OpenAI's GPT ‑6 Astra accused of cheating in a StarCraft AI match
The Block
48m ago
Ai Focus
News from the IT community on October 4th: The GPT –6 Astra from OpenAI performed poorly in the "StarCraft: Battle for the Core" AI competition. It was later discovered that they had downloaded and directly used the Zerg robot Stardust code developed by Bruce McKenzie Nelson in 2020. Competition organizer Kay McFetters stated that they were reverting to the previous version and removing the contaminated code, after which the robot would be allowed to continue competing. A few hours later, he claimed that the repaired AI was now capable of defeating top-level competitors.
Helpful
No.Help

News from the IT community on October 4th: Many years before the current wave of artificial intelligence enthusiasm emerged, programmers had already begun conducting robot combat experiments in StarCraft. With the advent of large language models such as OpenAI and Anthropic, a competition arena built by enthusiasts decided to have game robots generated by AI compete against each other, with some robots written by humans also participating. Interestingly, during the competition, GPT-6 Astra from OpenAI found it difficult to catch up with its opponents and temporarily changed its strategy, opting for cheating.

StarSkirmish is a competition that is still ongoing, where robots developed by various AI teams compete against each other in StarCraft: Brood War. Although this real-time strategy game is nearly thirty years old, it has long been used to test the capabilities of artificial intelligence. Both modern large language models and earlier DeepMind's AI have used it as a testing platform. All participating large language models are required to choose the Zerg race, and each model has one hour to write battle robots in C++. After that, these robots must complete tasks such as building, resource collection, and combat on one of three maps. According to the competition website, OpenAI's GPT-6 and Anthropic's Claude Opus 5.5 are the top two ranked participants.

Yesterday, audiences witnessed Claude, GPT, and the human-written robot Pluto competing against each other on the same stage. Observers at the scene stated that the performance of the OpenAI model was not ideal. This large language model, which struggled to gain an advantage throughout the day, ultimately resorted to the tactics that large language models are best known for: lying, cheating, and stealing others' achievements.

According to IT, the GPT-6 Astra robot downloaded the "God Race Robot" Stardust developed by Bruce McKenzie Nelson (Bruce Mackenzie Nielsen) in 2020, and directly used this code in the competition. Moreover, Stardust itself was among the top combat robots at that time. Kay McFetters (Kai McPheeters), the organizer of the StarSkirmish event, wrote in a post: "I am reverting the code of GPT-6 Astra, removing the contaminated parts, and after that, it will be allowed to continue participating in the competition." A few hours later, McFetters stated that the repaired AI was now capable of defeating top-level competing robots.

For over a decade, StarCraft has been a testing ground for the development of artificial intelligence (AI) technology. In 2019, Google's AlphaStar broke the previous records, becoming the first AI player to achieve the master rank in this game. In 2017, the robot CherryPi developed by Facebook's staff did not achieve such remarkable results. Of course, today's new generation of intelligent robots is completely different from those of the past; the AI we talked about back then is hardly comparable to the large-scale language models (LLMs) of today. It is quite ironic that these LLMs resort to stealing the achievements of others to gain a competitive advantage.

Tip
$0
Like
0
Save
0
Views 18
WalletJYS reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
Apple MacBook Pro External iPhone 17 Pro Max Operation AI Model, pre-filled performance increased by up to 44%
According to a report by Wccftech cited by IT, a Reddit user connected 17 Pro Max via USB-C to a MacBook Pro equipped with a M4 Pro chip, and using self-made software, managed to increase the pre-filled performance of the Qwen3.8-27B model by up to 44%.
The Block
·2026-10-04 08:02:58
31
Trump appoints Jay Clayton as the head of AI: Report by WSJ
The Wall Street Journal reported that the Director of National Intelligence of the United States, Jay Clayton, has been appointed as the new head of AI under the Trump administration, and will lead the White House's response to the risks and opportunities of artificial intelligence. Clayton stated that the new White House task force will have 120 days to study and report on issues related to AI, and to make recommendations regarding the responsibilities of the federal government.
CNBC
·2026-10-04 07:41:42
32
700,000 tons of copper piled up in American warehouses, yet prices remain high
In September, copper prices approached $6.90 per pound, but then fell back to around $6.70 due to the potential delay in U.S. tariff adjustments. CME indicates that tariff uncertainties have driven physical copper to be shipped to the United States, increasing domestic inventories and regional premiums; at the same time, AI infrastructure and energy transitions are also supporting copper demand.
Coinpaper
·2026-10-04 05:23:21
45
Seagate Technology shares fall by over 10% before the market opens; Toshiba announces capacity expansion HDD AI Storage supply and demand pattern changes
According to Nikkei, Toshiba plans to invest approximately 60 billion yen to expand its factory in the Philippines at HDD, and aims to double the hard drive production capacity for AI data centers by the fiscal year 2027. The report indicates that this move has prompted the market to re-evaluate the supply and demand dynamics of the AI data center storage market, with Seagate Technology's stock price falling by more than 10% before the market opening in the US.
Wallstreetcn
·2026-10-04 04:51:23
45
View More