth.
The policy was eventually relaxed in 2015, but the societal consequences have only just started to play out.
Sex Ratio At Birth
Many Asian cultures traditionally prefer male offspring due to expectations of future elderly care and family wealth inheritance by male descendants. This preference has unfortunately fostered gender-based discrimination, leading to a concerning practice of sex-selective abortions. In these cases, families often choose to abort female fetuses, aiming to ensure a male heir to their lineage. Consequently, this practice has contributed to a significant gender imbalance in the population, with a skewed male-to-female ratio.
The natural sex ratio at birth typically ranges from 103 to 105 boys for every 100 girls. This slight male birth prevalence is biological in nature and believed to be because of the relatively higher male mortality rates observed, from infancy through adulthood.
During the 1950s, China's sex birth ratios were within the natural range. However, the introduction of the One-Child Policy altered this balance. The growing accessibility and affordability of prenatal sex-determination technologies, like ultrasound and amniocentesis, enabled parents to predetermine their child's sex. This enabled the natural preference of parents for male offsprings to become exhibited in their offspring. This development contributed to a significant deviation in the sex ratio, favoring male births.
By the early 2000’s, China’s sex birth ratio reached as high as 120 boys for every 100 girls!
This problem is not too big of a deal, until these cohorts come of age.
The Marriage Squeeze
As the cohorts age, generally they want to get married, but due to decades earlier shifts in the sex-birth ratios, the problem of the “Marriage Squeeze” arises. A 120 to 100 ratio is bad enough on its own, but that only scratches the surface of the problem. As people pair off, there will be a segment of the population that remains in the market that continues to look for a spouse.
This growing imbalance between the number of men and women of marriageable age. According to research by Kuangshi Huang, there will be a surplus of 30 million excess males in 2028, and will reach as high as 41.4 million excess males by 2043.
China will have about as many men who can never get married as entire countries such as Canada, Poland, Argentina, and Spain have total population!
The Urban-Rural Divide
In general, people pursue significant others of similar or higher socio-economic status. In China, income, education, and ultimately socioeconomic status generally correlate with urban vs rural divides. Due to the huge surplus in marriage age males, females generally have the option of marrying higher-income and status urban males.
As such, there is a larger influx of rural females migrating to Tier 1 and Tier 2 cities. This keeps the sex ratio for single males manageable in cities. It is tougher for many males to immigrate to Tier 1 cities for work because of China’s Hukou System.
This system is essentially a household registration system that serves as a domestic passport, regulating population distribution and rural-to-urban migration. Every individual is assigned a hukou at birth, which is categorized into either a rural or urban hukou, and it is attached to their place of origin. While there is some relaxation of these rules in recent years, in general, it is still very difficult for a single male to immigrate to a city, while a rural female can immigrate if they are marrying an urban male.
In 2010, there were already ~500 single males for every 100 female at age 35 in rural China. In 2010, there were only 10 million the total number of surplus males.
Recall earlier, that the number of surplus males jumps 3 fold to 30 million surplus males in 2028. Furthermore, there will be as many as 41.4 million excess males by 2043. A 4-fold increase in surplus males from what the two charts above describe, so the problem is only getting worse, and rapidly. Some disagree the gender balance is quite this bad, but even the most optimistic claim it is still well over 10 million.
The demographic ratios reach a breaking point with cohorts that comes of age in 2028. 16% of males that come to marriage age in 2028 will never have a chance of getting married according to nuptiality tables calculated by Quanbao Jiang, Marcus W Feldman, and Shuzhuo Li.
The pronounced demographic shifts occurring in China present significant societal risks. Historically, societies with a surplus of men who have limited prospects for marriage, employment, and social advancement are susceptible to radicalization. These individuals often have little to lose and may be more willing to engage in activities that “rock the boat”. For many, the formation of a family and offspring acts as a stabilizing force, fostering a vested interest in societal well-being and disincentivizing actions that could jeopardize the status quo.
Without the hope of creating family units and securing a prosperous future for their offspring, many men could become disillusioned and detached from societal norms and expectations. This is exacerbated when the rate of economic growth slows significantly, further highlighting the lack of annual improvement in an individual’s life. These men would be concentrated in the aging rural areas, offering a potential area for radicalization, conveniently located next to critical raw materials and farming. This can be a risk for political stability.
On the other hand, this surplus of men can be mobilized for national causes that require a robust workforce or military presence. For instance, in facing economic slowdowns, a unified populace can be a potent force in driving economic revitalization and sustained growth. The surplus of men can be engaged in large-scale infrastructure projects, technology development, and other initiatives that not only provide employment but also stimulate economic activity and progress.
Furthermore, a unified country can leverage its surplus male population to advance specific political goals. For example, in hypothetical geopolitical scenarios where military force may be required, such as an invasion of Taiwan, having a large pool of potential draftees can be a strategic advantage. Engaging this population not only in military efforts but also in nation-building activities can foster a sense of purpose and belonging among these individuals, mitigating the risks associated with their demographic profile. This may be what certain leaders are even counting on.
