Скачать или смотреть OpenAI's GDPval: AI Officially Reaches Human-Level Performance

For years, we’ve speculated about when AI would hit human-level performance. That speculation is over 👇

OpenAI's new benchmark, GDPval, measures model performance on 1,320 real-world, economically valuable tasks across 44 occupations in the US.

The results are a wake-up call:

➡️ 𝐄𝐱𝐩𝐞𝐫𝐭 𝐏𝐚𝐫𝐢𝐭𝐲: Today's best models are already as good as or better than industry experts in about 48% of tasks. This suggests that AI is already outperforming 50 to 70% of your workforce.

➡️ 𝐉𝐮𝐧𝐢𝐨𝐫 𝐒𝐭𝐚𝐟𝐟 𝐋𝐚𝐠: If models are near parity with experts, they are likely outperforming 70-80% of junior staff (analysts, entry-level lawyers, marketing associates) on well-defined tasks.

➡️ 𝐇𝐢𝐫𝐢𝐧𝐠 𝐒𝐡𝐢𝐟𝐭: This isn't a future concern. Recent Harvard and Stanford research shows early-career hiring in AI-exposed sectors is falling, while demand for experienced workers rises.

The acceleration is staggering: performance on these tasks has more than tripled in just one year (GPT-4o to GPT-5), and models are completing the work 100x faster and 100x cheaper than humans.

In my view, GDPval is one of the most important benchmarks for tracking the impact of AI on the workforce. It shows us where AI can handle the routine, freeing human workers to focus on more complex tasks.

👉 𝐐𝐮𝐞𝐬𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐄𝐱𝐞𝐜𝐮𝐭𝐢𝐯𝐞𝐬: How does your company need to change its training strategy to ensure your team is AI literate?

👉 𝐐𝐮𝐞𝐬𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐢𝐨𝐧𝐚𝐥𝐬: What's the biggest change you're making to ensure you stay relevant in the age of AI?

#ai #futureofwork #skills #workforceplanning #gdpval