JOIN USDH Deciphers | ChatGPT storms social media with its smart repartee
Barely a fortnight since its launch, the ChatGPT (Generative Pre-trained Transformer) chatbot has already become a sensation on social media platforms. Intrinsically very intuitive, this chatbot has distinguished itself with witty descriptive answers to most complex questions. DH’s KVN Rohit, draws attention to what makes this chatbot developed by the Artificial Intelligence research team - OpenAI - so attractive.
What is it about ChatGPT that has piqued the interest of people across the world?
Interactions with the current crop of generic chatbots remain very mundane. Typically, queries are listed in the form of a menu and the user can pick one that matches their specific query closest.The response is a near-accurate answer and, more often than not, fails to resolve the issue. In contrast ChatGPT engages with humans with smart repartees and has shown the propensity to offer a more accurate response than seen to date.
The most striking feature of ChatGPT is the creative answers it churns out.
For instance, when asked to suggest a horse name that is a superhero, it offered three names– Super Stallion, Captain Colt and Mighty Mustang.
Guess what it suggested for superhero cats? Captain Sharpclaw, Agent Fluffball, and The Incredible Feline.
There are several other witty ChatGPT answers shared on Twitter and people are blown away by the chatbot's intelligence.
Also Read | Google rivals join forces in online maps
Besides suggestions, ChatGPT can offer language translations (limited to a handful of options), grammar correction, software coding, debugging software codes, factual answers to general knowledge questions, writing an essay on famous people, offering solutions to complex mathematical problems, create fictional horror stories, summarise a text or an article and more. This has led many to believe ChatGPT may one day replace human writers or even Google search engines.
"ChatGPT is scary good. We are not far from dangerously strong AI," Elon Musk cautioned Sam Altman (CEO, Open AI) on Twitter.
How is ChatGPT being trained to understand natural language and respond accurately?
The Open AI team says it uses Reinforcement Learning from Human Feedback (RLHF) algorithm to build ChatGPT. "We trained an initial model using supervised fine-tuning: human AI trainers provided conversations in which they played both sides—the user and an AI assistant. We gave the trainers access to model-written suggestions to help them compose their responses," says the OpenAI team on its website.
Every ChatGPT's reply to the human query was tested and fine-tuned based on several levels of feedback, to ensure the AI chatbot's response is concise and cogent for any user to clearly understand.
Will ChatGPT bring an end to Google or any search engine?
Well, it is too early to say this, as Google and other search engine developer companies are not sitting idle. They too have their own AI/ML-based tech to improve their search engines and are in different stages of progress and are being deployed regularly to improve the user experience.
Also, OpenAI says ChatGPT is still a work in progress. The current version of the GPT-3.5 series was completed in early 2022 and the company has admitted that some of ChatGPT's responses, though they sound believable, lack logic. Further, sometimes it declines questions that it can answer correctly, if the user asks the same question differently.
It should be noted that some negative test results are based on complex questions, but chatGPT is known to offer incredibly accurate information with simple day-to-day facts, which people usually ask Google.
Is ChatGPT available to the public?
Yes, ChatGPT currently is available for free to the public, to test its capabilities. Users have to just sign up on the OpenAI website (https://chat.openai.com). However, there is a limit on the number of queries you can post on the ChatGPT playground (query box) and beyond the threshold, you have to pay to increase the limit.