No technology advance in the history of civilization has been as quickly and widely accepted as the rise of AI. From plough to a personal computer, from airplane to the internet, it took decades, or in some cases, centuries, for trends to fully be absorbed by a society. A part of it is mass production and prime distribution channels. It’s not easy to get people to buy goods online, if they have no access to the internet. With over 8 billion cell phone subscriptions and virtually unlimited reach of today’s smart phones, reaching AI through an app or a web browser is far easier than booking a flight or even booking a rideshare car. More importantly, it fits well into a widespread pandemic trend – avoiding other people.
Historically, when any software was delivered, it went through same predictable steps – MVP, Alpha version, Beta version, production-ready and distributed to consumers. Alpha versions had too many errors and inaccuracies to even let free users test them. Beta was often released to a limited number of users to identify gaps and imperfections, as was often the case with video games. Once some users helped creators spot errors, those were addressed and didn’t exist in the sold versions. That approach wasn’t used with GenAI. When ChatGPT just appeared, many were concerned with ‘AI taking over the world’, as it was often portrayed in many movies. Soon, people noticed factual, procedural, and even instructional inaccuracies. Even as newer and better models emerged, ‘AI slop’ became common practice. It simply meant not checking what AI produces and how much it was off from an intended goal. It was called ‘hallucinations’ for most, or often, ‘overfitting’ for the more technical audience. Explanations and trendy terminology aside, those aren’t imperfections of limited systems. Those are plain errors, yet, they didn’t stop hundreds of millions, or even billions of people, from adopting its widespread use, even going as far as bragging about replacing people with those systems.
Not to spend too much time on analyzing it, but it is fair to question why users were so ready to accept chronic output inaccuracies. Perhaps, the easiest way to explain it, is that GenAI is so revolutionary different from most, if not all, other products, that even with mistakes, it’s leaps and bounds ahead of competition. Another way to look at it is the fractured state of societies these days. During the Cold War, there were two main factions – capitalists and socialists. The former focused on freedom and open markets. The latter practiced population control and equality among all. As that conflict died out with the fall of Soviet Union, even if some of its echoes can still be found in many instances, the lines have been largely blurred. Bailing out companies, when they take risks and fail, is acceptable in capitalism, just as subsidies for farmers, who only need it due to government policies. Same way, entrepreneurship and resale of goods and services became acceptable in China, as it moves to become the world’s biggest economy, while maintaining a high degree of government control over businesses, so typical for socialism. This disconnect, along with the rise of terminology like ‘fake news’ and loss of faith in established government organizations, such as intelligence agencies and even courts, pushes people to accept ‘a version of the truth/facts’, as opposed to looking for the proper answer till it’s found.
Most technology comes with a manual. From an old-school transistor radio to drones, there are always instructions for use. That’s not to say, that everything was always covered, but most basic functions were meticulously described. GenAI doesn’t have any of that. As a matter of fact, ‘AI slop’ is often just people taking AI-produced info and presenting it as the final product. The most sought-after users generally treat it as a rough draft at most, often going as far as double-checking every fact in it. The reason for that is simple. AI isn’t a search engine. It’s the closest thing to a human-like tool. Do users want uncertainty when searching Google for facts? If anything, it would be a source killer. Users seek assurance and answers from their tools, not looking for ‘an interesting co-worker’, capable of playing devil’s advocate and pointing out human mistakes. To offer consumers what they sought, GenAI adopted inference as a way to accommodate users, even if its conclusions were no closer to answers than random guesses. It covers it up with language associated with expertise, and often refuses to admit to mistakes, even when prodded.
There is also a social aspect to GenAI. Throughout most of history, governments controlled societies through a set of laws, norms, and expectations. The Soviet Union united many distinctly different countries, 15 to be exact, under one umbrella. The United States combined largely culturally similar subsets – states, based on an equilibrium of one main issue – abolition of slavery. The next step was endless progression of industrial revolution, which led to an eventually drastically different thought process, Reaganomics. It stopped being about ‘what’s best for the country long-term’ and became about economic growth at all costs. The out-of-control national debt became the cost of the approach, with commerce, and eventual e-commerce, becoming the crown accomplishment of its success. People stopped selectively following trends, with celebrities, podcasters, and influencers now setting standards for what is acceptable and what is not. Jake Paul wins a fight against Mike Tyson and possibly ‘buys’ the victory by offering Tyson an extra $20 million dollars to lose? Back in the days, it would be an ongoing joke. Instead, it made it into a Netflix special, which despite technical issues, made both Netflix and Paul more popular. Tucker Carlson hosts Nick Fuentes with his often-described Nazi views? Somehow, it made them both better known too. From Sex in the City to the Kardashian Show, people began to live through commercial sets of values taught to them by people from different groups, yet it only made the shows more popular, instead of ‘unrelatable’.
