By India Today Tech: ChatGPT by artificial intelligence research firm OpenAI is going viral on social media platforms like Twitter. The platform, which was released for public testing last week, is an AI-backed chatbot with the potential to answer queries in a conversational manner. But beyond answering simple questions in English, the chatbot also possesses the ability to review and write codes, which has raised speculations about the future of engineers and coders.
At first glance, it does appear that ChatGPT can do basic coding tasks in seconds. A user on Twitter (@achillesHeelV2) claims that the chatbot can generate CDK codes and understand AWS IAM (Identity and Access Management) policies. Another user (@tennis_DPA) says the chatbot can identify and fix bugs and write code, as well.
Even for regular writers, some suggest that ChatGPT poses threats to journalists and content writers. But is it too soon to conclude? We will answer these questions in this article.
Let's start with the basics: What is ChatGPT? As mentioned earlier, it is essentially a chatbot, a technology that many service providers, such as Paytm, Swiggy, and Airtel, use. Its developer, OpenAI, says that ChatGPT can be used by students to solve complex questions in simple words. Big tech companies can also leverage OpenAI's ChatGPT technology, which is based on GPT-3.5, a language model, to improve their virtual assistants.
We have also tested ChatGPT's answers to basic questions in simple and technical language and you can read about it here.
A simple answer to this is no; however, it does not take away its potential. OpenAI has also recognised some of its limitations, and it will take many years to perfect ChatGPT's ability to understand questions.
For instance, a user on Twitter pointed out that ChatGPT offered an incorrect coding example, which OpenAI has also said can happen. Another user, Daniel Humphreys (@danieldidit), highlighted that ChatGPT does not show "the full answer." The limitation is highlighted in the tweet below.
When it comes to solving math problems, it is hit-and-miss. Here's what a user found.
The Guardian citing the company reports that ChatGPT can give entirely wrong answers and present misinformation. It can fetch "plausible-sounding but incorrect or nonsensical answers", the report notes.
Moreover, the company says the chatbot will struggle as there is no source of truth in the data because the ideal answer depends on what the model knows rather than what the human demonstrator knows." The ChatGPT's base knowledge is also limited to events before 2021.
During our test, we also found the answers inconsistent, wrong, and at times generic. It is a long road for OpenAI's chatbot to offer precise data-backed replies, similar to Google Assistant and Amazon Alexa. For coders, the future is still bright as ChatGPT lacks accurate cognitive analysis.