Connect with us

AI December 6, 2022

AI Takeover in Stratego – Meet DeepNash

Avatar photo

Published

on

Another game long believed to be very challenging for artificial intelligence (AI) to conquer has fallen to bots: Stratego.

DeepNash, an AI made by London-based company DeepMind, now matches expert humans at Stratego, a board game requiring long-term strategic thinking against imperfect information.

This latest feat comes in the wake of yet another major win for the AIs in games previously thought to be the forte of humans.

Just last week, Meta’s Cicero, an AI that can outsmart human players at the game of Diplomacy, made history for outsmarting opponents online.

“The rate at which qualitatively different game features have been conquered — or mastered to new levels — by AI in recent years is quite remarkable,” says Michael Wellman at the University of Michigan in Ann Arbor, a computer scientist who studies strategic reasoning and game theory.

“Stratego and Diplomacy are quite different from each other, and also possess challenging features notably different from games for which analogous milestones have been reached,” said Wellman.

Imperfect information

The game has characteristics that are generally much more complicated than chess, Go or poker. Chess, Go and Poker have all been mastered by AIs.

AI Beats Humans in Stratego – Meet DeepMash

In the game of Stratego, two players put 40 pieces each on a board, but must not see what their opponent’s pieces are.

The objective of the game is to move pieces in turns to eliminate those of the opponent and capture a flag. 

Stratego’s game tree — a graph of all possible ways the game could possibly go — has 10535 states against Go’s 10360. 

When it comes to imperfect information at the beginning of a game, Stratego has1066 possible private positions, a figure that dwarfs only 106 such starting situations in two-player Texas hold’em poker.

“The sheer complexity of the number of possible outcomes in Stratego means algorithms that perform well on perfect-information games, and even those that work for poker, don’t work,” says Julien Perolat, a DeepMind researcher based in Paris.

DeepNash was developed by Perolat and his colleagues.

Nash inspired bot

The bot’s name is a tribute to the famous US mathematician John Nash, who came up with the Nash equilibrium theory that supposes that there are a “stable set of strategies” that can be followed by players in a manner that no player benefits by changing strategy on their own. As such, games tend to have zero, one or many Nash equilibria.

DeepNash combines reinforcement-learning algorithm and a deep neural network to find a Nash equilibrium. 

Generally, reinforcement learning is where an intelligent agent (computer program) interacts with the environment and learns the best policy to dictate action for every state of a game. 

In order to have an optimal policy, DeepNash played a total 5.5 billion games against itself. 

In essence, if one side gets penalised, the other is rewarded, and the variables of the neural network — which represent the policy — are tweaked accordingly.

AI Beats Humans in Stratego – Meet DeepMash

At some stage, DeepNash converges on an approximate Nash equilibrium. Unlike other Bots, DeepNash optimises itself without searching through the game tree.

For a duration of two weeks, DeepNash played against human Stratego players on online games platform, Gravon.

After competing in 50 matches, the Ai was ranked third among all Gravon Stratego players since 2002. 

“Our work shows that such a complex game as Stratego, involving imperfect information, does not require search techniques to solve it,” says team member Karl Tuyls, a DeepMind researcher based in Paris. “This is a really big step forward in AI.”

Other researchers are impressed as well by this feat.

Impressive results

“The results are impressive,” agrees Noam Brown, a researcher at Meta AI, headquartered in New York City, and a member of the team that in 2019 reported the poker-playing AI Pluribus4.

At Meta, the parent company of Facebook, Brown and her colleagues built an AI that can play Diplomacy, a game where seven players compete for geographic control of Europe by moving pieces around on a map.

In Diplomacy, the goal is to take control of supply centres by moving units (fleets and armies). 

Meta says Cicero is quite significant because the AI relies on non-adversarial environments.

Unlike in the past where prior major successes for multi-agent AI have been in purely adversarial environments, such as Chess, Go, and Poker, where communication has no value, Cicero employs a strategic reasoning engine and controllable dialogue module.

