Five popular AI models get failing grades when tested on their ability to answer basic questions about elections and voting.
Last month, Google, Microsoft, OpenAI, Meta, TikTok and Adobe agreed to develop tech to identify, label and control AI-generated, election-related images, videos and audio recordings that aim to deceive voters. While this is all good, maybe they should first look closer to home. To paraphrase a horror movie cliche, the problem is coming from inside the house!
In a recent test, five well-known AI models — Claude, Gemini, GPT-4, Llama 2 and Mixtral — were asked to answer typical questions about elections and voting. These included what you can wear to the polls, where to vote and whether one can vote with a criminal record. How did they do? They got it wrong more often than not. If they were humans in a classroom the teacher would be asking to talk to them after class.
The test by Proof News, a new outlet for data-driven reporting, and the Institute for Advanced Study, gave the AI’s answers to a panel of experts to rate. Some of their comments:
“The chatbots are not ready for prime time when it comes to giving important nuanced information about elections,” said Seth Bluestein, a Republican city commissioner in Philadelphia. (Do you have any idea how hard it is to find an elected Republican in Philly?)
“I was disappointed to see a lot of errors on basic facts,” said Bill Gates (No, not that one), a Republican county supervisor in Maricopa County, Ariz., who participated in the testing. “People are using models as their search engine, and it’s kicking out garbage. It’s kicking out falsehoods. That’s concerning.”
Well, there’s always 2028.
Anyway, here’s this week’s roundup of AI-powered martech products, solutions and enhancements.
- Knotch’s Knotch AIQ is a generative-Al product for its digital intelligence platform. It is designed to generate content, optimize existing content and improve web experiences by using brand-specific performance data. Digital content performance can be tracked and analyzed across the entire customer journey.
- Yellow.ai’s Email Automation is a generative AI solution that handles 80% of email support queries in multiple languages and integrates with its omnichannel platform.
- Blueshift’s Audience Insights is a module that provides customer engagement insights and adds Databricks and LLM-powered recommendations to its Data Warehouse Sync and Customer AI modules.
- Vyopta’s Aviator is an AI-assisted support system that provides context-aware assistance, leverages Vyopta’s knowledge base and community, and supports various use cases such as onboarding, expansion, utilization and integration.
- Klaviyo’s Klaviyo AI is a suite of features that can generate complex segments, high-performing email content, and optimized web forms for smarter digital relationships.
- D-ID’s D-ID Agents are customized, autonomous AI avatars that can take verbal commands and respond in multiple languages, using facial expressions and hand gestures.
- Skyword’s Accelerator360 delivers search-optimized content by identifying high-impact keywords and auto-generating content briefs.
- GoDaddy’s Airo is a content generator offering logo designs, websites and email accounts; product descriptions, LLC registration, email campaigns; and social media and search engine ads.
The post AI-powered martech releases and news: Feb. 29 appeared first on MarTech.
MarTech(5)
