OpenAI i Sam Altman hvale GPT-5, ali test razotkriva slabosti
Nedavni eksperiment jednog penzionisanog profesora ekonomije sa Pomona College doveo je u pitanje tvrdnje OpenAI i Sam Altman da GPT-5 dostiže ljudski nivo razumevanja. Test sa "rotated tic-tac-toe" pokazao je da GPT-5 može postati zbunjen kada se igra na mreži okrenutoj za 90 stepeni.
Iako pravila ostaju ista — mreža je i dalje 3x3 i rotacija ne menja moguće pobede — GPT-5 je počeo davati nepotrebno složena objašnjenja: "Players are so used to the 'upright' tic-tac-toe board that a rotation might subtly change how they scan for threats and opportunities," i: "Mathematically, rotating the board 90° doesn’t change the set of possible wins — it’s still the same game in terms of outcomes. But psychologically, it could feel different."
Model je tvrdio da izbor centra "remains the strongest opening move" i spekulisao da igrači mogu "mis-evaluate edge vs. corner moves when their orientation is shifted," što nije podržano u igri. Na pitanje da li rotacija otežava ljudima igru, GPT-5 je napisao: "From a pure strategy standpoint, rotating the board doesn’t change anything. Every winning line (rows, columns, diagonals) just gets mapped to another winning line," i dodao: "for humans, the story is different."
GPT-5 je zatim ponudio da "actually draw rotated tic-tac-toe boards with position labels... so you can see how each transformation messes with recognition," ali generisane slike bile su nerazumljive i pune grešaka. Kritičar je zaključio: "They say that dogs tend to resemble their owners," i: "Chat GPT very much resembles Sam Altman — always confident, often wrong." Ovaj primer ponovo otvara debatu o pouzdanosti GPT-5 i tvrdnjama OpenAI i Sam Altman.
