Meta Agrees to $18 Billion Settlement, Tightens Teen Social Media Rules 
OpenAI Data Center Chief Chris Malone Exits Ahead of Planned 2027 Listing 
SpaceX Mobile Network Could Cost Up to $130 Billion, Bernstein Says 
South Korea Unveils Record $597 Billion 2027 Budget to Boost AI and Chips 
HP Stock Drops 9% Despite Q3 Earnings Beat and Raised 2026 Outlook 
BofA Names ASML Top Semiconductor Equipment Pick on Growth, Valuation 
PayPal Shares Sink as $50 Billion Takeover Bid Collapses 
Nvidia Revenue Soars 106% as AI Demand Hits ‘Inflection Point’ 
Nvidia Pauses AI Cloud Financing Deals Amid Antitrust Concerns 
U.S. Probes Apex Logistics Over Nvidia AI Chip Shipments to China 
U.S. Disrupts China-Linked Hacking Campaign Targeting Government Agencies 
Pinterest Stock Falls as CFO Julia Brau Donnelly Resigns 
Google Expands Gemini Enterprise With AI Tools for Legal Sector 
Unitree Shares Plunge 45% After Blockbuster IPO, Raising China Tech Bubble Fears 
Luxshare Shares Rise as First-Half Profit Jumps 18% 
Add as a preferred source on Google
Last year, OpenAI proved that its artificial intelligence can win against professional “DOTA 2” players in a 1v1 setting. However, a lot of concessions had to be made to make that happen, including reducing the challenges of the game. This time, the AI is ready to compete in 5v5 matches after training for the equivalent of 180 years in a single day.
In its announcement, OpenAI made it clear that it wants to challenge actual pros during the “DOTA 2” tournament that will be held in August, Quartz reports. This will be a new hurdle for the technology because the matches will be held as is. There will be no restrictions to make things easier for the AI.
Not that it really needs it this time. Apparently, the machine can make over 20,000 moves in over 45 minutes while playing the game. For a title with as much complexity as this MOBA, this was simply necessary.
On that note, the team behind the AI was not even sure if it could actually be scaled to participate in 5v5 matches. After all, it will be five bots working together to take down experienced professionals in the game. In order to prepare these AI units, they were allowed to learn by matching against each other for extended periods of time.
Apparently, one bot even trained for four weeks. It racked up over 5,000 years of game time. The idea behind why this was done was simply to allow the AI to test things out and explore. This was what Greg Brockman, CTO of OpenAI, told Quartz.
“As long as the AI can explore, it will learn, given enough time,” Brockman said.
The CTO also spoke with The Verge to explain what this experiment is meant to accomplish. After all, OpenAI is not an entity that was created to focus solely on “DOTA 2.”
“This an exciting milestone, and it’s really because it’s about transitioning to real-life applications,” Brockman told the publication. “If you’ve got a simulation [of a problem] and you can run it large enough scale, there’s no barrier to what you can do with this.”
