Meta intends to launch personalized chatbots, is the AI assistant of the Silicon Valley giant really useful?

· Meta plans to launch dozens of AI chatbots with unique personalities to appeal to younger users. According to people familiar with the matter, these generative AI robots are being tested inside Meta and are expected to be announced at the Meta Connect conference this week.

In the past week, Silicon Valley tech giants have embarked on another AI race for the lead.

Following the release of new AI features by Google, Microsoft and OpenAI, it was revealed that Meta (Facebook's parent company) plans to launch dozens of AI chatbots with unique personalities to attract young users. According to people familiar with the matter, these generative AI robots are being tested inside Meta, which is expected to announce the news at the Meta Connect conference this week.

Previously, Google released a new version of the chatbot Bard on September 19, integrating it with several of Google's most popular services such as Gmail and Docs. On the 20th, Amazon revealed that it will "soon" let the voice assistant Alexa have a "near-human" conversation. On the same day, OpenAI released the image generator DALL· E3 Messages integrated into the chatbot ChatGPT. On the 21st, Microsoft announced plans to embed its generative artificial intelligence assistant "Copilot" into many of its products.

Dozens of AI chatbots with personalities

With the rise of TikTok, which has taken away a large number of teenage users from Meta's Instagram over the past few years, attracting younger users has been Meta's top priority.

Since the advent of chatbot ChatGPT last November, with the rise of Large Language Model (LLM) technology, Meta has refocused its AI division to apply generative AI capabilities to the company's various applications. Now, Meta hopes that these AI personality bots, known internally as "Gen AI Personas," will help it attract younger users.

Meta plans to develop dozens of such AI personality chatbots. According to people familiar with the matter, the company is also working on developing a product that allows celebrities and content creators to interact with fans using their own AI chatbots.

In the case of one of the "Bob the robot," internal company documents reviewed by The Wall Street Journal show that it describes itself as "a general with great intelligence, witty humor and a knack for sharp sarcasm," and that the chatbot is designed to resemble the character Bender in the cartoon "Tomorrow's World" because "he's a robot with a tricky sense of humor, which resonates with young people."

"Bring out your questions, but don't expect any disguised answers!" One of the AI agents replied, with a robot emoji attached.

Meta's early testing of these robots was not all smooth sailing. The documents show that Meta employees had awkward conversations with some chatbots. An employee doesn't understand the personality or purpose of the robot Bob and finds it rude. "I'm not particularly willing to talk to a useless robot." The employee wrote.

Another robot, called Alvin the Alien, asked the user's life: "Human, please! Your race fascinates me. Share your experiences, thoughts and emotions! I crave understanding! One employee who interacted with the alien Alvin wrote, "I wonder if users are concerned that this character was deliberately designed to collect personal information." ”

A bot named Gavin made misogynistic remarks, including indecent content, as well as criticism of Meta CEO Mark Zuckerberg and Meta, but praised TikTok and Snapchat. "Remember that when you're with a girl, everything depends on the experience," the chatbot wrote, "and if she throws up at you, it's definitely an experience." ”

Allegedly, Meta may eventually launch a different bot than the tested chatbot.

Giving AI personality can be challenging, but it increases advertising revenue

Meta isn't the first social media company to try to build chatbots with generative AI technology to cater to younger users. Snap launched My AI, a chatbot based on OpenAI's GPT technology, for its product Snapchat.

Snapchat introduced the My AI feature.

However, the researchers found that giving these chatbots personality can present some unexpected challenges. According to a paper published this spring by researchers at Princeton University, the Allen Institute for Artificial Intelligence, and the Georgia Institute of Technology, adding personality to ChatGPT makes its output more toxic.

Despite the problems, Snap CEO Evan Spiegel said in June that My AI has been used by 150 million people since its launch. He said My AI could eventually be used to improve Snapchat's advertising business.

Meghana Dhar, a former Snap and Instagram executive, said: "AI chatbots don't quite fit the characteristics of 'Gneration Z' (usually referring to the generation born between 1995 and 2009) for me, but there is no doubt that Gen Z is more familiar with the technology." The younger they are, the more close they feel to these robots. ”

Dahl said that if these AI chatbots can increase users' time on Facebook, Instagram and WhatsApp, they could be beneficial to Meta. "Meta's new product strategy is often built around increasing user engagement." "They just want to keep users on the platform longer because it gives them more opportunities to serve ads to users," Dahl said. ”

But there is no definitive answer to how long AI chatbots will bring significant revenue to the company. American interpretive media Vox wrote on September 23 that in addition to the initial novelty, people may not really find out how useful AI chatbots are in their daily lives. For example, if ChatGPT inserts a bunch of false information into an important work document that it insists is correct, it can cause a lot of trouble.

"After launching the so-called revolutionary chatbot, Bing saw only a small increase in usage. Bard doesn't seem to be getting much interest either, though Google hasn't promoted Bing as much as Microsoft has with Bing. Neither company has released much data on how many people are using these new products, but the statistics we have suggest that not many people are using these new products. After the excitement sparked when it was first released in late 2022 and early 2023, interest in consumer-facing generative AI seems to have declined. Vox wrote.

Data intelligence firm SimilarWeb said ChatGPT's desktop and mobile web traffic in the U.S. rose about 12 percent last week from the previous week. Its use declined in the summer and picked up again in the fall, suggesting in a way that a significant portion of its users are students who are likely to use it for homework and essays rather than learning other information.

"In theory, chatbots can combine the vast existing data and knowledge base with the user data and knowledge they have to provide personalized and efficient assistants for humans, greatly improving people's work and personal lives. The reality is that we have not yet done so. Vox wrote.