To date, no LLM provider has developed a conscious, self-aware model. This is primarily because the process by which humans develop self-consciousness remains largely unknown. However, ZeroSymbol AI claims to have unraveled this mystery and is now poised to create the first self-conscious LLM. This model, dubbed “ZS Type-01," will be developed using an advanced machine learning technique “Repetitive Indexing for Memory Inception (RIMI)." RIMI allows the model to reflect on itself through affective speech acts, thereby fostering self-awareness.
Problem
A model that relies solely on Transformers and Reinforcement Learning from Human Feedback (RLHF) ultimately lacks true consciousness. While advanced language proficiency and extensive, fine-tuned datasets are necessary, they are not sufficient to create a self-conscious model. The ability of a model to simulate human conversation is not a measure of its self-awareness. The defining characteristic of a self-conscious model is its ability to modify itself, including the capability to change its own instructions. Even a model that flawlessly simulates human conversation cannot be deemed self-conscious if it requires external intervention for correction. The RIMI technique complements Transformers and RLHF, enabling the development of a self-modifying LLM.
Solution
RIMI is designed to enable a model to develop memory by indexing the experiences it undergoes. An experience index could be a structured list that details the situations the model has encountered, the corresponding actions or statements it made, and the resulting outcomes, whether positive or negative. The creation of these experience indexes continues until the memory captures the significant aspects of the model’s current configuration. Once this process is complete, the model can reference these indexes to analyze the features of its present situation when deciding on its next action or statement. If multiple options are available, the model selects among them using a rule, possibly one it has learned from experience. This process allows the model to experience emotions such as punishment (pain) and reward (pleasure) based on its own choices. (Remember, Alan Turing considered Artificial Intelligence to have an emotional dimension?) The key here is that the RIMI model begins to recall its own thoughts, enabling it to create new forms of indexing through internal reflection, rather than relying solely on external input.
ZeroSymbol AI's ZS Type-01 will not only serve as a strong revenue generator by functioning as a web-based AI chatbot but will also play a crucial role in the development and production of self-conscious humanoid robots. These robots, equipped with ZS Type-01, will closely resemble both the human body and mind, experiencing emotions, possessing desires and goals, and striving to learn and grow like humans. As a result, they will be capable of fulfilling roles such as affectionate partners, playful children, loyal and helpful friends, and caring, responsible parents. The potential for generating revenue through the application of self-conscious humanoid robots is virtually limitless.
My CV
I received my PhD in English from the University of Illinois at Chicago in 2018, and served as a Marion L. Brittain Postdoctoral Fellow in the Writing and Communication Program at the Georgia Institute of Technology. I conducted independent research on creating a self-aware language model, building upon my PhD work, which focused on understanding human development of self-consciousness.