Indian-origin American computer engineer Krishna C. Mukherjee has spent much of his career in the field of artificial intelligence – and he thinks the relationship between humans and AI is a symbiotic one. For Mukherjee, AI is a collaboration between humans and machines to achieve optimal results and that hasn’t changed since the time he started work on the new technology stream in the late 1980s. “Microsoft has been introducing AI into products like Word and Excel, and then into the cloud computing environments such as Azure for a long time. This edition of ChatGPT into [Bing] is just an extension of that,” the technology pioneer tells indianexpress.com in a call.
Born and brought up in India, Mukherjee was recruited by Microsoft in 1988 after he did his master’s in computer science and engineering from the Indian Institute of Technology, Kanpur. Mukherjee is known for laying the groundwork for the architecture and design of the Windows operating system and office applications at the Redmond giant. He and his team were also credited with building AI-based features in Microsoft Office like the spelling and grammar checks, AutoCorrect and AutoFormat features, and IntelliSense.
For years, big tech companies like Microsoft, Meta (formerly Facebook) and Google have been adding AI into their products, but never managing to demonstrate use cases of artificial intelligence at scale. However, OpenAI became the first startup to show the power of generative AI with the launch of the GPT-3 (Generative Pre-Trained Transformer 3) large language model in 2020. But that was just the beginning.
Last year, came the iPhone moment in the tech landscape when OpenAI introduced ChatGPT, a conversation bot that can be used to generate original essays, stories and song lyrics in response to user prompts. Two months after launch, ChatGPT became the fastest application to reach 100 million users. Microsoft’s collaboration with San Francisco-based OpenAI and the launch of the new Bing search engine and Edge browser with ChatGPT-like tech has once again shown artificial intelligence is changing our relationship with computers as we know it.
ChatGPT can create college essays, fictional stories, and even job resignation letters.
Mukherjee says the availability of data as well as improvements in both hardware and algorithms have led to important breakthroughs in AI in the past 20 years. “The more data is available, AI will become better as it learns from the data.”
“Algorithmic improvements have also helped to create this infrastructure, also called a neural network because it allows computers to simulate the data and train themselves utilising billions of parameters and trillions of tokens to assimilate all the data that is available on the internet and train themselves so that they can actually pretend to be a human being,” explained Mukherjee.
But the biggest question around AI is whether it will replace humans at jobs. “AI is nowhere close to replacing humans, but humans using AI in a smart way will definitely be going to replace humans who don’t use it,” Albert Meige, Global Technology Futurist at Arthur D. Little chimes in, adding that AI bots can help boost creativity if you know how to take advantage of them.
“Artificial intelligence is artificial, it’s not creative… but that’s not true anymore,” he says. Meige sees AI as a tool that will push creativity forward and not the other way around. “It’s the same story as when photography was invented. Painters thought that photographers are taking the easy path. In reality, however, the cameras are still the brain but a lot depends on the intention of the artist.”
Meige agrees that AI-based bots might be fast to respond, but they are not emotional, culturally sensitive or intuitive like humans.
It’s impressive how ChatGPT or AI-powered Bing can produce remarkable results, making searching faster and smoother, given their ability to converse like humans. But Meige warns against trusting these AI chatbots blindly as they still make factual mistakes. “[These] chatbots output words in a probabilistic manner…sometimes the answer is factually wrong.” No matter how advanced these chatbots become, they cannot replace a journalist yet. “When you are a journalist, fact-checking is one of your key skills and that is not yet implemented in AI tools like ChatGPT,” Meige said.
Meige, however, added that he expects AI will take over those roles that are mundane and do not require expert skills like drafting basic press releases or legal documents. “It will destroy jobs, but it will allow people who have lost their jobs to be more human and do more interesting things.”
Bard is Google’s take on OpenAI’s ChatGPT.
Mukherjee, too, is of the opinion that AI will not make human skills obsolete as many have been predicting for years. Rather, AI will replace certain tasks where data is manually fed and managed. He recalls how the foundation of the Intelligent Filing Manager, which combined logic programming and applications like Microsoft Excel, was to eliminate the paperwork at large corporations way before artificial intelligence became the buzzword.
In simple terms, AI is essentially mimicking a human brain; it imitates how humans think, understand and act. This type of computer intelligence is what many organisations need to streamline their businesses. “The primary use case of [Chatbots] hasn’t changed,” says Jaya Kishore Gollareddy CTO and Co-founder of Yellow AI.
Gollareddy said his customers want to automate collaborations to reduce support costs or increase their engagement with the customer. For such tasks, AI is perfectly suited. For example, think of how the airline industry can use AI to improve the experience of flying including day-to-day flight operations and addressing issues that arise with plane maintenance or predicting weather forecasts.
Humans and AI are like a group from the same team working towards a common goal. But like humans, AI is also not free from bias, after all, the data used to train the model has been human-generated. Misrepresentation of data, lack of diversity or not enough data could be the sources of bias in AI. If our training data ignores a certain demographic group or ethnicity, there is the risk that beliefs and biases can influence how the chatbot responds to the queries and deliver results based on them.