“When you go beyond two-player zero-sum games, the idea of Nash equilibrium is no longer that useful for playing well with humans,” says Brown.

Brown and her team trained Cicero using data from 125,261 games of an online version of Diplomacy involving human players. 

Using self-play data and a strategic reasoning module (SRM),Cicero learnt to predict judgubg by the state of the game and the accumulated messages, the likely moves and policies of the other players. 

AI Beats Humans in Stratego – Meet DeepMash

Meta says it collected data from 125,261 games of Diplomacy played online at webDiplomacy.net. Of these games, a total 40,408 games contained dialogue, with a total of 12,901,662 messages exchanged between players.

Real-world behaviour

Brown believes game-playing Bots like Cicero can interact with humans and account for “suboptimal or even irrational human actions could pave the way for real-world applications.”

“If you’re making a self-driving car, you don’t want to assume that all the other drivers on the road are perfectly rational, and going to behave optimally,” he says.

Cicero, he adds, is a big step in this direction. “We still have one foot in the game world, but now we have one foot in the real world as well.”

Others such as Wellman agree, but insist more work still needs to done. “Many of these techniques are indeed relevant beyond recreational games” to real-world applications, he says. “Nevertheless, at some point, the leading AI research labs need to get beyond recreational settings, and figure out how to measure scientific progress on the squishier real-world ‘games’ that we actually care about.”

/MetaNews.

SHARE THIS POST
Image credits: Shutterstock, CC images, Midjourney.

AI

ChatGPT Makes History With Fastest 100M Users as StackOverflow Stumbles

Published

on

ChatGPT Makes History With Fastest 100M Users as StackOverflow Stumbles

ChatGPT, the popular chatbot from OpenAI, is now the fastest growing Internet application in history with a record 100 million monthly active users in January, according to a new report by Swiss investment bank UBS.

This comes as web traffic at developer knowledge-sharing platform StackOverflow slumped 11.5% to 247 million since the launch of ChatGPT. OpenAI, a private company backed by Microsoft Corp., launched ChatGPT as a free web interface in late November.

Also read: OpenAI Develops Tool to Spot AI-Written Texts

The AI-powered chatbot became incredibly popular with users thanks to its ability to perform a variety of tasks within seconds. In the U.S. tech hub of Silicon Valley, the conversation tool has virtually sparked off an artificial intelligence (AI) ‘arms race.’

13M ChatGPT users per day

Citing data from analytics firm Similarweb, UBS said in its report that the number of people using ChatGPT averaged about 13 million per day in January – double the number recorded in December. These are all distinct individuals, or unique users, it stated.

“In 20 years following the internet space, we cannot recall a faster ramp in a consumer internet app,” UBS analysts noted, according to Reuters.

By comparison, TikTok took about nine months after its global launch to get to 100 million users and Meta’s Instagram two and half years.

ChatGPT is a large language model developed by OpenAI. It has been trained on a diverse range of internet text to generate human-like responses to user prompts. The tool is based on OpenAI’s GPT-3.5 language technology, an upgrade of the GPT-3 model released in 2020.

The chatbot was trained with a lot of data, from books and articles to conversations. It can understand a variety of topics and contexts, and can be fine-tuned for multiple language tasks such as question answering, language translation, and text summarization.

ChatGPT accumulated more than one million users within its first week. The bot has transformed several industries since it was released just over two months ago. It can write complex essays, poetry, code and even pass MBA exams at the Wharton School of Business.

While ChatGPT is popular with some, others are concerned about the impact it will have on education and other industries. Gmail creator Paul Buchheit said recently ChatGPT-like tech will destroy Google’s $149 billion search engine business and dominance in two years.

StackOverflow traffic declines

Meanwhile, the rise of ChatGPT has started to affect rival StackOverflow, the Naspers-owned question and answer website for professional and enthusiast programmers.

According to Similarweb, the number of people visiting the StackOverflow website fell 11.5% just weeks after the launch of ChatGPT. On average, users dropped from almost 279 million in November 2022 to 247.4 million by the end of December 2022.

