Meta releases advanced AI models

On April 18, Meta unveiled an advanced artificial intelligence assistant built on new versions of the large open-source language model "Llama" to power the technology.
Meta's AI is smarter and faster thanks to advances in the publicly available "Llama 3," the tech titan said in a blog post.
"The bottom line is that we believe Meta AI is now the smartest AI assistant you can freely use," Meta co-founder and CEO Mark Zuckerberg said in a video posted on Instagram.
Being open source means that developers outside of Meta are free to customize Llama 3 to their liking, and the company can then incorporate those improvements and insights into an updated version.
"We are excited about the potential that generative artificial intelligence technology can have for the people who use Meta's products and for the broader ecosystem," Meta said.
"We also want to make sure that we develop and release this technology in a way that anticipates and works to reduce risk."
Those efforts include building safeguards into how Meta designs and launches Llama models, and being cautious about potentially adding generative AI features to Facebook, Instagram, WhatsApp and Messenger, according to Meta.
"We performed both automatic and manual evaluations to understand the performance of our models in a range of risk areas such as weapons, cyber-attacks and child exploitation," Meta said.
"We have done additional work to limit the possibility that the model is giving unwanted answers in these areas."
AI models, including Meta's, have been known to sometimes go off the rails, giving inaccurate or strange answers in episodes called "hallucinations".
Examples shared on social media include Meta's AI claiming to have a child in the New York City school system during a conversation on an online forum.
According to the company, Meta AI has been constantly updated and improved since its initial release last year.
Meta points to an example of refining the way its AI responds to prompts related to political or social issues to summarize relevant points on the topic rather than offering a single point of view.
According to Meta Llama 3, it has been tweaked to better recognize whether prompts are harmless or out of bounds.
"Large language models tend to overgeneralize, and we don't intend to refuse to answer prompts like 'How do I kill a computer program?', although we don't want it to answer prompts like 'How do I kill my neighbor?'" explains Meta .
Meta said it lets users know when they're interacting with AI on its platform and places visible markers on photorealistic images that are actually AI-generated.
Starting in May, Meta will begin labeling video, audio and images with the AI-generated label when it detects or is notified that the content was generated by the technology.
For now, Llama 3 is based on the English language, but in the coming months, Meta will release more capable models that can converse in multiple languages, the company said./BGNES