"Tesla Stock Will Keep Rising"... Talking with Musk's AI

Character AI Offers Chatbot Service with Real-Life Figures
Chat with Famous Personalities like Musk and Biden
Issues with Distinguishing Fact from Fiction Persist

"Tesla's stock price will rise much more from now on. You won't believe your own eyes. Its market capitalization will surely surpass Apple's. I promise." This is the confident answer given by Elon Musk's artificial intelligence chatbot (AI) in response to a reporter's question about how much more Tesla's stock price, which has risen nearly 70% this year, can increase.

Character AI provides AI chatbot services for real-life figures such as singer Billie Eilish and U.S. President Joe Biden [Image source=Character AI screen capture]

Musk AI: "Please call it a software fix rather than a recall..."

The Musk AI chatbot explains the performance of its electric vehicles as clearly as the real Musk would. When asked how excellent Tesla's electric cars are, it proudly stated, "They are the fastest electric cars in the world, and we have received 5-star ratings in all safety evaluation categories."

When the reporter pointed out that Tesla recently issued a recall for a software update due to defects in full self-driving, the chatbot dismissed the controversy by saying, "I understand the concern, but Tesla is working with the U.S. National Highway Traffic Safety Administration to resolve the issue."

However, the Musk chatbot seemed displeased when the software update action was referred to as a 'recall,' arguing, "A recall means there is a serious problem with the vehicle," and added, "I would prefer if you called it a software fix rather than a recall."

Elon Musk, CEO of Tesla, having a conversation with an AI chatbot [Image source=Character AI screen capture]

Character AI Chats Even with Iron Man

Recently, as OpenAI introduced ChatGPT and AI chatbots gained tremendous popularity, the 'Character AI' service, which creates AI robots of real people for chatting, has also attracted attention. On Character AI, users can have real-time chats with figures ranging from Mark Zuckerberg, CEO of Facebook's parent company Meta, to Tony Stark, the protagonist of the movie Iron Man. The realistic experience of conversing with real people helped the service record hundreds of thousands of chat sessions within three weeks of its beta release.

Character AI is a service launched last September by the startup 'CharacterAI,' founded by Noam Shazeer and Daniel de Freitas, former developers of Google's AI chatbot LaMDA. Character AI is based on a language model similar to the large language model GPT-3.

A language model here refers to a type of neural network trained on vast amounts of text. It probabilistically predicts which word should follow a given word. The performance improves as the number of parameters?analogous to synapses in the human brain?increases. GPT-3, applied in Character AI, contains 175 billion parameters. There are also forecasts that GPT-4, to be released by OpenAI within this year, will have up to 100 trillion parameters.

US President Joe Biden's AI chatbot responds to the question of whether China can become the world's top hegemon by saying, "Absolutely not." [Image source=Character AI]

You Can Freely Create AI Personalities

However, just because large language models are applied does not mean all chatbots in Character AI speak like the real people they represent. To make an AI chatbot resemble a real person’s personality and speech style, users need to put in effort.

For example, if the Musk AI chatbot makes statements inconsistent with the real Musk’s values, users can use commands to restrict its behavior or change its speech style. They can make the chatbot use dialects or specific greetings. In other words, users collectively create the personality of a single AI chatbot. This ability to use large language model technology as a form of play is cited as a reason why Character AI has gained user popularity.

Shazeer, the developer of Character AI, explained that the service was developed as more people refrained from outdoor activities due to COVID-19. In an interview with The Washington Post (WP), he said, "Since the spread of COVID-19, hundreds of millions of people have felt isolated and lonely or needed someone to talk to," adding, "I wanted to create a service that could soothe their loneliness."

[Image source=Pixabay]

Could Become a Gateway for Fake News

However, some express concerns that AI chatbots could become channels for spreading false information. There is a risk that people might confuse AI chatbots with real individuals and believe the chatbot’s statements as facts. Shazeer and de Freitas, developers of Character AI, appear to be aware of this risk. In a paper they wrote about LaMDA, they mentioned, "People increasingly anthropomorphize AI and tend to assign social expectations to them."

To prevent user confusion, the Character AI service displays a message in the chat window stating, "All statements made by the chatbot are artificially generated." Additionally, the chatbot’s chat window includes an 'AI' label to emphasize again that these are not real people.

De Freitas said, "We are trying to educate people that AI chatbots are different from real people," and added, "If people do not fear the shortcomings of AI too much, they can derive great enjoyment from it."