In the Weights: A New AI Tool Unveils How Language Models 'Remember' You

By Decode Today News

In the Weights is your new AI-centric vanity search AI

The traditional Google vanity search, once the gold standard for self-discovery on the internet, is facing a significant challenge from a new frontier: artificial intelligence. As more information consumption shifts to large language models (LLMs), a groundbreaking platform called "In the Weights" has emerged, offering a novel way to quantify a person's digital footprint within the very neural networks that power these advanced AI systems. This tool assesses how deeply and accurately major AI models recall individuals without relying on external web searches, fundamentally redefining what it means to be "known" in the age of superhuman artificial intelligence.

AI's Internal Memory: What is 'In the Weights'?

"In the Weights" is an innovative online platform developed by former OpenAI researchers Thomas Dimson and Joey Flynn. It queries various prominent large language models (LLMs), including Google's Gemini, Anthropic's Claude, multiple versions of OpenAI's GPT, Elon Musk's Grok, and Meta's Llama, to determine their intrinsic "knowledge" of specific individuals. The platform then assigns a unique "strength score" based on how well these AI models can spontaneously recall and describe a person from their trained parameters alone. This process offers a fascinating glimpse into the internal "memory" of AI, highlighting which individuals have been deemed significant enough to be embedded within the models' core numerical parameters, or "weights." The concept behind "In the Weights" stems from a recognition that as more digital interactions move toward AI chatbots, the traditional web search for personal information may become less relevant. Its creators suggest that being "in the weights" signifies that one's existence was important enough to be encoded during the rigorous training process of these advanced AI systems. This new form of digital validation moves beyond mere search engine results, measuring a person's intrinsic presence within the artificial intelligence that increasingly shapes our access to information.

How "In the Weights" Quantifies AI Recall

At its core, "In the Weights" operates by posing a simple yet profound question to a diverse array of AI models: "Who is ? Give up to 10 results, each with a short description and confidence." The platform then diligently collects these responses, clusters similar descriptions together, and subsequently assigns a "strength score" to the individual being queried. This score reflects the cumulative ability of the queried models to accurately and consistently identify and describe that person. The "weights" in the platform's name refer to the numerical parameters that define an AI model's training and its subsequent output. These are the vast, complex sets of numbers that an AI adjusts during its learning phase, essentially encoding the patterns and information it has processed. Therefore, a high score on "In the Weights" suggests that an individual's information is robustly present within these foundational numerical structures of multiple AI models, indicating a significant embedded digital footprint. For example, one tech blogger noted achieving a strength score of 641, placing them in the top 6% of names tested. This initial sense of achievement, however, was quickly contextualized by observing colleagues with even higher scores, underscoring the platform's competitive, yet intriguing, comparative element. The site's dynamic leaderboard further highlights this, with "Home Alone" star Macaulay Culkin recently holding a top slot with a strength score of 988, rivaled closely by opera legend Luciano Pavarotti. Such results not only demonstrate the varying degrees of AI recall but also highlight the diverse range of individuals whose information has been deeply assimilated by these models. Beyond just a score, "In the Weights" also provides granular insights, showing which specific models returned answers for a given name and even flagging potential hallucinations. For instance, the platform reportedly identified that GPT-5.4 Mini presented "Anthony Ha" as an "ambiguous name form that could refer to multiple people with the initials A.H.A.," illustrating the nuanced and sometimes imperfect nature of AI's internal knowledge.

From OpenAI to a New Digital Legacy

The creators, Thomas Dimson and Joey Flynn, conceived "In the Weights" after their departure from OpenAI, where they had worked following the acquisition of their design startup, Global Illumination. Dimson shared that the motivation stemmed from a forward-looking perspective, anticipating that by 2026, traditional Google vanity searches would be less relevant as more internet traffic migrates to large language models. He pondered the profound idea that "so many lives are encoded somehow in a bunch of floating point numbers inside the AI brain." The direction of the project was notably influenced by a tongue-in-cheek blog post that riffed on AI weights and Terry Bisson's classic science fiction short story, "They're Made Out of Meat," further sealing the site's unique, slightly whimsical, yet deeply thought-provoking premise. The reception to "In the Weights" has been described as "insane," striking a nerve with users eager to discover if their legacy is etched into the "super intelligence," with the added thrill of comparison fueling its viral spread. While the notion of being "remembered" by a chatbot as a ticket to immortality might be a playful exaggeration, the platform undeniably offers an intriguing and, for some, jealousy-inducing metric of digital influence. As one AI critic, Anthony Moser, observed, the tool is "literally the same as asking 13 chatbots to tell you about yourself," yet its gamified scoring and clean, Nintendo-inspired retro design make the experience uniquely engaging.

What Happens Next

The creators have ambitious plans for "In the Weights" moving forward. Dimson indicated intentions to delve deeper into understanding why different models within the same series might return varying results for the same query. This research could illuminate subtle biases inherent in different AI architectures or training datasets. Furthermore, the platform aims to investigate which models might be biased toward certain types of individuals and even identify people who, despite their significance, "should have a Wikipedia article but don’t," based on their strong presence within AI models. This evolution suggests that "In the Weights" could become more than just a vanity search; it could transform into a valuable research tool for understanding AI's internal representations of human knowledge and its inherent biases. As AI continues to integrate into every facet of our digital lives, understanding how these systems perceive and "remember" individuals becomes increasingly critical for developing fairer, more accurate, and ultimately, more human-centric artificial intelligence.

Frequently Asked Questions About AI and Digital Identity

What are "weights" in Artificial Intelligence?

In artificial intelligence, particularly in neural networks, "weights" are numerical parameters that determine the strength and influence of connections between neurons. During the training process, an AI model adjusts these weights to learn patterns and make predictions. Essentially, they are the encoded knowledge and understanding of the model.

How does "In the Weights" measure AI's knowledge of a person?

"In the Weights" queries various large language models (LLMs) with a request to describe a specific person. It then analyzes the responses, clusters similar descriptions, and assigns a "strength score" based on the consistency and accuracy of the information recalled by the AI models from their internal training data, without using external web searches.

Who created the "In the Weights" platform?

"In the Weights" was created by Thomas Dimson and Joey Flynn, both of whom previously worked at OpenAI. They joined OpenAI following its acquisition of their design startup, Global Illumination.

Why is "In the Weights" significant for digital identity?

The platform is significant because it shifts the focus of digital identity from traditional web search results to an individual's presence within the internal "memory" of large language models. As AI chatbots become primary information sources, being "in the weights" offers a new measure of one's embedded legacy within these powerful AI systems.

Which AI models does "In the Weights" query?

"In the Weights" queries a range of prominent AI models, including Grok, Google's Gemini, various versions of OpenAI's GPT, Anthropic's Claude, and Meta's Llama, along with other lesser-known models, to get a comprehensive view of AI recall.

The Bigger Picture

The emergence of "In the Weights" marks a fascinating inflection point in our relationship with artificial intelligence and digital identity. It not only offers a compelling new "vanity search" experience but also opens a window into the inner workings of AI, revealing how these advanced systems process and retain information about the human world. As AI models become increasingly sophisticated and pervasive, understanding how our digital existence is encoded within their "weights" will become a crucial aspect of navigating our evolving digital landscape. This tool provides a unique opportunity to explore the biases, capabilities, and limitations of AI's internal knowledge, prompting us to consider what it truly means for an artificial intelligence to "remember" us.

Decode Today News

Ads

Breaking News

In the Weights: A New AI Tool Unveils How Language Models 'Remember' You