Daily Tech Briefing
AI 科技速览

每天 5 分钟内学习 AI。获取最新的人工智能新闻,理解其重要性,并学习如何将其应用于您的工作。

AI 快讯
HackerNoon AI · 2026/7/29 16:47:57

Onton's New AI Trust Model Beats

AI 中文解读
Onton推出了一款全新自研AI模型,专门用来判断网上商品信息的真假,在准确性比拼中,居然把谷歌购物和亚马逊都甩在了身后。这款模型就像一个“商品信息鉴谎师”,能自动识别哪些产品描述和评论是可信的。如今AI代购越来越火,但网上大约30%的评论都是假的,每年给美国企业造成1520亿美元的损失。Onton的模型完全自主学习,还不局限于电商,未来可能成为整个互联网的信任层。对普通人来说,以后用AI助手购物会更放心,再也不用担心被虚假好评或夸大宣传带进坑里,买得安心,也逼得商家不敢轻易造假。
Onton, the San Francisco product discovery engine with 2M+ monthly active users, has launched a from-scratch AI model built to judge whether product information can be trusted, aimed at the agentic web where AI systems research and execute purchases for people. In head-to-head benchmarks against Google Shopping and Amazon, Onton says its model outperformed on accuracy across essentially every dimension tested, with the widest gap in interpreting the veracity of product information. The claim is backed by a whitepaper. The launch lands as AI-referred retail traffic explodes: up 693% YoY over Holiday 2025 and up 1,324% cumulatively since October 2024, per Adobe Analytics, while an estimated 30% of online reviews are fake, costing US businesses roughly $152B a year. Onton says the model learns entirely on its own and shows signs of generalizing beyond e-commerce, positioning the company to offer a trust layer for the broader internet, available today on Onton.com and case by case to partners. Onton's trust and authenticity model, launched July 29, 2026, is a from-scratch AI system that evaluates whether product information online can be trusted. In benchmarks it outperformed Google Shopping and Amazon on accuracy, especially at judging the veracity of product data. It learns autonomously, shows signs of generalizing beyond e-commerce, and is available on Onton.com and case by case to agentic web partners. What did Onton launch in July 2026? A from-scratch AI trust and authenticity model for the agentic web that evaluates whether product information can be trusted. How does Onton's model compare to Google Shopping and Amazon? In Onton's head-to-head benchmarks, backed by a whitepaper, the model outperformed both on accuracy across essentially every dimension tested. Why does agentic commerce need a trust layer? Roughly 30% of online reviews are estimated to be fake, and AI-referred retail traffic now converts up to 54% better than other channels, so manipulated inputs directly drive high-value purchases. Who can use Onton's trust model? All users on Onton.com today, and partners building on the agentic web on a case-by-case basis. Read All
分享
阅读原文