Charles Simonyi, a 77-year-old Technical Fellow at Microsoft and two-time space traveler, visited the Budapest University of Technology and Economics for its academic year opening and spoke to Portfolio in an exclusive interview about the current state, benefits and limits of artificial intelligence.
What he is working on and how he uses AI
Simonyi said he is working in a small team on a consumer-facing application that he expects will be widely used and solve a problem affecting many people. He still writes part of the code himself and described programming as "very pleasant" now because tools like Copilot assist him substantially. He answered "completely" when asked whether he actively uses AI for programming.
AI as a paradigm shift — and the role of precise intent
He believes AI constitutes a real technological and cognitive shift: "this will be a different world." At the same time, he stressed that human tasks will not disappear but transform. "If we tell the AI something and we are not satisfied with the result, then it is our fault, because we did not express our intent precisely enough," he said. According to him, expressing intent precisely requires effort and will be one of the main human challenges.
Simonyi noted a parallel with his earlier Intentional Programming concept: the idea that a computer could build machine code from an intention is now being realized in a different, practical way by large language models. He contrasted the unsuccessful theoretical "philosopher's stone" of that earlier idea with the present reality that AI actually works: you can tell the AI you need a BASIC interpreter that does certain things, and in moments get not only the computed result but the program that computes it.
How much code does AI write?
On claims that AI writes as much as 80% of code, Simonyi was skeptical and called such figures exaggerations. He emphasized that it matters how such percentages are measured. When working with existing systems, AI can help but cannot be allowed to do everything independently because it must adapt to what already exists.
Analogy to the early steam-engine era and the lack of theory
He compared the current state of AI theory to the period before Joule and thermodynamics, when engineers manipulated steam engines without a full theoretical understanding of energy. Similarly, he argued, AI needs a stronger theoretical foundation to understand why systems work and how to design them optimally.
Safety, slowing development, and regulation
While recognizing the importance of safety questions, Simonyi argued it would be premature to impose development restrictions now: first we should better understand how AI actually works. He warned against political maneuvers that use safety debates to harm competitors, saying competition should be about competing with rivals rather than using political means to damage them.
Regarding open source AI models, he said openness has produced benefits in other domains but is not a universal cure; the details matter for whether openness is advantageous.
On Elon Musk and technological integration
Simonyi praised Elon Musk as an engineer: "to me, Elon is a real engineer, almost the perfect engineer." He noted that linking different ventures (SpaceX, Tesla, Boring Company) is ultimately an ownership decision, but that many good things have emerged from activity across diverse fields.
AI in everyday life and the future
Simonyi said AI is already appearing in everyday devices: he gives the phone voice commands instead of tapping when driving. He believes AI will appear in more layers and directions across daily life. He also cautioned about a potential problem Elon Musk has described: if people’s desires are immediately fulfilled, they may struggle to express new wants accurately, leading to communication breakdowns with AI.
Long-term outlook
He expects the theoretical understanding of AI to be a long process, drawing a parallel with quantum mechanics, which underlies many technologies yet still leaves fundamental questions. Meanwhile, AI already delivers practical benefits, but the complete theoretical picture is not yet in place and will require ongoing research.
Short biography
Charles Simonyi was born in Budapest in 1948. He left Hungary in 1966, earned a doctorate at Stanford, and developed the first WYSIWYG word processor at Xerox PARC. From 1981 he led development of Word and Excel at Microsoft. He founded his own company in 2002, which Microsoft acquired in 2017; he has continued working as a developer there. In 2007 and 2009 he traveled to the International Space Station at his own expense and remains the only non-professional astronaut to have flown to space twice.
Conclusion
Simonyi sees AI not as a speculative bubble but as a fundamental technological change. He argues the immediate priorities are to deepen theoretical understanding of how AI systems work and to cultivate the human skill of expressing intent precisely, since that will define much of human-AI collaboration going forward.
