bodyType

customContents

entityId

entityUuid

articlePageQuery

filter

article

articleBaseAdvertisingInfoArticle

articleListArticle

articleSeoArticle

helpersCheckIsPostiesArticle

hooksArticleAcknowledgementGateOverlayArticle

hooksTrackPageViewArticle

hooksContentInterestArticle

articleLinkingDataContent

articleSchemaContent

liveArticle

advertZone

sentiment

contentLock

topics

types

sections

moreOnThisArticles

urlAlias

commentCount

sponsorType

authorLocations

authors

paywallTypes

corrections

displaySlideShow

multimediaEmbed

printHeadline

seriesContainer

socialHeadline

headline

flag

firstTopic

glossary

flattenTypeIds

publishedDate

additionalInfoBoxes

images

image

relatedLinks

relatedNewsletters

__typename

createdDate

sponsor

factSheets

series

disableOffPlatformContent

published

updatedDate

summary

subHeadline

body

keywords

copyrighted

customTimezone

liveContents

Stephen Chen

Chinese scientists say their artificial intelligence creation behaves just like a human when playing military war games.
Despite some military strategists with rich experience playing either with or against the AI for many rounds, they failed to identify it as a machine.
“AlphaWar passed the Turing test,” the team behind the AI said in a paper published in the Chinese-language journal Acta Automatica Sinica on February 17.
The machine was named after Google DeepMind’s AlphaGo, the first AI to defeat human champions in the complex Chinese board game.


The Turing test was proposed in 1950 by Alan Turing, the father of modern computers. It tried to answer a question that intrigued Turing: can machines think?
In his test, a machine would pretend to be a person and chat with a human interrogator. If it was identified as a machine it failed.
Eugene Goostman, a chatbot developed by Reading University in Britain, claimed to be the first machine to pass the Turing test after fooling 10 out of 30 judges in 2014. But the claim remains controversial, with some researchers calling the five-minute-chat design “nonsense”.
OpenAI’s much discussed ChatGPT and a chatbot developed by Google also claim to have passed the test last year.
AlphaWar was built by a team led by Professor Huang Kaiqi at the Institute of Automation, Chinese Academy of Sciences in Beijing. It passed the test in 2020, said Huang and his colleagues in the paper without giving details.
It is not clear why the news did not emerge for more than two years. Huang’s team could not be reached for comment.
A professional war game simulating a battle to predict possible outcomes is a serious exercise and is usually held behind closed doors.
A war game conducted by the US think tank Centre for Strategic and International Studies (CSIS) last year predicted Beijing was not likely to seize Taiwan. The decisions and moves in the game were mainly made by humans.
US, China, other nations urge ‘responsible’ use of military AI
Governments have invested heavily in the development of computer-assisted war game technology to improve speed and accuracy.
The US Defence Advanced Research Projects Agency (DARPA) launched Deep Green, a programme to develop war game AI in 2007. In 2020, after Deep Green was discontinued, DARPA launched another programme, Game Breaker, to resume its efforts.
Simulating a real-life battle has been a difficult task, even for the most powerful computers, according to Huang’s team.
A modern military campaign can involve a huge number of combat and support personnel with vastly different skills or equipment, increasing the scale of problems.
In recent years, AI technology has advanced rapidly and defeated human players in large-scale computer games such as Starcraft, a game considerably more complex than Go.
But some of the most critical pieces in the military war game are human beings, who can make unexpected mistakes or reach surprising achievements in a highly unbalanced situation offering limited knowledge about the opponents.
This uncertainty could change the course of a battle, and it is extremely difficult for AI to learn and mimic, according to Huang’s team.
Another human challenge is that a war game is not really a game – the outcomes can affect the life and death of many people, they added.


The researchers’ paper said AlphaWar could come up with strategies superior to those devised by humans.
The machine improves its performances by learning from military strategists or by playing against itself. But it still makes mistakes and trails slightly behind the top human strategists in a number of areas, including coordination between units and weapon use.
Huang’s team said that some new tools, including large language models similar to ChatGPT, were likely to be employed to improve the performance of AlphaWar and other AI war games.
Slaughterbots? International AI summit to focus on military use
The test proposed by Turing more than 70 years ago remains one of the most intriguing challenges for computer scientists, said a professor of computer science with Tsinghua University who asked not to be named because of the sensitivity of the topic.
The original version was a chat, but some researchers argued that the test could be conducted in other forms of human-computer interaction, as long as the computer’s identity was concealed.
“It really depends on how you define thinking. I personally don’t think AlphaWar has passed the Turing test in the strictest sense,” the professor said.
“I don’t think any machines have,” he added.


How a Chinese spacecraft will catch Mars rock samples for journey to Earth

Global climate expert Chen Deliang returns to China after decades in Europe

For Shanghai-born international maths star Yilin Wang, curiosity is what counts

China turns on hypergravity machine to ‘compress’ time and space

Science

Chinese AI plays war games like a human, with military strategists unable to identify it as a machine, developers say

Artificial intelligence is being inserted into military strategy. A Chinese team has developed AlphaWar, an AI creation it says has passed the Turing test and can behave like a human. Photo: US Navy via AP

Beijing to host artificial intelligence conference amid ChatGPT frenzy, joining other cities in promoting the industry

Men interact with a Baidu's artificial intelligence robot at the company's headquarters in Beijing on April 23, 2021. Baidu is a partner of its home city’s Beijing Artificial Intelligence Industry Innovation and Development Conference taking place on February 13. Photo: Reuters

People interact with the AI-powered Microsoft Bing search engine and Edge browser during an event at the company’s headquarters in Redmond, Washington, on February 7. Photo: Bloomberg

US unveils artificial intelligence ‘Bill of Rights’ to safeguard civil rights from technological abuse

A live demonstration uses artificial intelligence and facial recognition in a dense crowd at the Horizon Robotics exhibit at the Las Vegas Convention Center during CES 2019. Photo: TNS

Chinese military short of troops trained in hi-tech operations, PLA Daily reveals in rare show of candour

Chinese President Xi Jinping says the PLA must move faster to become a world-class fighting force. Photo: Xinhua

China’s Shandong aircraft carrier sails in formation in the South China Sea. Photo: Handout

Chinese military focuses on stronger NCO ‘backbone’ with longer service, more skills

A military parade in Beijing celebrates the founding of the People’s Republic of China. Photo: Xinhua

US weighs more China tech restrictions, targeting quantum computing and artificial intelligence

US President Joe Biden looks at the IBM System One quantum computer with New York Governor Kathy Hochul during a tour of an IBM facility in Poughkeepsie, New York, on October 6, 2022. Photo: AP

News

China

‘AlphaWar passed the Turing test,’ scientists from the Institute of Automation report in paper
It learns from military strategists or by playing against itself but trails behind top human strategists in unit coordination and weapons use


‘AlphaWar passed the Turing test,’ scientists from the Institute of Automation report in paper.


Chinese AI plays war games like a human, with military strategists unable to identify it as a machine, developers say

‘AlphaWar passed the Turing test,’ scientists from the Institute of Automation report in paper It learns from military strategists or by playing against itself but trails behind top human strategists in unit coordination and weapons use

‘AlphaWar passed the Turing test,’ scientists from the Institute of Automation report in paper

It learns from military strategists or by playing against itself but trails behind top human strategists in unit coordination and weapons use