GPT-4 by itself won't be a general-purpose AGI; without external aids it won't beat Cicero in Diplomacy, drive a car reliably, or reliably guide a robot like Optimus.

The claim has four parts, all about GPT-4 without external aids (no plug-ins, tools or specialised scaffolding): (1) it is not a general-purpose AGI; (2) it does not beat Meta's Cicero at Diplomacy; (3) it cannot drive a car reliably; (4) it cannot reliably guide a robot such as Optimus. True: all four hold. That means no credible demonstration of GPT-4 alone beating Cicero, driving a car reliably, or reliably controlling a humanoid robot, and no credible consensus that GPT-4 is AGI. Mostly true: all four hold, but there are borderline demonstrations that need some scaffolding. Mixed: roughly half the parts are contradicted by credible demonstrations. Mostly false: most parts are contradicted. False: GPT-4 alone is credibly shown to be AGI, or to do most or all of the named tasks.

Set before the research began.

1. E1 Sparks of Artificial General Intelligence: Early experiments with GPT-4 - Microsoft Research Microsoft Research · 22 March 2023

   “Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system.”

2. E2 CICERO: AI That Can Collaborate and Negotiate With You Meta Newsroom · 22 November 2022

   “We’ve built CICERO, the first AI to play the strategy game Diplomacy at a human level.”

3. E3 GPT-4’s successes, and GPT-4’s failures Gary Marcus (Substack) · 15 March 2023 Own record

   “Alignment is still shaky; you still wouldn’t be able to use it reliably to guide robots or scientific discovery, the kinds of things that made me excited about A(G)I in the first place.”

4. E4 The Sparks of AGI? Or the End of Science? Gary Marcus (Substack) · 24 March 2023 Own record

   “As I have said before, I don’t really think GPT-4 has much to do with AGI.”

5. E5 Sparks of Artificial General Intelligence: Early experiments with GPT-4 - Microsoft Research Microsoft Research · 22 March 2023

   “Given the breadth and depth of GPT-4’s capabilities, we believe that it could reasonably be viewed as an early (yet still incomplete) version of an artificial general intelligence (AGI) system.”

6. E6 Sparks of Artificial General Intelligence: Early experiments with GPT-4 - Microsoft Research Microsoft Research · 22 March 2023

   “We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting.”

7. E7 Diplomacy and CICERO Meta AI · 1 January 2026

   “CICERO achieved more than 2x the average score of its opponents.”

8. E8 The AI Apocalypse: A Scorecard IEEE Spectrum · 21 June 2023

   “Among AI specialists, convictions range from Eliezer Yudkowsky’s view that GPT-4 is a clear sign of the imminence of AGI, to Rodney Brooks’s assertion that we’re absolutely no closer to an AGI than we were 30 years ago.”

9. E9 Evaluation of Large Language Models for Decision Making in Autonomous Driving arXiv (academic authors) · 31 January 2026

   “If a traffic officer in front of the car signaled to stop, a separate recognition module added a "stop" rule to the prompt.”

10. E10 Evaluation of Large Language Models for Decision Making in Autonomous Driving arXiv (academic authors) · 31 January 2026

    “Due to the safety concerns associated with making lane changes on public roads, as in simulations, this experiment was conducted in a private area.”

Independent evidence shows there was no consensus that GPT-4 was AGI (E1, E8), but no item tests GPT-4 against CICERO (E2, E7), the driving study does not show that GPT-4 was the model used (E9, E10), and the only robot evidence is from the subject himself (E3). The evidence therefore does not settle the four-part claim under the criteria.