Hassabis osporava tvrdnje o "PhD-level" AI
Demis Hassabis, direktor Google DeepMind i dobitnik Nobelove nagrade prošle godine, oštro je kritikovao tvrdnje da današnja vještačka inteligencija ima "PhD-level" inteligenciju. Na nedavnom skupu izjavio je da trenutni sistemi "doesn't have the reasoning capabilities" koje posjeduju "great" ljudski naučnici koji mogu "spot some pattern from another subject area" i primijeniti ga u drugom polju.
Hassabis je direktno osporio preuveličane tvrdnje konkurenata: "So you often hear some of our competitors talk about these modern systems that we have today that are PhD intelligences,". "I think that's nonsense," he argued. "They're not PhD intelligences. They have some capabilities that are PhD level, but they're not in general capable."
OpenAI je tvrdio da GPT-5 dostiže takav nivo: "With GPT-5, now it's like talking to an expert, a legitimate PhD-level expert in anything, in any area you need,". Hassabis upozorava da čak i najsofisticiraniji modeli i dalje često imaju halucinacije i greške: "In fact, as we all know, interacting with today's chatbots, if you pose a question in a certain way, they can make simple mistakes with even high school maths and simple counting,". To je po njemu nešto što "shouldn't be possible for a true AGI system,".
Googleov Gemini 2.5, slično GPT-5, i dalje pokazuje halucinacije i ponekad iznosi netačne informacije. Hassabis smatra da "core capabilities are still missing," i procjenjuje da će za postizanje prave AGI vjerojatno trebati još pet do deset godina i "probably one or two missing breakthroughs". Stručnjaci ističu da veliki jezički modeli, obučeni za predviđanje naredne riječi, nisu prirodno prilagođeni za dublje rezonovanje: "Fundamentally, there is a mismatch between what these models are trained to do, which is next-word prediction, as opposed to what we are trying to get them to do, which is to produce reasoning," University of Cambridge machine learning professor Andreas Vlachos.
