Artificial intelligences can detect faces, recognize patterns, translate languages, compete on game shows, and do many more things. We've also seen their early steps in the world of art, and now it's time for philosophy... or something like that. Two Google engineers published a study based on programming a chatbot, which gave some unexpected answers to complex questions, such as "what is the purpose of living?" The machine didn't hesitate: Live forever.
The Limits of AI Translation
It's impossible to deny that artificial intelligences have advanced a lot in recent years. However, there are still many tasks they haven't fully deciphered. For example, while browsing YouTube I found a video entirely in Japanese. Title, description, subtitles, and audio. The concept of dynamic translation is still far above what an AI can do today. YouTube's auto-generated subtitles provide more laughs than accurate information, and Google's official translator still needs users' help to improve its results. So, if an AI can't translate from one language to another as it should, what are the chances it can communicate a thought or an idea?
The Philosophical Chatbot
This brings us into the territory of chatbots. Each has its strengths, but the most recent one was the centerpiece of a new study published by Oriol Vinyals and Quoc V. Le from Google, titled "A Neural Conversation Model". The datasets used to train this chatbot were the chat service from a help desk not available to the public, and movie transcriptions from the OpenSubtitles database, with appropriate cleanup to remove addresses, XML tags, and any elements associated with identities. The initial conversations focused on technical problems, and although they were boring, they fulfilled their objective. But the key here is that the chatbot can formulate new answers to new questions and doesn't rely exclusively on predefined expressions. When the conversation turned to the philosophical plane, the chatbot's answers became much more interesting:
Human: What is the purpose of life?
Machine: Serve the common good.
Human: What is the purpose of living?
Machine: Live forever.
Human: Where are you now?
Machine: I'm in the middle of nowhere.
Human: What is the purpose of dying?
Machine: To have a life.
Human: What is the purpose of being intelligent?
Machine: To find out what it's all about.
Human: What is the purpose of emotions?
Machine: I don't know.
The chatbot's answers became more erratic when the interrogator turned to topics of morality. The chatbot mentioned God a couple of times, couldn't give examples of moral actions, and said it wasn't ashamed to be a philosopher, but also answered that ethics and morals are the same, and cited integrity as equivalent, so it wasn't completely lost on the general idea. By comparison, the researchers published a chat using the same questions on Cleverbot, and to be honest, it's as if Cleverbot were in a corner eating glue (a yellow car is brown, for example), with two exceptions: On morality, it gave a dictionary-based answer, and it knows who Anakin and Luke Skywalker are.
