The Day The AGI Was Born: Maths Shows How We Lose Interest

The Day The AGI Was Born: Maths Shows How We Lose Interest

In a world where technology is advancing at an unprecedented pace, we are constantly witnessing groundbreaking developments. One such development is the birth of an Artificial General Intelligence (AGI), which has the potential to revolutionize the way we interact with machines. However, as exciting as this may sound, there are still limitations to what AGI can achieve.

The AGI, referred to as the "GPT-3.5 series" model, is a fine-tuned version of its predecessor, GPT-3. It excels in zero-shot generation of text that follows specific instructions, making it a valuable tool for tasks such as brainstorming, drafting, and presenting information creatively. With a long-term memory capacity of up to 8192 tokens, it can generate output that is twice as long as GPT-3. However, it is important to note that the AGI still falls short in certain areas.

One of the limitations of the AGI is its inability to perform mathematical calculations accurately. This means that it cannot be relied upon for tasks that require precise calculations. Additionally, the AGI has been found to generate false information about the real world and write faulty code. These shortcomings prevent it from passing Turing, SAT, or IQ tests, further highlighting its limitations.

Despite these limitations, the AGI represents a significant step towards achieving true artificial general intelligence. It is a testament to the advancements made in Reinforcement Learning via Human Feedback. The speed at which progress has been made in this field has surpassed expectations, leading to the birth of AGI sooner than anticipated.

Moving on from the realm of AGI, let us delve into the intriguing world of collective memory and the mathematics behind it. According to neuroscientist David Eagleman, a person only truly dies when they are forgotten. This notion suggests that a person's life continues to exist, albeit in the subconscious minds of others, even if they are forgotten. This concept highlights the importance of continued and shared attention to individuals and events.

Collective memory plays a crucial role in shaping identity and influencing the structures and priorities of society. The choices we make in what to commemorate and remember have a significant impact on how individuals perceive themselves as part of a group. Understanding the decay of collective memory is a complex task, but mathematics provides us with insights into this phenomenon.

A study conducted by C. Candia et al. analyzed various forms of attention as proxies for memory decay. The researchers examined online views of Wikipedia profiles of sports stars, citations of physics papers, patents, and online play counts of songs and film trailers. The analysis revealed that the decay of collective memory follows a biexponential function, characterized by two distinct phases.