This AI is capable of distinguishing between authentic online reviews and fraudulent ones, and its accuracy is quite impressive.

      AI is becoming increasingly adept at identifying untrustworthy reviews.

      Fake reviews pose a significant threat to online consumers. If you've ever made a purchase based on highly positive reviews only to find the product lacking, you understand the issue. A recent study published in the International Journal of Information and Communication Technology suggests an AI-driven system that can not only identify fake reviews but also track their dissemination.

      Why current tools are inadequate

      Most current fake review detection systems hone in on the text of a review. This strategy was effective for a time, but fake reviewers have evolved. They now combine meticulously crafted text with deceptive images to create the appearance of authenticity. Tools that rely solely on text often fail to catch this, which creates problems for both shoppers and genuine sellers.

      Structure of the fake review detection model by Bing Duan / International Journal of Information and Communication Technology

      The researchers tackled this issue by developing a system that assesses multiple signals simultaneously. It examines the review text through two distinct methods: a text convolutional neural network and pre-trained language models, capturing both superficial and deeper meanings. It also considers reviewer behavior, as fake accounts typically have default profile pictures and system-generated usernames, unlike real users who personalize their profiles.

      Can AI detect a fake image as well?

      The concise answer is yes. Images associated with reviews are assessed separately using a residual network, a common deep learning tool for visual processing. After all signals are collected, the system combines them to determine whether a review is authentic.

      When a review is identified as fake, a Transformer model is activated to trace its origin and monitor its spread throughout the network.

      Framework for tracking the dissemination of fake reviews by Bing Duan / International Journal of Information and Communication Technology

      Testing on a vast dataset from JD.com revealed that the system achieved a recognition accuracy of 94.2% and a tracing accuracy of 93.5%, surpassing all existing methods it was compared with. This level of precision could lead to fewer deceptive reviews and more reliable ratings for consumers.

      Rachit is an experienced tech journalist with over a decade of experience in the consumer technology field.

      You can now send emails directly from ChatGPT on the web.

      You can now draft, edit, and send emails directly from within ChatGPT on the web.

      If you've ever composed an email in ChatGPT and needed to copy it to Gmail or Outlook to send, you can now bypass that step. OpenAI has introduced the capability to send emails directly from writing blocks in ChatGPT on the web, allowing the entire process to remain within a single conversation.

      It turns out that teaching games like Battleship can significantly enhance small AI models.

      By using Battleship as a platform for AI training, researchers have improved the reasoning capabilities of smaller models.

      Small AI models received an unexpected boost from a classic game. MIT researchers employed a Battleship-inspired approach to test whether AI agents could enhance their information-gathering methods before making a move. The result was a noticeable increase in performance for smaller systems, including one model that improved from rarely defeating humans to winning most of its games after researchers modified its board-searching strategies.

      Steam Machine is confirmed to launch this summer, but its pricing remains unclear.

      Steam Machine is nearing its release, with expanded game verification set to arrive before Valve announces its price.

      Valve has confirmed that Steam Machine will be shipping this summer, providing PC gamers with a clear launch timeframe for its SteamOS living room PC. However, the crucial detail of pricing is still unknown, which many potential buyers need in order to determine if it fits their gaming setup. This update coincides with Valve's expansion of its Verified program to include Steam Machine and Steam Frame. For Steam Machine, games will be evaluated for default controller support, graphics settings, and performance without manual configuration. Valve claims the hardware is approximately six times more powerful than the Steam Deck while still utilizing SteamOS, the Steam interface, and Proton.