01-26-2021, 03:33 AM
![[Image: 0b37a061466bc876472ca710e6e139d8]](https://s.yimg.com/ny/api/res/1.2/DHepYIPql1o99bE2stgqpw--/YXBwaWQ9aGlnaGxhbmRlcjt3PTY0MDtoPTM2MA--/https://s.yimg.com/uu/api/res/1.2/rc_T0j8VuICo9cPSyIHQbA--~B/aD01NDk7dz05NzY7YXBwaWQ9eXRhY2h5b24-/https://media.zenfs.com/en/bbc_us_articles_995/0b37a061466bc876472ca710e6e139d8)
Earlier this month, the Olympics for Hackbots were held: the 11th annual Artificial Intelligence (AI) competition trained to negotiate.It's called an Auto-Negotiation Championships with more than 100 participants from Japan, France, Israel, Turkey and the United States against each other in five leagues.This will be held manually. (Or in silicon) in Japan as part of the International Conference on Artificial Intelligence But because of the corona virus, the competition is part of the virtual convention.Universities from Turkey and Japan were the big winners this year, with an altercation between humans and each other: simulating a factory manager who managed the supply chain and the Werewolf game.
In some leagues, they negotiate with real-life humans, recruited from the slotxo web, on others bots negotiate with other bots.In its early years, [AI] easily outperformed humans," said Tim Baarslag, co-founder of the hagglebot game at Centrum Wiskunde & Informatica, the Dutch National Institute of Mathematics and Computer Science.But now they Behavior is closer to humans, sometimes better than humans. It's only in the artificial domain, ”he said.Humans have a high level of emotional understanding and expertise. But can fail when there are many issues Diplomacy software has a long history - some bargaining support systems began to emerge in the early 1980s: with names such as Inspire and Negoisst.These are taken from the academic work of game theory, which simulates how rational decision-makers make decisions based on a perspective of what other decision makers will do.
The tool tries to help negotiators prepare proposals and strategies, and then help them reach a favorable agreement. "Don't leave money on the table," Baarslag said.But now artificial intelligence for trading horses that can work on its own has begun to emerge in the business.If you're selling to Walmart, you've probably already found the AI system developed by Pactum, tested by the major US retailer.Chatbot contacts suppliers and invites them to negotiate contracts together for things like prices and payment terms.For some suppliers, there are as many as 30 payment points.We now see that some sellers like to talk to bots. If there are tens of thousands of sellers, it's sometimes difficult to get human attention, ”said Martin Rand, chief executive of Pactum, based in Tallinn and Silicon Valley.This bot often offers sellers irrelevant offers asking if they would like to increase their checkout by a certain number of days or keep the product in their warehouse.
Then, from these answers, the bot learns their preferences, Mr Rand explains.If you know everything "The negotiations should not happen at all, you have to have zero rounds and the deal will happen immediately," Barlot said.The most common negotiation is figuring out which things are more important to your partner and where they'll walk away.Machine learning is now helping AI predict another's preferences based on little observation, as well as a wealth of experience in previous negotiations.And when you know these things, even thousands of side-deal-related negotiations. "It turned out to be a great computation, what a computer is amazingly good at," according to Baarslag.We have a problem at BP, we need 90 to 120 days to get the contract," said Michael O'Brien, who manages a $ 2 billion (£ 1.5 billion) budget for the oil company's IT services from 2012 to 2020.Mr O'Brien found that his department spent 80% of his time negotiating terms and conditions from the start with sellers.It takes just 20% of the time to sort out the toughest issues like product and service quality disputes.With thousands of custom contracts, there is no way to know how often a supplier agrees to standard terms or with a heavily negotiated counterpart.
There are 1,400 vendors. “I can't read 1,400 contracts,” O'Brien said.The contract would "go into the filing cabinet and I don't remember what we agreed upon," he said.A company might ask a contractor to insure $ 10 million for a $ 10,000 project, and then usually pay a tenth of that.So he created a machine learning tool with AI App Orchid developers that can learn what BP agreed with in past contracts.And tools can use that information to offer additional options to suppliers negotiating.Concept suppliers can negotiate with AI and make contracts with less time and effort, a revolution, O'Brien said.The supplier said to him: "If you tell me I chose this option, which is bad for me? But it's not always bad, we can make a contract tomorrow, I'll click there and let's get started. It was accomplished, ”he said.This resulted in an 80% reduction in jobs, with contracts between one and eight.