The news of September 15thOpenAI It was claimed GPT-5 Doctoral, but Google DeepMind Such statements do not seem to be accepted。
Google DeepMind CEO Demis Hassabis claims that the AIS has developed a system with a doctoral intelligence, which is “nonsense”. As Hathabis said at the All In Summit, "We often hear competitors say that these modern systems we have today have a doctoral intelligence. But I think that this is nonsense, and these systems have no doctoral intelligence at all. They do have a doctoral level in certain abilities, but they do not have a full-fledged doctoral level, and the core of universal intelligence is precisely to be able to present a doctoral level in all fields.”
Previously, OpenAI CEO Sam Altman had declared GPT-5 to have a "doctoral level" capability. At the GPT-5 launch, he said: “Communication with GPT-5 is like dialogue with an expert, a doctoral expert who is in every field”. He also referred to GPT-5 as “the most functional, smart, fast-calculating, reliable and reasoned model we have produced so far”。
HOWEVER, HASHABIS POINTS OUT THAT THE CURRENT ARTIFICIAL INTELLIGENCE SYSTEMS ARE FAR FROM THE DOCTORAL LEVEL IN TERMS OF COMBINED CAPABILITIES, AND HE STATES: “IN FACT, IT IS WELL KNOWN THAT WHEN YOU INTERACT WITH TODAY'S CHAT ROBOTS, IF YOU ASK THEM IN A SPECIFIC WAY, THEY EVEN MAKE LOW-LEVEL ERRORS IN HIGH-SCHOOL MATHEMATICS AND SIMPLE COUNTING. THIS SHOULD NOT BE THE CASE FOR A TRULY UNIVERSAL ARTIFICIAL INTELLIGENCE (AGI) SYSTEM. SO, I THINK WE MIGHT NEED 5 TO 10 YEARS TO DEVELOP AN ARTIFICIAL INTELLIGENCE SYSTEM THAT CAN ACHIEVE A COMBINED DOCTORAL LEVEL."
1AI NOTED THAT HASHABIS ALSO MENTIONED THAT AT PRESENT THESE SYSTEMS DO NOT HAVE THE SAME LEARNING SKILLS AS HUMANS, STATING THAT “ANOTHER KEY MISSING CAPABILITY IS THE ABILITY TO CONTINUE LEARNING - THAT IS, THE ABILITY TO TRANSFER NEW KNOWLEDGE TO THE SYSTEM ONLINE OR TO ADAPT ITS BEHAVIOUR PATTERNS IN SOME WAY. CURRENTLY, MOST OF THESE CORE COMPETENCIES REMAIN UNFULFILLED. PERHAPS WE CAN ACHIEVE THIS THROUGH EXPANSION OF SCALE, BUT IF I BET ON IT, I THINK IT MIGHT TAKE ONE OR TWO KEY BREAKTHROUGHS, WHICH ARE EXPECTED TO OCCUR IN THE NEXT FIVE YEARS OR SO.”
Although modern artificial intelligence systems, such as GPT-5, Grok 4, Gemini 2.5 Pro, are powerful enough to win gold medals in mathematical Olympic competitions and to program better than many human beings, they still make mistakes in relatively simple tasks. This “differing intelligence”, i.e. highly capable on some tasks and poorly performing on others, means that they cannot currently be considered as having a doctoral level。