When I used to primarily work in web design, creating high-quality websites would lead to them appearing at the top of search results, and many people would view them. However, over time, quantity started to be prioritized over quality. Websites began to focus more on the quantity of information, and it became clear that they could never compete with social media in terms of the volume of information they provided. Naturally, I also began to shift my focus from my own website to sharing information on social media. With the widespread adoption of smartphones, websites evolved into applications, and front-end engineers began to overshadow designers. As a result, we entered an era of design absence, with a proliferation of ad-filled, uninspiring websites. I, too, stopped visiting websites as frequently.
However, thanks to the advancements in AI, today we can access high-quality information instantly without wasting endless hours on repeated searches. This has brought meaning back to creating excellent websites. When users ask AI questions, it presents reference websites, making it inevitable for websites providing high-quality information to receive increased attention. The era has come full circle, where I once created high-quality websites in the early days of the internet and reached more people than many large corporations' websites. It's happening again!
Looking back, it all started a few years ago (in 2012) when I received an email from Lancers (a Japanese business matching site that connects companies and freelancers) as they were launching a new service called LancersPro, saying, "You will be featured on the top page as a top designer." I wondered, why me? While I had taken on jobs from other business matching sites a few times, I had never accepted any work through Lancers, especially at the time when they had just started. But then, it hit me. I had been focusing on AI from an early stage, and I had been paying attention to Python since the time when there were only a few books about it in Japan, mainly translations of foreign versions. So, I thought, "Ah, it must be AI!" However, it seemed unlikely that I, someone like a punk rock prodigy matched with typical clients who had strong connections, graduated from the University of Tokyo's Faculty of Law, started a consulting company after working at a famous foreign consultinfirm, and had well-known companies as clients, would be a good fit. So, I canceled it.
A few years later, during the chaos and uncertainty of the global pandemic, I received notifications that several credit card companies were reducing my credit limits. I suspected it was because I was a freelancer.
After that, as the situation with the pandemic began to settle down to some extent, I started receiving notifications one after another from various financial services that my credit limits were being raised. You might think it's because the pandemic was calming down, but I noticed that AI's performance had improved rapidly around that time, mainly from the accuracy of translations. So, once again, I thought, "AI, isn't it?"
I'm a creator and not an expert, so these are purely my intuitive conjectures, but considering that the Transformer paper was published in 2017 and advanced companies began to put it into practical business, it all makes sense.
A few years later, as if in a déjà vu, I suddenly received a notification from Facebook, only few days apart, that "you have been recognized as a top fan" of an official account of a very famous musician with tens of thousands to tens of millions of followers.
Scoring is a technique and method used by financial institutions to quickly determine the lending limits. For example, it corresponds to the non-fleet classification in auto insurance. By scoring users in advance in several ranks, you can get quotes in a few minutes with a few questions. It's a convenient feature for both companies and users. However, this doesn't concern high-scoring users, but for those with lower scores, it can be a troublesome presence. If your score drops a bit, you may be able to recover it, but it won't be easy for someone with the lowest score to raise it. Troubles with payments on e-commerce sites, frequent complaints about products or services, very poor ratings on auction sites, etc. Companies that provide such services often also offer some form of financial services, and there's a possibility that they share data to some extent. Furthermore, food delivery drivers are already being scored based on their delivery performance and ratings, with high-scoring drivers being prioritized for order offers . (Note1)
Please keep in mind that this is my personal analysis and has no basis in fact.
So, is it a logical leap to think that SNS platforms have begun using such a technique?
Comparing the current Facebook with X (old Twitter), if I leave a comment on a famous account, on X, there are often too many comments, and mine gets buried. But on Facebook, by default, comments are displayed based on relevance, and only a few comments are shown. Naturally, comments from top fans are given priority. Additionally, both platforms used to be noisy with ads and recommendations. Even if you repeatedly clicked "Not Interested" every day, it never seemed to have any effect. However, Facebook now, once you notify them that something is not relevant, it stops showing it, but X remains unchanged. The most annoying thing is notifications on the lock screen. Recently, in the case of Facebook, I receive notifications occasionally, but X remains as it was, and I was forced to turn off notifications because I kept receiving notifications related to uninteresting accounts such as “Mr./Ms. XX has done YY”.
I'm not a Facebook advocate, nor do I want to brag about being selected as a top fan. I'm just talking about the benefits that high-performance AI can offer to users and companies that have adopted it.
Lastly, why did I pay attention to this subtle but important change in Facebook? Before changing its name to META, Facebook planned a financial service called Libra, which was so amazing that it faced resistance from financial institutions and governments around the world, and the project was abandoned as a failure. Libra is undoubtedly a financial service. What if, even if it's not all 2.9 billion users, they had already been scored?
The world we are currently living in has already transitioned into a data-driven society, powered by information. And in the future, it is bound to accelerate and expand further. Those who used to think that it was clever to get by with a casual approach should consider changing their way of life.
However, it's not all good news. The reality is that when all technologies rapidly become prevalent, downsides come along with them.
What if credit information leaks externally?
How can we prove that the assessment of scores is fair?
Is it possible to establish methods for using scoring that won't widen the wealth gap?
