Advertisment

ChatGPT: Our Quest for Computer Conversation

ChatGPT can perform semi-complex coding tasks, and could solve up to level 3 of Google’s coding challenges with some iterative prompting

author-image
DQINDIA Online
New Update
ChatGPT

ChatGPT is making a splash, with a flurry of both positive and negative press. It has invited extensive conversation on the internet about the possible risks. However, the gains of ChatGPT are undeniable, especially after Microsoft’s investment in its parent company OpenAI. But before we talk about ChatGPT, we will briefly discuss the history and origins of chatbots.

Advertisment

The story of chatbots begins in the 1950s with Alan Turing designing the Turing Test - a test designed to see if machines could ever display human-like intelligence, and it involved a human evaluator chatting with both a human and a computer participant. The computer would only pass the Turing Test when the evaluator could not identify it as different from the human participant. Computer scientists have been trying to build a chatbot that passes the Turing test for over fifty years. One of the earliest chatbots was designed in the 1960s by Joseph Weizenbaum, and was a psychotherapist chatbot called ELIZA. Originally displayed to show the limitations of AI, many people found Eliza pretty convincing as a therapist. The fundamental principle that ELIZA used was keyword recognition, and asking questions based on a single keyword. 

ELIZA, and other chatbots in the subsequent fifty years, paved the way for the many chatbots in our everyday life - including customer service chatbots. They allowed for the creation of conversational AI. The most popular of these are virtual assistants - like Amazon’s Alexa or Apple’s Siri. The underlying principle of these virtual assistants is similar to that of a text based chatbot - keyword recognition - even as they are much more natural in the way they respond to queries.

However, the big limitation of these chatbots is the inability to carry conversational context forward. They can respond to immediate requests but cannot use past information to construct logical arguments. They search the internet for the closest possible direct response to the question, and often have several pre-programmed features that mimic human conversation, but they cannot construct new work based on context clues.

Advertisment

Also read: How ChatGPT is affecting industry dynamics

Here is where ChatGPT comes in. Designed by the company OpenAI, ChatGPT is technologically very easy to use, requiring no knowledge of code. This makes it a highly democratic tool. Further, the software is always learning and improving the responses, with every customer interaction. This makes it tremendously useful in a wide range of white-collar work - both routine and creative. This makes ChatGPT a potentially transformative intervention in organizations. 

 Some of the many things ChatGPT can do include:

Advertisment
  • Legal Work: ChatGPT can draft legal documents and contracts. However, these documents will need review to ensure they are compliant with country specific legal requirements.
  • Coding: ChatGPT can perform semi-complex coding tasks, and could solve up to level 3 of Google’s coding challenges with some iterative prompting. Further, ChatGPT can explain, comment on and document existing code that is pasted into the interface. This can be invaluable for new developers in understanding what legacy code does. But ChatGPT may not be able to write perfectly usable code from scratch.
  • Sales and Marketing Collateral: ChatGPT can help in the creation of emails, press releases, social media posts, slide decks, SEO keywords and more, and is likely to significantly affect the sales and marketing aspects of businesses.

This list is far from comprehensive and is evolving day to day. Other fields that are likely to experience disruption are management consulting and learning and development.

At the time of writing, ChatGPT has crossed a hundred million users in two months, making it the fastest application to do so. Every user represents new ways to engage and interact with ChatGPT, and in a work context, there are new use cases being identified and pursued every day. However, ChatGPT is far from perfect. In most of the cases listed above, ChatGPT continues to requires human guidance. It can only provide an outline for the code or the contract, it cannot provide you a final version. There are inaccuracies and ambiguities that the system cannot solve at this point, even as it gets better at learning everyday. In organizational contexts, there is also the challenge of recency of knowledge. ChatGPT has no access to copyrighted data, or intellectual property that is not in the public domain. Further, the question of who owns the content generated by ChatGPT is still under discussion. This is not including questions of ethical or regulatory concerns, which are currently the topic of debate particularly in educational institutions. 

Advertisment

Given these grey areas, it is unlikely that ChatGPT will replace human workers in entirety. It does however appear that most people will have to change the way they work. To fully harness the power of ChatGPT, and other AI interventions like it, will require a significant overhaul in our approach to work. The most critical aspect of working with ChatGPT will be how to generate the ideal prompt to yield best results. This field is now being called “prompt engineering” and is going to be integral across business functions. It is likely to become a critical skill in the future. 

Organizations, including ours, will have to think about how much they will engage with ChatGPT, how they might roll out these interventions, as well as guidelines for its use in the workplace. Since it is also such a new technology, the ways in which it will evolve and grow remain to be seen. We are keen to understand the ways it will affect many of our verticals, and it is important to us that we embrace technology interventions that could have long term repercussions and potential for disruption.

The article has been written by Roshan Shetty, Chief Revenue Officer, Sonata Software

Advertisment