Picture: Mariki Uitenweerde / University of Pretoria / via ANA File – UP launches SCU-B student counselling chatbot to boost student well-being. The revolutionary AI-powered ‘conversational chatbot’ marks a significant breakthrough in processing information, the writer says.
By Naledi Ramontja
AI chatbots have garnered widespread attention, with OpenAI’s ChatGPT leading the way when it was launched in November 2022. This revolutionary AI-powered ‘conversational chatbot’ marked a significant breakthrough in processing information. Following this, Microsoft introduced Bing, and Google introduced Bard chatbots, both powered by Large Language Models (LLMs), setting the stage for advanced conversational applications.
These chatbots are known for their ability to perform diverse functions across various sectors, providing coherent and human-like responses instantly, setting them apart from traditional search engines like Google and Wikipedia.
In terms of datasets, Bard boasts 1.37 billion parameters and a massive vocabulary base of over one and a half trillion words, while ChatGPT features 175 billion parameters for GPT-3.5 and 1 billion for the premium GPT-4 version, giving ChatGPT an edge. Bing and Bard are freely accessible, with ChatGPT offering a free basic version (GPT-3.5) and a premium GPT-4 version with a $20 monthly subscription fee, promising quicker and more accurate responses, reduced biases, enhanced processing speed, and additional features, including multimodal capabilities.
ChatGPT shines with its quick and concise responses across a broad spectrum of topics, while Bard emphasises creativity and contextuality. Bing offers an internet browsing feature and conversation continuity but is limited to Microsoft Edge, whereas both Bard and ChatGPT are accessible on any browser. Notably, ChatGPT and Bard have web-based interfaces, while Bard also stands out for launching mobile applications on IOS and Android.
Bard can provide real-time responses to current events, which ChatGPT cannot. Additionally, Bard employs different metrics to ensure answer accuracy. The reach and influence of Google via popular social networks like Twitter and YouTube work to Bard’s advantage.
What’s truly remarkable about these chatbots is their ability to tailor responses to specific age groups. For example, we tested Bing, ChatGPT, and Bard on explaining Algebra to a 10-year-old learner, and the results were noteworthy.
As for limitations, Bing allows up to 25 questions per day per account but lacks the option to save, share, or export conversations. ChatGPT stores conversations for reference. Bard’s visitor count has surged, with 142.6 million visitors in May, up from 49.7 million in April, largely due to Google’s widespread reach.
These chatbots are not without flaws. They inadvertently include biased information and sometimes struggle to grasp the full context of prompts. Both OpenAI and Google acknowledge these limitations and encourage the responsible use of chatbots and AI assistants.
The rapid evolution and growing adoption of AI chatbots raise important questions about their potential and limitations. Establishing regulations for AI use is a fundamental starting point. As the AI landscape continues to evolve, it is crucial to consider ethical principles and responsibilities to ensure the responsible development and use of AI technologies.
The ongoing boom in generative AI could follow the path of the dotcom bubble, and stakeholders must be mindful of potential risks and costs associated with data-driven AI. Policymakers and academics play a crucial role in shaping the future of AI, emphasising ethics and responsibility as guiding principles.
Naledi Ramontja is with the Institute for Pan-African Thought and Conversation at the University of Johannesburg.