Moreover, if data like surveillance camera footage is used, there's a possibility of a surveillance society.
Furthermore, what if governments started scoring their citizens?
The challenges are indeed mounting.
So, how should we, as data survivors, navigate society in its current state? That's something I'll discuss another day.
Note1: In Japan, a company has already put such a system to practical use.
Bias and Noise
Decision-making is often referred to as "judgment," and it goes without saying that making the right judgments is essential for project success. There are factors, bias, and noise that can lead to incorrect judgments. Bias involves a skewed perspective, while noise represents variability. As organizations grow, it's easy to imagine that these factors also become more significant.
In corporate meetings where conclusions aren't reached, and discussions are carried over to the next session, and productivity suffers, these factors often play a role.
Fortunately, I am a freelance creator, and biases are often tied to one's personality (I have the innate ability to view my own bias to some extent objectively, and I've strengthened it through years of learning and training). Noise, on the other hand, blurs the concept and weakens appeal. However, this discussion revolves around my unique approach, so for understanding noise that typically occurs in organizations and systems, I recommend reading books by authors like Daniel Kahneman.
In my case, noise differs from the noise discussed in behavioral economics. Instead, it refers to the factors that can disrupt judgments when one often needs to make quick decisions in a profession that requires it. For example, in creative work, targeting the kind of user with a particular personality for a piece and how to reach them through media, or in practical situations, responding promptly and almost reflexively to unexpected client questions in presentations. As a lecturer, it means answering students' questions on the spot.
However, I recognize that explaining such specialized cases wouldn't be helpful to readers, so the main topic this time is the variability of information on social media, often referred to as "noise."
Noise is amplified by groups. The current internet, especially social media, is filled with noise. Selecting only high-quality information from this clutter is extremely difficult, but this ability can create a massive information gap. "In an information-aggregating society, information is the greatest asset," as Peter Drucker said. Nowadays, society is already an information-aggregating society. It used to be "money is trust," but now it's "information is trust." The scoring system I mentioned in a previous column is an example of this. Even if person who work for a large corporation and have a stable income and some savings, AI might still refer to the person's social media posts and e-commerce purchase history, among other information. Such actions are called "behavior," and AI analyzes individual behavior, calculating the level of trustworthiness and investment suitability. Someone who purchases expensive luxury brands that don't match their income to appear wealthier on social media may be considered a valuable customer by the luxury brand. But if that person asks you for a loan, how would you respond?
So, the main point is that I don't need followers because I want to eliminate the noise generated by noisy individuals. If I had tens of thousands of followers, most of them would undoubtedly become noise generators. "Tonight's dinner is..." "I went to a café in Daikanyama." "I took it with a Leica camera," etc. To me, amidst these posts, scrolling through thousands of timelines every day to find a handful of high-quality information is a significant waste of time. Furthermore, the number of views and "likes" on my posts becomes ambiguous, turning into noise. As a result, it can lead to misjudgment, as mentioned earlier. Kahneman also discusses how noise is amplified by groups in his book.
Noise requires a filter
The numbers on X (formerly Twitter) are unreliable. In the past, there were services offered by numerous businesses like "5,000 followers for 5,000 yen" or "3,000 likes for 1,000 yen." Such services were prevalent on various business matching apps. (There may still be such providers today.) Back then, it wasn't uncommon for university students with nothing special to offer to have 5,000 followers.
As of October 2023 when I am writing this article, looking at X (formerly Twitter), the recommended timeline includes posts in English by a popular overseas musician with around 30,000 views, posts related to stock investments by Japanese accounts considered unimportant with 1.15 million views, posts by Japanese accounts about to ChatGPT with 320,000 views, and two words posts in English by Elon Musk with nearly 1 million views. It's beyond analysis!
Based on my experience up to this point (I joined Twitter in March 2009), aside from posts by popular English musician and Elon Musk, the numbers are extraordinarily high, at the level of a singular point. This trend seems to have become the norm on X (formerly Twitter). These posts are quite lengthy, and it appears they are only available to paying users. How many users will actually read such long posts to the end? Even if it's one in a thousand, the actual views are probably closer to 100,000 views with a count of 1,000.
X (formerly Twitter) has a long history of inundating users with posts that hold no interest, notifying them massively as "recommendations" and displaying them in their timelines. While the number of views might increase, the fact that the posts are read is questionable. This can be seen as nothing more than material to help post authors hype themselves, but for me, it's material that can lead to misjudgment and noise.
That’s when I had an idea.
Sharing Posts from Facebook to X, does not count the views when you simply see the post. I noticed this and realized it was a perfect solution. Furthermore, I can post long texts for free. It's like hitting two birds with one stone.
By engaging in creative thinking daily and training my brain, I come up with ideas for activities other than creative work.
To obtain data with minimal noise, it is essential to devise how to filter out the noise. Filtering is often mentioned, but it's mostly about filtering spam emails and harmful websites. There are few discussions from the perspectives of behavioral economics and media analysis.
In a party full of noise, only the loudest voices are heard, regardless of their relevance. The internet is exactly like that. In such a situation, how can you effectively convey the right information to the right target without raising your voice too loudly, given that the current internet is full of noise?
For this purpose, some method to remove noise is necessary.