Even ignoring the demographic nightmare that is playing out in many countries, there is an immense need for companionship in our society. The natural relationships people have with other humans will not always be achievable for tens of millions if not hundreds of millions across the globe, and as such they will be incredibly lonely.
Many have already turned to fostering parasocial relationships with streamers, influencers, and micro-celebrities, but those alone aren’t fulfilling. AI girlfriends and AI boyfriends will be far more fulfilling emotionally and mentally with personalized responses, especially with diffusion and voice models sitting behind them.
The above was more AI-doomer, but let’s get back to our preferred thought pattern, techno-optimism. There is a potential for AI to help alleviate not just this problem, but many others in terms of companionship, therapy, emotional support, personal development, and behavioral change for all types of people, regardless of gender.
Here’s a list of ways they could be used, ironically generated by ChatGPT. While the list is quite optimistic, it’s noteworthy that there are plenty of people already reporting using models in some of these ways.
Virtual Companions: LLMs can serve as virtual companions, providing conversation and interaction for individuals who may feel lonely or isolated.
Social Simulation: By simulating social interactions, LLMs can help individuals practice social skills in a safe and controlled environment.
Counseling and Therapy:
Mental Health Screening: LLMs can be used to conduct initial mental health screenings, helping to identify individuals who may need professional help.
Cognitive Behavioral Therapy (CBT) Support: LLMs can assist in delivering cognitive behavioral therapy exercises, helping individuals to manage symptoms of disorders like anxiety or depression.
Mood Monitoring: LLMs can be used to track an individual’s mood over time, providing insights into emotional patterns and triggers.
Crisis Support: While not a replacement for professional intervention, LLMs can provide immediate responses in crisis situations, offering support until professional help can be accessed.
Education and Awareness:
Mental Health Education: LLMs can provide information and resources on mental health topics, helping to raise awareness and reduce stigma.
Stress Management Techniques: They can educate individuals on various stress management techniques such as mindfulness, breathing exercises, and relaxation techniques.
Mindfulness and Meditation Guidance: LLMs can guide individuals through mindfulness and meditation exercises to promote mental well-being.
Motivational Support: By offering encouragement and tracking progress, LLMs can help individuals stay motivated towards achieving personal goals.
Habit Formation: LLMs can assist in the formation of new habits by providing reminders and positive reinforcement.
Smoking Cessation and Substance Abuse Support: LLMs can provide support and resources for individuals looking to overcome addiction.
This is all hypothetical, but it the usage statistics are very clear. There’s millions of Character AI users who uses the product for more than 2 hours a day. They simply love talking to it. The potential benefits are not yet known.
Influencers, Incels, and Insurrectionists
Going back to the AI doomer viewpoint, there is clearly going to be a decline in social and soft skills. The end state of the unchecked optimization pressure brought about by deep learning recommendation systems and generative AI will result in a subpopulation of terminally addicted individuals that are isolated from each other by the design of the system. Many of the use cases, especially with Chai and Character AI are sometimes just softcore erotic content.
In the case of the China example, MiniMax, raised $250 million dollars and was seeing soaring usage with a very similar product. Once AI girlfriend screenshots started circulating, it got taken off the internet. The government there has a huge morality issue with erotic and gaming content. China's technological control clearly affects the development and distribution of AI.
On the other hand, it will just take time for them to understand how to align these models to be powerful and subservient to their needs. As a single male in his 20s, I think the generalization that young single men are the most vulnerable and manipulable population group is probably accurate given we are the least likely to think critically before acting.
What happens when models are aligned by bad actors who can utilize an LLM to spread propaganda through deep-rooted psychological methods?
In this context, it’s easy to imagine a world where leaking model weights much in the same way that the original LLAMA model weights were leaked could be a vector of attack for cyber warfare. Imagine a model that can run on device that appears to be tailored for a certain use case, such as companionship at first sight, but actually have a deeper-seated motives such as radicalizing incels or convincing a segment of the population of certain political views.
Countries can be overthrown by radicalizing their young males, and personalized vectors of doing so make it much easier. There was significant bemoaning of 4Chan and potential foreign interference in the 2016 US election, but what’s happening right now on Taiwanese internet is just a taste of what will be coming down the pipe due to the massive decline in the cost of human-level intelligence as a result of AI.
The AGI risk that no one seems to discuss but is entirely reasonable is that a vulnerable and manipulable group listens to the artificial intelligence they view as their only companion. Simp armies already exist in the context of parasocial relationships, why not one that is weaponized with personalized companions.
Startups, AI, Mass Media, Semiconductor Demand, and the Coming Wave
Enough philosophical ranting, let’s loop this back to semiconductors, AI, startups, mass media, and what’s currently happening in the market.