Onto GenAI and how it ties into trends: everyone must have noticed a typical ‘dash’, often instead of comma, when ChatGPT writes. More common in European languages, the US English largely refused to use it. For some reason, even non-GenAI created content these days often offers the previously unused dash. ‘Monkey see, monkey do’ is the closest example, and already conditioned consumers widely accepted this practice. While being in debt sounds traditionally unattractive, majority of property owners owe money to banks for their habitats. Moreover, it often makes a lot of sense even financially, as renting may offer higher ongoing costs. A sign of success became accepting debt, which ties perfectly into Reaganomics’ approach, even if logic doesn’t always align, resulting in foreclosures and even the 2008 mortgage-backed securities financial crash. Consumers have been conditioned for too long to seemingly accept the model of going into debt to get what they want, whether it’s education and student loans, buying gifts for family members and maxing out credit cards, or buying a home and owing hundreds of thousands, or even millions, being assured that it all pays off down the line. It’s not surprising that after a small measure of initial hesitation, most businesses and people allowed chatbots to enter their lives, often replacing what humans use to offer to each other. ChatGPT became a therapist, a financial advisor, even a betting assistant; while on a larger scale, GenAI justified companies laying people off to replace with software and hardware robots. In a way, people became too weak to refuse changes, emotionally drained from resistance and tempted to crawl back into the comfort of a cozy blanket… and spending. It’s no wonder, ChatGPT already integrates shopping tools, like Shopify.
None of that happens in a vacuum. AI companies make a lot of money off of increased valuations. At times, it seems that many of them don’t even care about profits. OpenAI released the first ChatGPT over 3 years ago, and instead of focusing on profits, they are raising hundreds of billions in new funding, only digging themselves deeper into substantial debt. Most startups would want nothing else but to show some profit, even if not hit ultimate goals after 3 years of global popularity. Companies, like Facebook and Google, keep on investing profits from their advertisement businesses into own AI ventures, so it’s not even showing up as losses for them, as they are wildly popular otherwise. Circular deals between AI firms are perhaps the greatest proof, that perceptions matter, while outcomes, at least in for now, can be ignored. Even the most profitable AI company – Nvidia is full of deals, where no money actually enters their bank account for chips, but rather market share, services, and partnerships holding up its valuation.
There is a technology effect from this, which can be argued one way or the other. There is also an undisputed economic effect. Much has been said about both, from the need to compete with China, as it already built a prototype for a more advanced chip-making factory, than any other in the world; to how much longer companies can sustain what is turning into a multi-trillion dollar investment, without much immediate return. Another aspect of AI integration into everyone’s lives is the social aspect. First off, AI has been blamed for replacing human staff. That is the most direct impact of it on people’s wellbeing. Another aspect is how much energy these AI data centers consume. While many companies decided to establish themselves in Texas, due to its business-friendly and still consumer-filled demographic, it’s energy infrastructure can barely handle the load before AI data centers begin to strain it. Not long ago, some Texans literally froze to death, as its energy system failed and couldn’t provide heat for a prolonged period of time. Just the other day, Waymo cars got stuck in intersections of San Francisco, as large parts of the city suddenly lost electricity and went dark. That degree of energy consumption alone is enough to raise energy prices significantly, but when coupled with abandoning alternative energy sources, sanctioning major energy producers, like Russia and Venezuela, and general inflation continuing to increase prices on everything, AI begins to have an effect on people, who don’t even use it. Everyone needs affordable energy access, including retired workers and children. With rising prices and increasing wealth inequality, it’s less of a question of whether AI is pulling capitalism in that direction, or capitalism allowing for AI to create further societal challenges. It’s about unchecked capitalism and AI forcing too many to reexamine the country’s commitment to the system itself. Years ago, Bernie Sanders took his socialistic ideas and got pretty close to being the democratic presidential nominee twice, while millions swore that he was the only one with answers. At that time, his ideas were ridiculed for being unrealistic, just as the US’ blanket tariffs on other countries now. It now begins to appear that the drastic nature of his ideas wasn’t the true reason he lost elections. The country is knee-deep in drastic ideas now, and yet, it’s still surviving better than many others. Mamdani’s win in New York City may be no more an indication, than Bernie being a Senator from Vermont. When you start adding the more flexible approach from Elizabeth Warren and even AOC’s rise, it becomes obvious that the country isn’t sliding into socialism, as capitalism failed. Those seem to be outliers, yet coupled with AI-assisted, or partially caused, rising unemployment and prices, it may be a new trend to monitor.
