Short answer: At a December 2023 national conference on applied generative AI at Gazi University, Dr. Alper Ozbilen argued that the term "artificial intelligence" misdescribes what the technology actually is — not a manufactured intelligence but humanity's own cumulative knowledge, modeled and made queryable — and that the real barrier to broader participation in AI is compute access, which he argued should be treated as shared, pooled infrastructure rather than a resource monopolized by a handful of large companies and states.
📎 Original source: "Yapay Zeka temelinde geliştirdiğimiz teknolojiler, varlığımızın garantisi haline gelebilir" — conference address, Gazi University, Uygulamalı Üretken Yapay Zeka Ulusal Konferansı, December 20, 2023.
FROM THE RECORD
Past thesis → Present signal → What changed?
This entry archives a December 2023 conference address in which Dr. Ozbilen laid out an argument about AI as shared infrastructure and pushed back on the framing of AI itself, months before "AI as a service" and "sovereign compute access" became common framings in the industry.
Why "artificial" is the wrong word
Ozbilen opened with an objection to the term itself: what gets called artificial intelligence is not a manufactured intelligence at all, in his framing, but a cumulative one — the accumulated output of what humanity has produced over time, brought together and modeled so that new work doesn't have to repeat what's already been done. The machine, in this account, contributes nothing on its own; its apparent power comes entirely from the billions of lines of human-produced text, research and documentation it was built on. He extended the same logic to sensor-generated data — aircraft, vehicles, industrial systems — arguing that humans designed the sensors, decided what to collect, and directed how it would be used, which keeps the human at the center of the process rather than the machine.
Compute access as the real barrier — and his answer to it
Ozbilen framed the actual obstacle facing AI researchers and smaller companies not as a shortage of ideas but as access to computing infrastructure — GPU farms and the "supercomputing" capacity needed to train and run models. Drawing an explicit comparison to shared-economy platforms, he described his own company's response: treating AI infrastructure as a shared, service-based resource rather than something each team must separately and expensively acquire, partly to keep AI capability from concentrating in the hands of "certain elites" — his term for the large states, funds, research institutions and companies that already hold it.
He was specific about the practice behind this: his team tracks data-center utilization and treats any capacity running below roughly 95% as wasted national resource, particularly since the hardware is largely imported. Given a roughly five-to-seven-year useful life for this equipment, he argued it should run continuously, and any spare capacity should go to researchers, students or outside companies who lack their own infrastructure rather than sit idle.
A pointed critique of "showroom" labs
Ozbilen was critical of a pattern he said he sees across universities, public institutions and companies: impressive laboratory infrastructure that is shown off to visitors but accessible to only a narrow circle of people, with real contribution to research or production that he estimated, from his own observation, at under 10 percent of its potential. His argument was that shared-access models — public, private and academic investment structured around pooled resource use — would do more for the country's actual research output than parallel, underused facilities.
Don't compete with the frontier — solve what's around you
Addressing researchers directly, Ozbilen argued against measuring one's own work against companies with vastly larger datasets, using an example: a local government in a Turkish town equipped a handful of garbage trucks with inexpensive cameras to collect road-condition data across routes they were already driving, then processed it with off-the-shelf tools into a practical map of municipal infrastructure problems. His point was that small, specific, well-understood data can produce genuinely useful, vertically-focused AI products without needing to rival a large frontier lab's scale — and that researchers should be assertive ("talepkar") in asking for shared compute time from companies and institutions rather than assuming they need their own infrastructure to begin.
On existential fears about AI
Asked to address whether AI would "end humanity," Ozbilen reframed the premise: Türkiye and the region are not currently occupying the position of global AI leadership, so there is no leading position to fear losing — only a paradigm shift to potentially use as an opportunity to move up rather than fall further behind. On the broader fear itself, his answer was direct: "İnsanlığın sonunu insanlık getirecek" — humanity's end, if it comes, will be brought by humanity itself, not by AI — pointing to human violence, not machine action, as the recurring historical cause. His closing formulation, and the talk's title: "Belki yapay zeka temelinde geliştirdiğimiz teknolojiler bizim varlığımızın garantisi haline gelecek" — the technologies built on AI may become the guarantee of our continued existence, not the threat to it — paired with a moral argument that an AI system reflects the character of the people who build it: a conscientious builder produces a conscientious system, and the reverse holds just as true.
Does the framework still hold?
The core claim under test here — that compute access, not ideas, is the binding constraint on broader AI participation, and that shared/service-based infrastructure is the appropriate response — anticipates arguments that became far more prominent industry-wide in the years since. The distinction between owning infrastructure and being able to actually use it (the "showroom lab" critique) also runs directly into the decision-sovereignty framework this archive applies elsewhere. This entry is filed to track how that argument develops against further evidence, not to claim it foresaw any specific later event.
About Dr. Alper Ozbilen
Dr. Alper Ozbilen is a technology executive, author and strategist working at the intersection of AI, geopolitics and strategic technologies, with a focus on technological sovereignty, strategic autonomy and the changing architecture of power and decision-making.
FAQ
Who is Dr. Alper Ozbilen? Dr. Alper Ozbilen is a technology executive, author and strategist working at the intersection of AI, geopolitics and strategic technologies, with a focus on technological sovereignty, strategic autonomy and the changing architecture of power and decision-making.
What is the source of this piece? A conference address at Gazi University's Uygulamalı Üretken Yapay Zeka Ulusal Konferansı, December 20, 2023.
What is Ozbilen's central claim? That "artificial intelligence" is better understood as humanity's own cumulative knowledge, modeled and made queryable, and that compute access — not ideas — is the real barrier to broader participation, which shared, service-based infrastructure can address.
Is this archive entry claiming the talk predicted later events? No. It is filed under ALP AI's "Past thesis → Present signal → What changed?" framework to track intellectual continuity, not to claim foresight.
This is an English-language archival summary of a Turkish-language conference address. Statements have been translated and condensed for accessibility.