Recently, I came across an interesting story about one of the leading architects of modern AI. Andrey Karpaty — a Slovak with a Canadian passport who has gone from academic researcher to a key figure in the artificial intelligence industry. His career reads like a textbook on how to influence technological development.

It all started in Bratislava in 1986. At 15, his family moved to Canada, where he studied computer science, physics, and mathematics. He studied seriously — in 2015, he defended his dissertation on neural networks and computer vision. Later, he taught a course on convolutional neural networks at Stanford, which became very popular among students.

Interestingly, Karpaty also interned at Google, and in 2015, he became a co-founder of OpenAI. But he didn’t stay there long. He moved to Tesla, where he led the development of autopilot for six years. In 2022, he left Musk’s company and then returned to OpenAI (although for a short period). Currently, he runs a YouTube channel about neural network creation and founded Eureka Labs, a startup focused on integrating AI into education.

Regarding patents, Karpaty holds six. Five belong to Tesla — these are developments related to predicting 3D characteristics for autonomous driving, creating data for machine learning, and annotating videos with deep neural networks. One patent (the first) is registered to Google. It turns out that his former employers will continue to receive dividends from his developments for a long time, while Karpaty himself has not yet registered anything in his name.

His article "Software 2.0" has become quite well-known in professional circles. In it, he discusses how neural networks will enable the creation of software that humans won’t be able to understand. He even introduced a new term — vibecoding (vibecoding) — to describe development with the help of AI assistants. Karpaty himself admitted that he feels like a lagging programmer for the first time, comparing AI tools to “powerful alien technology without an instruction manual.”

According to him, the development of neural networks is progressing much faster than expected. While the initial prediction was the emergence of True Artificial Intelligence by 2043, after the release of GPT-4, the timeline shifted to 2028. Karpaty urges schools to restructure the educational process to match AI realities because it’s simply impossible to detect its use in homework.

It looks like humanity is really on the verge of a new technological revolution. And people like Andrey Karpaty play far from the least role in this process. His patents, research, and videos are already shaping how we will interact with AI in the coming years.