StackOverflow dropped 30 positions in the “Programming and Developer Software” websites category, falling from the 202nd position in October 2022 to 229th by the end of December 2022. It is the company’s lowest ranking in that category since October 2019, the data says.

The website serves as a platform for users to ask and answer questions, and, through membership and active participation, to vote questions and answers up or down similar to Reddit. They can also edit questions and answers in a fashion similar to a Wiki.

In December, StackOverflow banned the use of answers generated using ChatGPT, citing concerns about the accuracy of such responses.

“…because the average rate of getting correct answers from ChatGPT is too low, the posting of answers created by ChatGPT is substantially harmful to the site and to users who are asking or looking for correct answers,” the company said in a statement.

As MetaNews previously reported, the computational resources required to run ChatGPT over large scales are very costly. OpenAI has now launched a paid subscription service known as ChatGPT Plus for $20 per month, hoping that will help to cover costs.

Continue Reading

AI

Microsoft Warns Employees Not to Share Sensitive Data with ChatGPT

Published

on

Microsoft Warns Employees Not to Share Sensitive Data with ChatGPT

Microsoft has warned its employees not to share sensitive data with an artificially intelligent (AI) chatbot, ChatGPT from OpenAI. Employees of American multinational tech giants had asked in an internal forum whether ChatGPT or any other AI tools from OpenAI were appropriate to use at their work, Business Insider reported.

Also read: 30% of College Students Use ChatGPT

In response to that inquiry, a senior engineer from Microsoft’s CTO office allowed to use ChatGPT but couldn’t share confidential information with the AI chatbot.

“Please don’t send sensitive data to an OpenAI endpoint, as they may use it for training future models,” the senior engineer wrote in an internal post, per Insider.

ChatGPT, here only for two months, is already raising concerns in the academic sector. Microsoft has become a partner of OpenAI, the parent company of ChatGPT, and has confirmed an investment of ten billion dollars.

Microsoft is planning to integrate OpenAI’s technology into its products, including the Bing search engine and other software, to enhance their capabilities, as reported previously.

The major concern of Microsoft regarding “sensitive information” may include sharing internal software code and seeking checks and advice from the chatbot.

Amazon’s Same Concern

ChatGPT has continuously made headlines since its launch last November but has also faced bans, especially in the academic sector as it became the cheating partner for students’ schoolwork. Recently, the tech giants have also raised their concerns over its use.

Amazon warned its employees to beware of ChatGPT last week, as reported by Insider. Insider claims that an Amazon lawyer has urged employees not to share code with ChatGPT via an internal communication form.

“This is important because your inputs may be used as training data for a further iteration of ChatGPT, and we wouldn’t want its output to include or resemble our confidential information (and I’ve already seen instances where its output closely matches existing material),” the lawyer wrote.

The lawyer placed more emphasis on requesting that employees not share “any Amazon confidential information” (including Amazon code they are working on) with ChatGPT via Slack.

Personal Data Concern

As concerns about data privacy grow among large corporations, an OpenAI representative has directed questions about the company’s data and privacy policy to ChatGPT’s FAQ page. The terms of service of OpenAI grant the company the right to use all input and output generated by ChatGPT users, with the stipulation that personally identifiable information (PII) is removed from the used data.

However, it’s quite impossible for OpenAI to identify and remove all the personal information from the data provided to ChatGPT, says Emily Bender, who teaches computational linguistics at the University of Washington.

“OpenAI is far from transparent about how they use the data, but if it’s being folded into training data, I would expect corporations to wonder: After a few months of widespread use of ChatGPT, will it become possible to extract private corporate information with cleverly crafted prompts?” said Bender.

Vincent Conitzer, a computer science professor and director of an AI lab at Carnegie Mellon University, said, “All of us together are going to have to figure out what should be expected of everyone in these situations. Is the responsibility on employees to not share sensitive information, or is the responsibility on OpenAI to use information carefully, or some combination?”

