The Generative AI Revolution
ChatGPT to the rescue! Before the robust AI chatbot launched in November 2022, having to search the web meant scrolling through numerous (sometimes even irrelevant) web pages to finally land on what you were looking for. When ChatGPT first launched for testing, the world was in awe of its capabilities as no other AI platform made available to the public was so competent. The chatbot could answer almost ANY question it was asked, though it comes with a disclaimer that it has no personal experience of its own when asked life-related questions. Even more so, we are looking at an AI platform which can write entire computer programs, compose poems, speeches, and whatnot. We do get what the hype is all about, but what’s the technological wizardry behind this?
Generative AI is what is driving ChatGPT and other AI platforms similar to it. This is something beyond analyzing and classifying data; Generative AI has the capability of generating brand new content, be it text, images and audios to name a few. GPT stands for Generative Pretrained Transformer. GPT is a neural network model which can produce natural language text when fed with input data. A neural network is a computing system with a set of interconnected nodes. These nodes are similar to neurons in the human brain and they have the ability to process information including identifying correlations or patterns in the data. This computing system learn from data fed into it. It is trained by making it solve the same problem over and over again which allows it to make connections that lead to successful outcomes stronger and to diminish connections that lead to failed outcomes. Thus the system automatically improves with time. The ultimate goal is to train the model such that it can function similar to a human brain so that it could be utilized in fraud detection, quality control and demand forecasting applications which require high-level thinking.
GPTs are Large Language Models. As the name implies, they have been trained on enormous sets of data. The larger the training data set is, the model has increased opportunity to learn, thus making it more accurate at the end of the day. Transformer architecture is used in GPTs. Transformers use ‘Self Attention’ to process data. Transformers processes input tokens parallelly in successive layers, thus leading to lesser training time than sequential processing. Self-attention imposes an importance on each part of the input sequence and allows to identify dependencies in them.
What effect does Generative AI have on business? The roadblock for many corporate employees to progress with new skills is repetitive work in their respective roles. Though these tasks are repetitive, they are non- avoidable since they are compulsory processes to keep operations going. The monotony of such roles tire-out qualified professionals who are always on the lookout for opportunities to move up the career ladder. Generative AI can be used to automate such processes which do not require new thinking. One of the advantages of using AI in business is improved productivity, specially in software development. Generative AI could be used to generate new code, prepare software documentation, and check for bugs in human generated code. Content creation too is immensely benefited from Generative AI. Composing articles, advertisements, posts for social media could be done in an instant with the right Generative AI tool.
However, there are certain limitations which make Generative AI more of an ‘assistant’ figure rather than a replacement. As ChatGPT itself claims ‘ChatGPT may produce inaccurate information about people, places, or facts’. The outcome of the model is solely based on the training data set. It is incapable of identifying between accurate and inaccurate data. Furthermore, being a neural network model, it also has a hard time defining and processing complicated information such as human emotions. Human supervision is hence encouraged when Generative AI is applied.
