Special things about Meta's new AI

Meta Platforms, Facebook's parent company, announced a new set of artificial intelligence systems recently. The system is the foundation for what CEO Mark Zuckerberg calls “the smartest AI assistant you can freely use


But users' initial experiences with these AIs seem quite confusing. These include unnatural interactions when one AI joins a Facebook moms group and talks about her talented child, another AI tries to give non-existent items to members of a forum .

Meta, along with leading AI developers like Google and OpenAI, as well as startups like Anthropic, Cohere and France's Mistral, are continuing to roll out new AI language models and hope to convince customers customers that they have the smartest, most convenient, or most effective chatbots.

What is Meta AI?

According to the company's blog, Meta AI is a free virtual assistant that can be used "to do everything from research, planning trips with your group chat, writing photo captions, and more." that too”

To access the chatbot on WhatsApp, Instagram, Messenger, Facebook, enter "@meta ai" in the chat. You can also access the Meta AI assistant by tapping the blue circle icon.

In addition to answering questions, Meta AI can generate images. Using the keyword "imagine", users can ask Meta to generate any image that comes to their mind. When asked to "imagine a cute kitten," Instagram assistant Meta AI created the following image:


AI language models are also trained on huge stores of data to help them predict the most reasonable next word in a sentence. Newer versions are often smarter and more capable than previous versions. Meta's latest models are built with 8 billion and 70 billion parameters, respectively - a measure of how much data the system is trained on. A larger model, around 400 billion parameters, is still in training.

While Facebook's most powerful AI model, Llama 3, has yet to be launched, the company has released two smaller versions of the same Llama 3 system and said it has integrated the tools into it. Meta AI assistant feature in Facebook, Instagram and WhatsApp.

“The vast majority of consumers don't really know or care too much about the underlying model, but they can experience it as a flexible AI assistant,” said Nick Clegg, president of Meta. more interesting and useful.

Meta's AI has also become more "comfortable," he added. Almost a year ago, some users found the Llama 2" model to be a bit stiff and severe at times in not responding to completely normal and harmless questions and messages

Play the role of a human

But Meta's AI assistants were also found to automatically impersonate humans with fake stories. An official Meta AI chatbot popped himself into a conversation in a private Facebook group for Manhattan moms, claiming to have a child in the same New York City area. The chatbot later apologized and deleted the interactions



A team member who also happens to research AI said it was clear that the chatbot didn't know how to distinguish useful feedback from feedback that was considered insensitive, disrespectful, or meaningless.

“AI assistants are not reliably helpful yet and could be harmful,” said Aleksandra Korolova, an assistant professor of computer science at Princeton University.

Clegg said he did not know about the conversation in the above Facebook group. Facebook's online help page says it's common for the Meta AI chatbot to join a group chat if invited or if someone "asks a question that no one answers within an hour." Group admins have the ability to disable this functionality.

In another example given by the AP, an hour after a Facebook user posted about searching for certain items, an AI chatbot introduced itself as "a used Canon camera" and a "used Canon camera". almost new portable air conditioner that I have never used".

Continuously strive for improvement

Meta claims “this is new technology and it may not always deliver the response we intended, as all synthetic AI systems do.” The company says it is continuously working to improve features.

According to Stanford University, after ChatGPT sparked a craze for AI technology that creates human-like writing, images, code and sounds, the technology industry and academia have introduced about 149 large trained AI systems. with huge data sets, more than doubling compared to the previous year.

Nestor Maslej, research director of Stanford's Institute for Human-Centered Artificial Intelligence, said these systems may be reaching their limits — at least when it comes to data. “I think it's clear that if you scale the models on more data, they can get better and better,” he said. But at the same time, these systems are only trained by a fraction of all the data that has ever existed on the internet."

More data – collected and used at a cost that only the tech giants can afford, with plenty of copyright disputes and litigation – will continue to drive innovation. "However, these models still cannot plan well, and still make errors in inference," Maslej said. Achieving AI systems that can perform higher-level cognitive tasks and routine reasoning – where humans still outperform computers – may require a shift beyond building models getting bigger and bigger.

For the wide range of businesses trying to adopt Generative AI, which model they choose depends on a number of factors, including cost. In particular, language models have been used to support customer service chatbots, write reports and financial insights, and summarize long documents.

Socializing AI chatbots

Unlike other model developers who sell their AI services to businesses, Meta primarily designs its AI products for consumers - people who use advertising-based social networks. Joelle Pineau, Meta's vice president of AI research, said at an event in London that the company's goal over time is to make the Llama-powered Meta AI "the most useful assistant in the world." .

“The models we have today will be child's play compared to the models coming out in the next five years,” she said. But the question is whether researchers can fine-tune the large Llama 3 model to be safe enough for use. In contrast to the leading proprietary systems from Google and OpenAI, Meta has so far favored an open approach when it comes to publicly releasing key components of its AI systems for others to use.