The preview of ChatGPT became available to the public in November last year. It quickly gained traction and went viral on social media, newspapers and TV shows. Over one million users signed up to experience the latest advancements in artificial intelligence through ChatGPT in just five days. This is a monumental achievement, especially compared to services such as Facebook and Spotify, which took 10 and 5 months respectively to reach the same number of users.
Unlocking a world of possibilities with ChatGPT
The release of ChatGPT fits a trend of the rise of large (language) models such as GPT-3, DALL-E, BERT and LaMDA. ChatGPT was launched for the public in a very accessible and conversational way to give a large audience the chance to experience how artificial intelligence could contribute to their daily life. Impressed by the humanlike answers and conversations that ChatGPT produced, users came up with thousands of use cases in a wide range of domains. To name just a few:
- Content creation: Ask ChatGPT for a draft, let it finetune and update your text, and within seconds, you have co-created blog posts, funny stories and even homework on any topic imaginable
- Programming: ChatGPT can assist in finding bugs in code, writing SQL queries and turning unstructured data into structured data
- Text processing: ChatGPT has the potential to revolutionise customer service by summarising long e-mails and helping contact centre agents write responses to customers
- Chatbot: ChatGPT can improve current chatbots by drafting personalised answers, creating training data and responding to non-business questions
ChatGPT: A powerful tool to be used wisely
ChatGPT seems to be at the peak of (inflated) expectations and its impressive performance can make one impatiently jump on the ChatGPT bandwagon. However, when exploring the use of ChatGPT for professional and large-scale applications within the enterprise, we see three key factors to consider:
- Uneven factual accuracy: One major issue is the uneven factual accuracy of the output. The answers generated often sound very plausible but are not guaranteed to be true. This might not be a problem when answers can be validated easily, as with code. However, when for example giving financial advice to a client, the answer should be true and apply to the client’s situation. A ChatGPT-based application that spits out false information could be a major liability for a company
- Inconsistency: Another drawback is the inconsistent output generated by ChatGPT. The model’s output is dependent on chat history, prompts and chance, which can lead to different answers to the same question. This feature is great for brainstorming sessions or content creation, but not so much for situations where users need consistent instructions such as in legal, financial or medical advice
- Explainability: The uneven factual accuracy and inconsistent output raise the question of how the output was generated. The nature of the model behind ChatGPT makes it difficult to trace back which features caused unwanted output, making it hard to adjust specific answers
These may sound like showstoppers, but there are still many applications where ChatGPT can prove to be valuable. It shows great potential in its capabilities to process natural language where it can perform complicated tasks like summarisation, structured text extraction, sentiment analysis and more. Furthermore, we see its incredible value as a smart assistant, where ChatGPT improves code, suggests titles for presentations or comes up with creative ideas for hackathons.
Elevate your business by leveraging AI solutions like ChatGPT
At Valcon, we are certainly impressed by the performance of ChatGPT, but it is not just about being impressed, it is about finding the right and valuable solutions for our clients. That is where our expertise comes in. It is the first time we have seen the public interact with a large language model on this scale. However, this is not the first time we have seen the functionalities that are on display with ChatGPT. We are constantly evaluating new technologies and solutions to ensure that we provide our clients with the best options for their specific needs.
Want to learn more? Are you interested in how ChatGPT can add value to your business? Do you wonder what the impact will be when your employees use these tools? Or do you already have a specific use case in mind? Please reach out, to our team of experts who can help you find valuable solutions:
Bram Rohlfs, Specialist at [email protected]
Herman Stehouwer, Senior Principal at [email protected]
About Conversational AI at Valcon:
https://valcon.com/conversational-ai/
Links to relevant blogs:
https://valcon.com/conversational-ai-science-fact-or-science-fiction/
https://valcon.com/case-studies/open-source-chatbot-that-handles-sensitive-information-on-premise/
