goats

Can an artificial intelligence model “feel” and be “self-aware”? It seems like an absurd question, but there are people who are beginning to think that this is the case. Technology is becoming more compelling and precise, and conversations with many models can raise questions for those who use them. And to disprove them all, someone has created a goat in ‘Age of Empires II‘. Believe us: the idea makes sense. The men who believed that AI was conscious. Richard Dawkins is not just any scientist. This biologist and zoologist is one of the most renowned experts in his field. Precisely for that reason his statements of May 2026 They surprised everyone and everyone: after having multiple conversations with Claude, he came to say: “You may not know that you are aware (of yourself), but I already believe that you are.” An old debate. Many criticized those statements, but the message was not in fact new. In June 2022, months before ChatGPT made its appearance, Google engineer Blake Lemoine stated that already at that time chatbots were developing their own consciousness. Google suspended him from his position. Anthropomorphism in the 21st century. A Microsoft researcher named Adrian de Wynter collaborated with New York University to try to answer that question. The result was a study with a promising title: ‘If LLMs have human attributes, so does Age of Empires II.’ De Wynter argues in it that the conversational capabilities of chatbots have caused that people anthropomorphize them: ChatGPT, Gemini or Claude are no longer machines: they are almost like people. No? No. The goat from Age of Empires II. In his study, this researcher wanted to demonstrate this phenomenon with a unique analogy. He showed how he had built an Age of Empires II map editor to create NAND gates using goats. As I explained, “The goal of the paper is to formally demonstrate that we tend to anthropomorphize too easily and that sometimes the claims we make about the capabilities of large-scale language models (LLMs) are too strong. This is not an easy task, given that the concept of ‘human-like attributes’ is a somewhat abstract term.” A super basic and very revealing LLM. The Age of Empires II map or scenario editor has an isolated ‘sandbox’ mode in which players can create their own maps and objectives taking advantage of the video game’s digital resources. De Wynter used these resources to create NAND logic gates in the game, so that in this “raw LLM” the grass was a 0, the jumpers were a 1, and the goats were the bits. Games to create simple LLMs. The operation, explained on their GitHub pagewas unique. De Wynter took advantage of the concept of the perceptron, the simplest neural network that exists: an algorithm that sorts an input into binary classes. People, as explained in 404 Media, have been taking advantage of the idea for some time. in other games like Minecraftso de Wynter came up with the idea of ​​taking advantage of the concept in Age of Empires II to try to answer the question of whether AI can be conscious. AI inside. No matter how complex ChatGPT or Claude may seem, behind it there is nothing more than a gigantic network of mathematical operations based on logic gates and perceptrons. These operations are carried out on chips like those from Nvidia, but Wynter changed them for the map of a video game. He was able to replicate the structure of a basic AI in the scenario editor and revealed something important: If you interact with an AI through a chatbot and the machine responds to you empathically, you tend to humanize it and think that it is conscious because the interface imitates a human conversation. But if you remove the chatbot and put that same neural network to work in a game of Age of Empires II, the only thing you see on your monitor is a bunch of goats moving in a virtual meadow. Conclusion. For the Microsoft engineer the conclusion is clear: the underlying software is the same in both cases. If the illusion of consciousness disappears when we replace ChatGPT’s conversational interface with virtual goats, the supposed consciousness is not in the system: it is something we assign to the persuasive interface of ChatGPT, Gemini or Claude. The AI ​​doesn’t understand you; just pretend it. An essay by Ted Chiang published in The Atlantic At the beginning of June 2026 he also denied the idea that AI could be conscious. The author made his conclusions clear: “The only reason a large language model (LLM) generates phrases like ‘I understand’ is to make it more attractive than a search engine and increase the likelihood that the user will return; that is, it is another way to maximize customer engagement. This benefits the company marketing the LLM, but not the users.” In Xataka | As far as we know, the agency that supervises AI in Spain is not supervising anything. What it does have is an Ideas Laboratory