According to PANews, OpenAI CEO Sam Altman recently shared insights on the emerging influence of Artificial General Intelligence (AGI), suggesting its impact could surpass previous technological transformations. Altman highlighted three key observations regarding the economics of AI:
1. The intelligence of AI models is logarithmically related to computational resources, with investments in training computation, data, and inference computation yielding continuous and predictable returns.
2. The cost of using AI decreases approximately tenfold each year. For instance, the cost per word of GPT-4 at the beginning of 2023 was about 150 times higher than that of GPT-4o in mid-2024, significantly outpacing Moore's Law.
3. Linear growth in intelligence results in a super-exponential socio-economic value, driving increased investment that is unlikely to cease in the short term.
Altman anticipates that AI agents, such as software engineering assistants, will become "virtual colleagues" and permeate various industries. He compares the economic impact of AI to that of transistors, noting that while the changes may be subtle, they are profoundly transformative. The widespread adoption of AI could lead to a sharp decline in the cost of certain goods, while the prices of scarce resources like land may rise.
He underscores the importance of policy and societal consensus for the future of AGI. To ensure the benefits of AGI are widely distributed, new economic models might be necessary, such as providing a "computational budget" to every individual globally. Altman envisions that by 2035, everyone should be able to harness intellectual resources equivalent to the global intelligence of 2025, significantly unleashing human creativity.