Musk claimed in 2020 that GPT-5 or 6 might be indistinguishable from humans
Elon Musk wrote on August 14, 2020, that the improvement speed from the initial GPT to GPT-3 is impressive. If this speed continues, GPT-5 or GPT-6 might be indistinguishable from the smartest humans. He immediately clarified that this is just a personal opinion, not an endorsement.
In the same post, he positioned himself as a neutral outsider. His exact words were that he left OpenAI two to three years ago and is no longer on the inside. Public records show he resigned from the OpenAI board in February 2018, reportedly due to differences in company direction and potential conflicts of interest with Tesla's AI business. OpenAI was co-founded by him in 2015 with his funding. At the time of the post, Sam Altman and Greg Brockman were already managing an organization different from the one he left.
The timing was right after GPT-3 had a limited release. At that time, OpenAI had not yet established the subsequent naming rhythm. The gap between GPT-3 and GPT-4 was about two and a half years. He did not provide testing standards or specify whether "indistinguishable" referred to conversation, exams, or work. The sentence only set one condition: the improvement speed remains unchanged.
This statement did not involve funding or price predictions. By 2026, OpenAI had released GPT-6 Astra. Musk later founded xAI and repeatedly referred to Grok 5 as general artificial intelligence or indistinguishable, without providing actionable thresholds. The phrase "neutral outsider" did not continue to reflect his later position.
The post was a reply, not a press release. The focus was on the generational gap of OpenAI, with the speaker emphasizing that he had left. The event was driven by the limited visibility of GPT-3's capabilities. The beneficiaries were readers who extrapolated expectations based on speed. The pressure was on OpenAI, which he excluded from endorsement, and anyone who took a single opinion as a roadmap.
This is not a transaction. There were no buyers, sellers, or deliveries. No funds flowed because of this statement. It was later repeatedly quoted because the naming coincidentally reached GPT-5 and GPT-6. The quote does not change the two limitations in the original sentence: it is just an opinion, and the speed must continue.
Source: Public information
ABAB AI Insight
The split between Musk and OpenAI had already occurred before the post. He co-founded and funded it in 2015 and left the board in February 2018. In August 2020, he portrayed himself as an outsider to judge the speed of GPT-3 while refusing to endorse it. This position was not maintained later. After founding xAI, he began comparing his Grok numbers with similar claims.
No capital followed this post. What was mobilized was expectation. Musk set GPT-5 or 6 as potential points that might not show differences from the smartest humans, but there were no tests or score thresholds. The motivation was to extrapolate from the leap between two generations rather than to announce internal routes. At that time, OpenAI's naming had not yet fixed on 5 and 6, yet the sentence preemptively wrote these two numbers as endpoints.
Comparable is Kurzweil's extrapolation of human levels using computational curves, and DeepMind's claim of general learning after AlphaGo. The difference is that Musk was both a former funder and a later competitor. The industry position in 2020 was still in the demonstration phase. GPT-3 was only open to a limited number of testers, and the product intervals later proved not every generation advanced by the same number of months.
Structurally, it is a transfer of pricing power. The narrative of capability pricing shifted from laboratory announcements to an outsider's extrapolation. The mechanism is that "if speed continues" leaves the responsibility in the future, and the speaker only uses opinion to disclaim liability. By the time GPT-6 Astra was released, indistinguishability still lacked a common definition. The reason the term neutral became ineffective is very specific: he later created his own model and evaluated it with similar sentences.
ABAB News · Cognitive Laws
The premise for extrapolation to hold is that speed must not change.
Those who claim neutrality may later enter the field themselves.
Indistinguishability without tests is just an opinion.