Operating costs of OpenAI's GPT-3: Do they impact the future of artificial intelligence?

It seems that running the new AI model O3 from OpenAI will be more costly than expected, as recent reviews have shown costs could be ten times higher than initial estimates, raising questions about its feasibility for widespread use. Upon its launch in December, OpenAI introduced model O3 as a groundbreaking advancement in the field of artificial intelligence, especially after testing it in collaboration with ARC-AGI developers. However, a report published by "TechCrunch" revealed that running the more powerful version O3 High could cost $30,000 per task, compared to the initial estimate that did not exceed $3,000. Mike Knob, co-founder of the ARC Prize Foundation, explained that these costs make O3 High almost as expensive as running O1-Pro, which is currently the most expensive OpenAI model. While the company has not yet announced the official pricing for O3, the available figures indicate a massive consumption of computing resources, using 172 times the resources compared to the O3 Low version, which may limit its use outside of large companies. These high costs align with previous reports indicating that OpenAI is considering charging $20,000 monthly for some AI services, such as "specialized AI agents." Despite these steep prices, some analysts believe that using O3 may still be more cost-effective than hiring humans, but it still faces efficiency challenges, as tests have shown it needed 1024 attempts to solve each task in the ARC-AGI test, reflecting the ongoing gap between artificial intelligence and human capabilities.