It’s incredibly easy to contaminate large-scale language models (LLMs). What’s even more disturbing is that you don’t need access to do that.
A new study by scientists at New York University delves into how much medical misinformation AI can handle before it starts giving incorrect answers. They found that even a small amount of misinformation, just 0.001% of the data it’s learning from, can confuse the AI.
While this research focuses on people deliberately trying to trick AI during training, it also highlights a larger issue. It’s all the misinformation that’s already circulating online that these AIs are learning from. Additionally, some databases still contain outdated medical information. Sci-Hub, a free access/piracy site for scientific research, found that only 15% of papers retracted by publishers were removed from the site.
The good news is that contaminated data has to compete for attention with potentially accurate information. Therefore, the ability to poison an LLM may depend on the topic. Bad news: LLM is trained on data pulled from the internet, much of which is garbage.
Here’s a roundup of this week’s AI-powered martech releases.
Amazon Ads announced new features within Amazon Marketing Cloud (AMC). This new feature allows advertisers to use natural language to generate SQL queries targeted to their desired audience. This eliminates the need to write code and significantly reduces the time required to develop audience queries. Advertisers can run these queries within AMC to create new audiences and activate them across Amazon DSP and advertising consoles. Cordial’s Cordial Edge uses multimodal AI to generate personalized marketing models for individual retail brands. These models consider various data types, including brand creative, illustrations, photos, and text, to optimize marketing messages for increased purchases. PearlMountain Limited has released an update to its video creation platform FlexClip. This update includes two new tools: an AI video generator and a motion tracking feature. AI video generators enhance creative possibilities, and motion tracking features simplify complex editing tasks. Jellypod launches an AI podcast studio. The platform offers features such as AI-powered voice cloning, content control, and robust content integration. Jellypod aims to simplify podcasting by helping creators with every aspect of production, from ideation to final editing and publishing. Dappier announced the availability of AskAI banner ads. This feature allows digital publishers to integrate conversational AI response engines within their banner ads. This in-ad unit provides contextual knowledge of user content and provides advertisers with a high-profile environment to connect with consumers. HUMAN Security partners with TollBit to protect against rogue AI scraping bots, enforce robots.txt directives, and help publishers monetize their content. Jivox has updated the Jivox IQ DaVinci platform. This update includes deep API integration with leading retailer DSPs such as Amazon DSP and Walmart Connect DSP. The platform combines your audience with creative and media in a single workflow. Integral Ad Science has launched an AI-powered Total Media Performance (TMP) solution. TMP integrates media quality signals and on-the-fly optimization to help advertisers achieve better campaign outcomes. Creative Realities launches AdLogic CPM+TM, a campaign planning and management platform. The platform provides programmatic functionality and is designed for in-store retail media networks. AdLogic CPM+TM allows advertisers and publishers to plan, manage, and optimize campaigns within a unified AdTech stack. BrandRep has launched an AI-powered keyword generator. This tool uses artificial intelligence to identify keywords related to small businesses. The goal is to improve your SEO performance and accelerate your growth in the competitive digital space.
Add MarTech to your Google News Feed.