OpenAI, the research and development giant behind the hugely popular AI platform ChatGPT, says it is looking forward to investing in reducing ‘reflective and implicit biases’ in how AI responds to different inputs.
Source: The Economic Times
The Indian government has on several occasions indicated that the country will develop a framework to prevent the misuse of AI, and the algorithm has come under international scrutiny to ensure accountability. A few weeks ago, the Department of Telecommunications released draft standards to ensure the fairness of artificial intelligence. A small team at the Ministry of Electronics and IT (MeitY) called Bhashini is currently developing a chatbot for WhatsApp that uses data from ChatGPT. Moreover, some people, especially farmers who live in rural areas, do not always want to write down their questions, the chatbot also allows voice notes.
Basically, voice notes can be used to make a request to a chatbot, which will respond based on a voice generated by ChatGPT.
“We are investing in research and engineering to reduce both explicit and implicit biases in how ChatGPT responds to different inputs,” OpenAI said in a February 16 blog post. In some cases, ChatGPT will reject output that it shouldn’t, and in some cases, it can’t reject it when it should. We believe that improvement is possible on both sides.
OpenAI also said that they are working on an update to ChatGPT that will allow users to easily customize the feature as per their convenience. “This means allowing system results that other people (including us) strongly disagree with,” OpenAI said.
Risks of Customization “It’s important to ensure that there is enough balance in AI as it reinforces people’s existing beliefs about our technology and sycophantic AI,” he said.
OpenAI is also considering public reviews to make decisions on ChatGPT’s hard limits.
He added, “As a starting point, we sought external input on our technology during the Red Team integration. We also recently started asking for public feedback on AI in education (an important context in which our technology is being developed).
ChatGPT’s algorithm is initially ‘pre-trained’, in which the model learns to predict the next word in a sentence. To do this, the model is exposed to the vast literature on the Internet. Next, open the model AI “fine-tunes” to fine-tune system behavior.