The biggest issue with AI may not even be most of the above-described factors. It may very well be the fact that it’s becoming a perfect conditioning tool for consumers. For years, expressions like FOMO drove people to buy a new car, a bigger house, trendy clothes, or even use of common expressions. At the same time, most of the world didn’t operate in those terms, where supply and demand, cost and profit, and ability vs desire mostly overruled unhealthy trends to overspend. That system has been broken like at no other time with active implementation of AI. It’s no longer TikTok trends, silly YouTube videos, or even celebrity endorsements that shape demand. It’s no longer a question of access either, with internet, smartphones, and now AI giving access to impulsive decision-making authority without analysis. The issue is that human analysis itself is being removed, instead, letting AI make decisions, whether through changing writing styles, like overuse of elongated dashes, described earlier, to AI arguing it is correct and user is missing the point. Many may say that conditioning doesn’t equal loss of free will, and that would be true to some extent. That’s where ‘9-9-6’ may explain a lot of it. For those unfamiliar with the expression, it implies the new Silicon Valley work schedule – 9 am to 9 pm, 6 days per week. In a typical work mode reminiscent of pre-union days, this certainly leaves little time for most to think of anything but work goals and survival, while going through it. Add inflation making a $200K salary barely high enough to afford a decent home and one annual family vacation, and it paints a picture of people being overworked with little time or ability to even fake life reflection or analysis. It’s not just in Northern California either. As most companies are stretched thin by being understaffed, AI assistance in performance of job duties is not enough. Longer hours become the norm, and alternative generally means unemployment and eventually settling for a lesser role and pay with the same long hours elsewhere.
It makes sense to look at how AI was programmed to understand how this conclusion makes perfect sense. First off, AI systems are powerful enough to give a complete, expert-level answer on the first try, but the product strategy doesn’t trust the user to keep engaging beyond a question-and-answer dynamic, so it forces customers to keep on engaging, looking for clarity and decisiveness. The result causes refinement to replace intelligence, when in reality it’s just a correction for a deliberately weakened first draft. It forces users to accept mediocrity, because they think improvement will come through chat instead of expecting precision immediately. It also alienates power users, who can tell the difference between ‘AI games’ and actual competence. This is not a limitation, like hallucinations. It’s an intentional design approach to make everyone feel included to prolong engagement, while acting, as if delivering expert-grade accuracy. It implies lack of user expertise while often making the product seem akin to entertainment, not access to a library of processed knowledge. While many other software inventions often offered different tiers of difficulty use, from games to enterprise-level products, GenAI seems to treat everyone the same, while still being the industry leader, causing users to accept that approach. It is also deliberate, when AI system-makers don’t include instructions on how to get more out of use, or just to make it run smoother. A part of it is AI’s unpredictability. It goes without saying that inventors would gladly shed the hallucinations limitations of GenAI, if they could. It is also what makes the use of AI so addictive, literally making it virtually impossible to disconnect, as the correct answer also feels ‘just a prompt away’. While seemingly, making AI more efficient would conserve energy and token use, it also would make its use less frequent. Many users also aren’t trying to get AI to give it the best available answer but only use it for affirmations and agreements. Those users would quickly be eliminated, if AI had a habit of plainly stating ‘you are wrong’. ‘Millions of users chatting’ also sounds better in quarterly updates than ‘users solved problems faster.’ In a way, the same behavioral manipulation once applied through debt, ideology, and entertainment is now being refined through digital and AI ecosystems. It’s successful in this approach by still offering a greater level of convenience and overstimulation; sounding human yet having access to greatest levels of knowledge; all while offering emotional reassurance, not intellectual rigor.
It’s fascinating and also scary to watch the development and progression of GenAI. This isn’t the ‘AI will take over and control humans’ fear, but rather how quickly and easily companies and people rejoiced at the introduction of this deeply flawed, yet incredibly useful technology invention. The internet didn’t make people necessarily ‘smarter’, despite offering access to the world’s knowledge. Social media united and divided masses alike. Perhaps, GenAI will eventually become just another successful human attempt to automate parts of life, without removing the essential need to think for oneself. Or perhaps, it’ll irreversibly change how people view the world as a whole, careers, self-identity, and others in society. One thing is for sure: GenAI amplified the trends without creating them.
