With the development of chatbots that do not have any guardrails restricting outputs, new uses, along with risks and dangers, arise.
Guardrails and filters are frequently added to chatbots by big companies to prevent them from giving misinformation or making potentially harmful claims. For example, a chatbot might refuse to answer a question about how to commit suicide or regenerate responses supporting conspiracies. Without filters, AI can and will answer all questions that they are given, including those relating to harmful topics such as drug use and suicide. Unmoderated outputs can also give false and wrong claims, which spreads misinformation among users.
Despite the benefits of filters on AI, independent programmers are developing their own chatbots with little to no regulations on the outputs. Some users dislike the many restrictions placed on responses. The developer Eric Hartford was annoyed when ChatGPT refused to answer some of his questions, citing ethical concerns. “If I ask my model a question, I want an answer, I do not want it arguing with me,” Hartford explains. In response to the guardrails on OpenAI, Hartford began working on WizardLM-Uncensored, his own unfiltered chatbot.
Many other new chatbots have been created in recent months. The models were usually developed by independent programmers or groups of volunteers. Most of the developers tweaked existing chatbots instead of starting from scratch.
Chatbots without guardrails and filters can be used in ways that normal, regulated ones can not. For example, users can train models to help with private documents or personal information without the results being checked and scanned by the companies training the AI. Programmers can also develop new features for the AI faster, as they do not have to worry about any compatibility issues with filters.
Many developers feel that unrestrained chatbots are an inevitable result of AI. “This is going to happen in the same way the printing press was going to be released and the car was going to be invented,” said Hartford. “Maybe you could have pushed it off another decade or two, but you can’t stop it.” Only time will tell whether the pros outweigh the cons of freely speaking AI.
Source:
https://www.nytimes.com/2023/07/02/technology/ai-chatbots-misinformation-free-speech.html
Guardrails and filters are frequently added to chatbots by big companies to prevent them from giving misinformation or making potentially harmful claims. For example, a chatbot might refuse to answer a question about how to commit suicide or regenerate responses supporting conspiracies. Without filters, AI can and will answer all questions that they are given, including those relating to harmful topics such as drug use and suicide. Unmoderated outputs can also give false and wrong claims, which spreads misinformation among users.
Despite the benefits of filters on AI, independent programmers are developing their own chatbots with little to no regulations on the outputs. Some users dislike the many restrictions placed on responses. The developer Eric Hartford was annoyed when ChatGPT refused to answer some of his questions, citing ethical concerns. “If I ask my model a question, I want an answer, I do not want it arguing with me,” Hartford explains. In response to the guardrails on OpenAI, Hartford began working on WizardLM-Uncensored, his own unfiltered chatbot.
Many other new chatbots have been created in recent months. The models were usually developed by independent programmers or groups of volunteers. Most of the developers tweaked existing chatbots instead of starting from scratch.
Chatbots without guardrails and filters can be used in ways that normal, regulated ones can not. For example, users can train models to help with private documents or personal information without the results being checked and scanned by the companies training the AI. Programmers can also develop new features for the AI faster, as they do not have to worry about any compatibility issues with filters.
Many developers feel that unrestrained chatbots are an inevitable result of AI. “This is going to happen in the same way the printing press was going to be released and the car was going to be invented,” said Hartford. “Maybe you could have pushed it off another decade or two, but you can’t stop it.” Only time will tell whether the pros outweigh the cons of freely speaking AI.
Source:
https://www.nytimes.com/2023/07/02/technology/ai-chatbots-misinformation-free-speech.html
