The University of Science and Technology of China (USTC) and Tencent YouTu Lab have released a research paper on a new framework called Woodpecker, designed to correct hallucinations in multimodal large language AI models. “Hallucination is a big shadow hanging over the rapidly evolving MLLMs,” writes the group, describing the phenomenon as when MLLMs “output descriptions that are inconsistent with the input image.” Solutions to date focus mainly on “instruction-tuning,” a form of retraining that is data and computation intensive. Woodpecker takes a training-free approach that purports to correct hallucinations from the basis of the generated text. Continue reading Woodpecker: Chinese Researchers Combat AI Hallucinations

The algorithm powering TikTok’s vaunted For You page is reportedly getting help from human collaborators. Although the personalized feed was said to be based on user interests and selections, “employees regularly engage in ‘heating,’ a manual push that ensures specific videos ‘achieve a certain number of video views,’ according to six sources and documents reviewed by Forbes.” What’s more, while the algorithm does have a say in what goes viral, staff at TikTok and ByteDance are also hand-picking specific videos to give preferential treatment, saturating their distribution throughout the user base. Continue reading TikTok Is Accused of Manually ‘Heating’ Personalization Feed

Twitter has unveiled a new feature that will let users directly report misinformation and fake news with regard to elections. Among the elections in India on April 25 and Europe on April 29, Twitter now offers the option “It’s misleading about voting” in the “Report Tweet” dropdown menu. On its blog, Twitter states that “voting is a fundamental human right … any attempts to undermine the process of registering to vote or engaging in the electoral process is contrary to our company’s core values.” Continue reading Twitter Launches Feature to Report Election Misinformation

Twitter chief executive Jack Dorsey has brought his own company to task, for what he said is a less-than-stellar performance in handling malicious activity. More specifically, he said that he did not move quickly enough to take action against the Russian efforts to create divisions between Americans. Dorsey, who has previously expressed contrition for his slow reaction, has now made it clear that the company needs to take actions to prevent this from happening again, rather than just reacting after the fact. Continue reading Twitter to Hire Experts, Accept Proposals to Stop Bad Actors