Latest AI Developments: Meta’s AI Personas, Pearson’s AI Study Buddy and More
Today, on the first day of August, significant developments have been revealed in the realm of artificial intelligence. YouTube has unveiled its plans to introduce experimental AI capabilities that will enable the platform to automatically generate video summaries. Additionally, Meta, the parent company of Facebook and Instagram, is currently in the process of creating AI personas that will revolutionize the way users engage with these social media platforms. These are just a few highlights from today’s AI roundup, so let’s delve deeper into the details.
YouTube AI summary generation
According to Google’s support page, YouTube has begun testing a new AI feature that automatically creates summaries for the platform. As part of the test, summaries will only be created for a limited number of English-language videos, and they will only be viewable by a few users at this stage. However, the feature is expected to be rolled out to a wider user base if it gives good results.
These summaries can be read on the watch and search pages and give you a quick overview of the video. This feature is not intended to replace video descriptions.
Meta may soon add AI persona to Facebook, Instagram
Meta is developing a variety of AI-powered chatbots that could change the way its social media platforms work. Notable among them are Facebook and Instagram’s chatbots, which are internally called “persona,” according to a Financial Times report. These features can be implemented as soon as September 2023. The purpose of these individuals is to improve and provide a new search, recommendation and interaction with the platform.
The reasoning ability of the GPT-3 is similar to that of an undergraduate student
According to a PTI report, researchers at the University of California, Los Angeles (UCLA) conducted an AI experiment which highlighted that the GPT-3 AI model has the same reasoning ability as an undergraduate student. During the experiment, the researchers asked GPT-3 to predict the next shape that followed several complex arrangements. They also asked the AI to answer SAT analogy questions while making sure the AI had never encountered these questions before. The same problems were also given to 40 UCLA students.
GPT-3 was able to solve 80 percent of the questions, which is between the average human score (which was just under 60 percent) and the highest scores. Interestingly, the test found that the AI also made similar mistakes as humans.
Pearson may soon start a fellow student in artificial intelligence
One of the world’s top five textbook makers, Pearson, may soon launch an AI researcher chatbot that is “free from the noise and corruption of web-based AI models,” according to a report from Gizmodo. The company also revealed that this AI tool will be featured in its Pearson+ and Mastering subscription services and will be available in time for US schools to reopen after the summer holidays.
Cybercriminals are training AI chatbots for attacks
A new hacking AI tool called FraudGPT has appeared, while another based on Google Bard is in development, reports BleepingComputer. The chatbot was designed by the same person and is believed to have been building malicious AI tools for some time. The AI tool was first seen advertised on hacker forums, where the tool’s description stated that it was intended for scammers, hackers and spammers.
