The world today is supposedly in the middle of another technological leap. ChatGPT-3, released by OpenAI, the Artificial Intelligence Company in California, took the world by storm as tech enthusiasts have been raving about the potential use cases of this revolutionary AI bot. It is said to produce AI-generated, insanely accurate search results, write job application letters, code, stories, and essays, and answer questions about almost everything under the sun. Users say it’s an extremely useful chatbot, producing mind-boggling results that are often hard to differentiate from the work of humans. The AI bot has got everyone marveling, with entrepreneurs willing to dive deep and invest in companies eager to harness its fullest potential.
ChatGPT, an extraordinary hit, has already garnered a massive fan following, drawing more than one million users within a few days of its launch since November 2022, sparking experimentation by the public. Powered by a staggering amount of data and computing techniques, the chatbot can devise sophisticated responses, meaningfully string words together and tap into an ocean of information and vocabulary by mimicking speech patterns and disseminating encyclopedic knowledge. Recognized as a large language model or LLM, the chatbot is fed a mammoth amount of data in the form of conversations, articles, and books which are used to build models based on the probability of the possible words or sentences that would follow texts that came before. These predictive texts have been massively scaled, which allows them to generate complete sentences instead of single, isolated words.
However, the success of ChatGPT isn’t without some form of human intervention. It has received training from human AI trainers, incorporating a plethora of questions and answers into its dataset. Programs were run to produce different responses and human experts ranked in the order of best to worst. Such fine-tuning has helped the chatbot frame highly impressive information as naturally as possible. Some claim that ChatGPT is also more difficult to corrupt than other chatbots. It has also been programmed to decline inappropriate questions and avoid making up answers it is not trained to respond accurately.
But even though ChatGPT holds the ability to solve the most complex mathematical problems, write a song or poem on command or explain quantum physics, a full takeover by the generative AI is a far cry. Using the chatbot has also invited trouble for many companies. Koko, a mental health company, was in the direct line of fire after its founder, Rob Morris declared how they used GPT-3 to respond to users, using AI to “help craft” swifter and smarter responses as an experiment. While tools like these hold massive potential, experts believe it’s a long haul before we finally use them at their optimum levels. Human intervention is still a critical part of the AI’s journey, and all said and done, it’s not magic after all! ChatGPt is also clueless about what’s true or false. Some do state that even though the answers from Chat GPT seem fluent and confident, they are sometimes spectacularly wrong. So, much care needs to be taken before buying whatever the AI churns out. Among its other fundamental limitations are that it does not know about happenings post 2021, as no new information was fed or updated on its system post 2021.
Despite all the hype, ChatGPT is still at its nascent stage. And as someone rightly pointed out, it is undoubtedly transformative but very much a work in progress.