UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Марія БровінськаAI Eng
20 October 2025, 09:04
2025-10-20
OpenAI Co-Founder Andrey Karpaty: “The industry is living in a future where fully autonomous systems write all the code and humans become redundant. I don’t want to live in such a world.”
Even in a world where artificial intelligence is developing at lightning speed, it is worth being patient. In his opinion, full-fledged AI agents will not work correctly for at least 10 years. At least, that is what Andrey Karpaty, co-founder of OpenAI and one of the most famous researchers in the field of machine learning, director of artificial intelligence and computer vision for autopilot at Tesla, believes.
Even in a world where artificial intelligence is developing at lightning speed, it is worth being patient. In his opinion, full-fledged AI agents will not work correctly for at least 10 years. At least, that is what Andrey Karpaty, co-founder of OpenAI and one of the most famous researchers in the field of machine learning, director of artificial intelligence and computer vision for autopilot at Tesla, believes.
What are AI agents?
Agents are one of the hottest topics in the field of artificial intelligence. We are talking about virtual assistants that can independently perform tasks, break them down into substages, plan actions, and act without user prompts. Investors have already dubbed 2025 the «year of agents,» but Karpaty is convinced that they are still far from true autonomy.
In a post on the X platform (Twitter), Karpaty developed his opinion: «My criticism of the industry is that it overestimates its capabilities. The industry lives in a future where fully autonomous systems write all the code, and humans become unnecessary. I don’t want to live in such a world.»
He believes the future should be collaborative: humans and AI should work together. «I want an agent to be able to show me how it uses an API, make fewer assumptions, and collaborate with me when it’s not sure. I want to learn in the process and become a better programmer, not just be given mountains of code that ‘supposedly works.’»
Karpaty warns: if we create systems that completely replace people, the result will be «AI porridge» — a huge amount of low-quality content that will flood the Internet.
In a conversation on the Dwarkesh Podcast, Karpaty stated that fully functional AI agents will appear no earlier than in ten years.
«They just don’t work. They don’t have enough intelligence, they’re not multi-modal, they can’t use a computer, they don’t have continuous learning. You can’t just tell them something and have them remember it. They’re cognitively limited, and it doesn’t work yet,» explained Karpaty, who is currently developing the Eureka Labs project, a school built on the basis of AI.
Scientists created a fake company that employs only AI agents from Google, OpenAI, Anthropic, and Meta. The experiment showed which AI developments are the most effective, and whether there is a chance of a machine uprising
2025 is called the year of AI agents. Are AI agents really a must-have for everyone, and what AI assistants are the Ukrainian divisions of EPAM, Intetics, Levi9, P2H and Railsware working on?
