Is AI panic valid? Hungary’s AI commissioner answers the question
Artificial intelligence is unlikely to become self-aware in the current structure, László Palkovics, the government commissioner for AI, said in The Bold Truth About Hungary podcast on Thursday.
Palkovics told Zoltán Kovács, the state secretary for international communication and relations, that self-aware AI could not be created “by accident”, only deliberately. “But why would humanity want an enemy for itself?” the commissioner said. “We have plenty as it is, so what we’re looking for is tools or things that aid us, and those can’t become self-aware.”
“So I think these fears are interesting and lend themselves to good discussions … less so with mathematicians and engineers, but mainly with philosophers or religious scholars,” Palkovics said.
He said the kinds of models that could be developed and how fast they could be trained depended on data, computing capacity, a clear legal and financial environment as well as “the intent and talent that outline the possibilities” for the use of AI. Hungary, he added, met all of those requirements.
Palkovics noted Hungary’s Komondor supercomputer with a computing capacity of 5 petaflops, and that a public procurement procedure will soon be launched for a 20 petaflop supercomputer dubbed Levente.
He said Hungary was in a good position when it came to computing and storage capacity, and the country’s energy mix covered the energy needs of a supercomputer.
Asked if he thought Hungary should develop its own large language model (LLM), Palkovics said such an AI would not be competitive against international models as an LLM, but would be competitive in a variety of other areas.
He said what mattered was the data such a model was trained on, explaining that Hungary was fortunate because it had structured data on agriculture, forestry, game management and the database of the Central Statistical Office.
Read more AI-related news on Daily News Hungary.
Read also:
- Hungarian motorways enter the digital age with AI-powered monitoring system
- Artificial intelligence could unleash nuclear weapons on us, researchers warn