Continue Reading

AI

Gmail Creator Says AI Will Replace Search Engines Like Google in 2 Years

Published

on

Gmail Creator Says AI Will Replace Search Engines Like Google in 2 Years

Paul Buchheit, the creator of Gmail, said artificial intelligence (AI) will replace search engines such as Google in two years.

“My prediction is that AI (not necessarily ChatGPT) will replace search, maybe within the next two years (current AI not yet good enough),” Buchheit tweeted on Feb. 1.

Google facing ‘total disruption’

The software engineer said this as a form of clarification on an earlier statement he made regarding ChatGPT, the widely popular OpenAI bot. He believes that his comments were taken out of context and “sensationalized” by the media.

Also read: AI Could Kill Everyone, Researchers Warn MPs

However, the argument remains fundamentally the same in that Paul Buchheit’s prediction will still impact his former employer. Google is already reacting to the threat with plans to release 20 new AI related products to improve search this year, according to reports.

In his original comments from early December, Buchheit stated: “Google may be only a year or two away from total disruption. AI will eliminate the search engine result page, which is where they make most of their money,” he wrote on Twitter.

“Even if they catch up on AI, they can’t fully deploy it without destroying the most valuable part of their business.”

Google gets much of its revenue from search through online advertisements. The tech giant, which dominates the search business, charges advertisers a fee for ad placements that sit next to the search results. This improves the chances of ads being seen by consumers.

In 2021, Google reported revenues rose 41% to $258 billion from a year ago, its biggest to date. Ads made up 81% of total revenues that year. However, Google is grappling with a slowdown in online ad spending amid mounting antitrust lawsuits.

According to market research firm Insider Intelligence, the company’s share of total online advertising income in the U.S. fell from 37% in 2016 to 29% in 2022. Alphabet-owned Google controls up to 90% of the search business’ total market share.

Killing Web search ChatGPT way

Culture is now a sanctuary in this posthuman future. But the upcoming dominance of AI in search has even the most cultured corporate organizations in panic, and none more so than Google. OpenAI’s buzzy chatbot ChatGPT has been all the rage since its launch in November.

It can write complex essays, poetry, code and even pass MBA exams at the Wharton School of Business. OpenAI has also received a reported $10 billion cash boost from Microsoft, which is looking to add the chatbot’s abilities into its own search engine.

ChatGPT is shaking things up. It has spawned an AI arms race of sorts in Sillicon Valley. As Paul Buchheit said, tech like AI can remove the need for search engine result pages. He elaborates on how search businesses like Google could become useless, in exactly the same way that the company “killed” pre-internet business The Yellow Pages.

“The way I imagine this happening is that the URL/Search bar of the browser gets replaced with AI that autocompletes my thought/question as I type it while also providing the best answer (which may be a link to a website or product),” he explained.

“The old search engine backend will be used by the AI to gather relevant information and links, which will then be summarized for the user. It’s like asking a professional human researcher to do the work, except the AI will instantly do what would take many minutes for a human.”

Google joins AI race

The new artificial intelligence technology has been a wake up call for Google. CEO Sundar Pichai reportedly declared a “code red”, steamrolling over the company’s existing plans to kickstart AI development, The New York Times reports.

Google is now targeting to release around 20 new products and “demonstrate a version of its search engine with chatbot features this year,” it said, quoting sources with knowledge of the plans.

One of these projects is “Apprentice Bard,” as per a CNBC report. Apprentice Bard is a chatbot that is able to provide responses to questions asked using natural language. It uses Google’s conversation technology called Language Model for Dialogue Applications (LaMDA). The model is similar to the GPT language model utilized by ChatGPT .

Continue Reading

News Feed

Advertise With Us

Unlock a wide range of advertising
opportunities with MetaNews to reach the
fast-paced meta world.

Publish Your PR

Share your press release with
MetaNews’s growing global audience,
fans, and followers.

Subscribe

Sign up here to get news & updates right to your inbox!

Copyright © 1997 – 2023 MetaNews All Rights Reserved

Copyright © 1997 - 2023 MetaNews All Rights Reserved