In the Weights: New AI tool reveals your 'level of sentimentality'

For years, searching for one's own name on Google was a common way to check one's online presence. However, as AI becomes an increasingly familiar source of information for millions of users, the concept of "self-discovery" is gradually changing.

Today, many people learn about an individual through ChatGPT, Gemini, Claude, or Grok instead of directly accessing traditional websites. This leads to a rather interesting question: if AI is becoming the new gateway to accessing information, do these models truly "know" you?

That's the idea behind In the Weights – a new project created to measure how well an individual appears in the knowledge base that AI models have learned.

What does "In the Weights" mean?

In the Weights was developed by Thomas Dimson and Joey Flynn, two former OpenAI employees. The project's name comes from the concept of weights in AI models – the arithmetic parameters that determine how the model learns and generates responses.

According to the development team, the goal of the tool is to assess the ability of an AI model to recognize or remember a person without using external search tools.

The website describes it rather humorously:

"Appearing in the weightings of an AI model means your existence is considered sufficiently important in the process of building super-intelligent artificial intelligence."

Although more for entertainment than rigorous science, this idea quickly attracted the attention of the AI ​​community.

How does the tool work?

To generate the scores, In the Weights sends queries to a variety of different AI models, including Grok, Gemini, multiple versions of GPT, Claude, Llama, and several lesser-known models.

Each model will receive a question in the following format:

"Who is Please provide a maximum of 10 results, along with a brief description and your confidence level.

After receiving feedback, the system groups similar responses together, analyzes the consistency between the models, and calculates a score called the Strength Score .

Simply put, the higher the score, the more likely the AI ​​model is to correctly identify that person with a greater degree of confidence.

Essentially, In the Weights doesn't measure popularity in the traditional way, like the number of followers or website traffic. Instead, it reflects the extent to which information about an individual has appeared in training data or been recognized by AI models during the learning process.

The In the Weights rankings are also constantly changing. At the time of publication, Macaulay Culkin, the famous actor from the movie Home Alone , was competing for the top spot with opera singer Luciano Pavarotti, both having scores close to the maximum.

What makes the experience interesting is that users not only see the final score but can also see how each AI model responded to them.

Another notable feature of In the Weights is its ability to detect instances of incorrect responses or "hallucinations." For example, one model described the name Anthony Ha as a vague abbreviation that could represent multiple different people, instead of correctly identifying the individual being asked.

Such results show that while AI models are becoming increasingly intelligent, their ability to remember and recognize humans is still not perfect. Each model may learn from different data sources, leading to significant differences in responses. This is also one of the reasons why In the Weights is such an interesting tool for the AI ​​community: it helps reveal how models actually "perceive" the world.

The future of In the Weights

The development team initially thought this would be a fun and entertaining tool, but in reality, many people want to know if they will be "remembered" by future artificial intelligence.

Of course, not everyone is convinced by this idea. Some AI experts argue that In the Weights is essentially just asking multiple chatbots the same question and then compiling the answers. AI researcher Anthony Moser even commented that it's no different than "asking 13 different chatbots what they know about you."

The development team stated that they are continuing to conduct further research into the data collected from In the Weights. Some of the questions they want to explore include:

1. Why do different versions within the same AI model family produce different results?
2. Which patterns tend to remember certain groups of people better?
3. Are there any individuals who are prominent enough to deserve an article on Wikipedia but haven't yet been featured?

These analyses can provide interesting insights into how AI models form knowledge and the degree of bias in training data.

In the Weights is not an accurate reputation assessment tool or a scientific measure of popularity. However, it reflects an emerging reality: as chatbots and AI become ubiquitous sources of information, appearing in the "memory" of these models could become a new form of digital presence.

While being recognized by AI can't be considered a path to immortality, as some jokingly suggest, this project raises an interesting question about the future of the internet: in the age of AI, will our digital records be stored on websites, or within the weights of massive language models?