SpaceX IPO Sparks Market Optimism as Shares Surge 19% on Trading Debut 
Google Gemini Co-Lead Noam Shazeer Leaves for OpenAI Amid AI Talent Race 
Anthropic Restricts Global Access to AI Models After U.S. Security Review 
Kingboard Holdings Shares Surge After HK$11.77 Billion Block Trade to Expand PCB and AI Supply Chain Business 
US Raises Concerns Over Possible ASML EUV Machine Transfer to China 
SoftBank Shares Drop as OpenAI Losses and Rising Costs Spark Investor Concerns 
SpaceX Stock Slides After IPO Rally as Valuation Concerns Grow 
Hanmi Semicon Shares Surge After $33 Million SpaceX Investment 
Samsung Gains Interest from BYD, Google, AMD as AI Chip Demand Strains TSMC Capacity 
Apple Signals Product Price Hikes Amid Rising Memory Chip Costs 
G7 Explores AI Access Deal With U.S. Amid Anthropic Restrictions 
SK Hynix Shares Hit Record High After Shipping Next-Generation HBM4E AI Memory Samples 
Chinese Social Media Giant Xiaohongshu Eyes Hong Kong IPO at Over $70 Billion Valuation 
Trump Administration Delays DeepSeek and CXMT Trade Blacklist Designations Amid U.S.-China Tensions 
Anthropic Officials Meet White House Over AI Model Outage 
Trump Says Anthropic No Longer Seen as National Security Threat 
In a world where it is becoming increasingly clear that the age of artificial intelligence is inevitable, much of the fear directed towards the technology has to do with machines eliminating humans. This hasn’t happened yet, but what better way to speed up the process than by having AIs practice fighting each other first? This is exactly what Google’s DeepMind is trying to do just to see what would happen.
Now, it’s important to keep in mind that DeepMind is not out to create violent robots on purpose. Rather, what it is trying to accomplish is determine how AIs react to particular situations where they are required to engage opposing forces or challenges, Engadget reports.
By pitting two AIs against each other, both of which have differing goals that they need to accomplish, though, not necessarily in direct conflict, it’s important to find out if they decide to fight each other or cooperate. DeepMind is doing this through two games that were created exactly for this scenario.
One is called Gathering and the other is called Wolfpack. The former involves two units tasked to collect apples, thus the title Gathering. The units are also given laser weaponry, which allows them to remove the other from the game in order to collect as many of the apples as possible.
As for Wolfpack, the idea is for the two AIs to work together in order to capture a third unit by navigating a series of blocky obstacles. The idea is to have the two AIs figure out the best way to corner their prey.
So, what were the results? In Gathering, the two AIs displayed high levels of aggression, particularly when the apples start running out. In contrast, the algorithms worked together well in Wolfpack. Interestingly enough, the AIs also became aggressive and cooperative in the two games respectively when given a boost in computation power.
This suggests that with current AI technology, the behavior of the machines will depend largely on the rules imposed on them, The Verge reports. Depending on the context, the machines will either work with others or eliminate them.
