If your LinkedIn feed is drowning in ‘how to use ChatGPT’ posts and your leadership team are already looking for ways to incorporate ChatGPT into your workflow, you’re not alone. ChatGPT is the buzzword on everyone’s lips right now and rightly so. The release of this new technology is set to revolutionise the way we work. But to get to grips with ChatGPT and what it means for you and your role, you don’t need to have technical know-how. Instead, it’s just worth knowing the basics.
Let’s start with LLMs. ChatGPT is an LLM, but what does that actually mean?
The term LLM stands for Large Language Model. This is a type of machine learning model, which is designed to work with natural language. No complex codes or scripts. Just the simple language we all use in our day-to-day lives.
Large Language Models are nothing new. In fact, while ChatGPT 4 is the version we’re most familiar with today, its predecessor, ChatGPT 2, was released back in 2019. LLMs like this have previously been used to classify, translate and summarise text. But how is a more recent LLM like ChatGPT 4 able to do so much more than this?
“Well, since 2019, these large language models have grown a thousand fold in size, and that gives them capabilities that just weren’t present before. Now, their key functions are to recognize, summarise, translate, predict and generate text.” said Gavin Harcourt, our Head of Platforms here at Streetbees.
“That’s all done through the huge knowledge fed into the LLMs training data sets and this gives it a wide range of context and learning that lets it generate entirely new and unseen output.”
By training these LLMs on huge amounts of text data via books, articles and web pages, the model is able to learn patterns and relationships between words within the natural language. For example, it can recognise that the terms ‘cup’ and ‘tea’ have a strong relationship, as they’re often used in the same sentence. The more data it is trained on, the better it will be at processing and generating content.
While your business might already be using this type of technology for functions like innovation and operations, how can you use it to get richer and more valuable insights from your consumers?
Well, market research often involves a huge amount of data, so there are lots of opportunities to incorporate LLMs throughout the lifecycle of a research project. Here are just a few:
Investigating your business problem and validating your hypothesis
- By web scraping - searching through all of the relevant sources on the internet - an LLM like ChatGPT could generate themes around a specific business question to help you validate and refine your hypothesis.
Generating your survey in a conversational format
- Once you’ve settled on your business question and your hypothesis, the LLM could generate a conversational-style survey for you to test your hypothesis.
Conducting intelligent interviewing
- Through the use of conversational AI, LLMs could conduct the survey and could respond to the consumer’s previous answer, routing or generating questions accordingly.
Classifying your data
- LLMs could not only clean your data, they could also classify it. Instead of someone having to manually sift through data to look for mentions of certain brands or products, for example, this could now be done automatically.
Creating your dashboards in a ‘dash’
- LLMs have the ability to put the ‘dash’ back into ‘dashboard’ by translating the data into charts and visualisations in minutes, rather than days.
Automating your reporting
- LLMs could also quickly summarise your data into draft headlines and executive summaries to help you create a concise narrative for your reporting.
Time is another huge factor.
“Currently, if you have a market research question or hypothesis, it could take two or three months before you get your answer once you’ve decided to go to the market with that question. With large language models, it could now take only a handful of days to get those answers and that speed is going to be really powerful.” said Martin Richie, our Senior Machine Learning Engineer at Streetbees.
“With traditional research methods, you might go to the market with a question and get an answer that you weren’t expecting. You might then realise that you need to modify your original question or go in a slightly different direction to get the answer you need. On the flip side, the answer you get might inspire you to ask another question off the back of it. In both instances, you would need to keep iterating on the research cycle to get to where you need to be.”
“Sometimes, this can take so long that your insights can end up being out of date and no longer relevant. So, being able to iterate on this cycle and hone down on the things you’re actually interested in at speed is what I think makes this so powerful.”
To summarise, a Large Language Model or LLM is:
- A type of machine learning model, which can recognize, summarise, translate, predict and generate text and is designed to work with natural language. No complex codes or scripts. Just the simple language we all use in our day-to-day lives.
- A technology that could have huge implications for market research, from helping you generate a hypothesis to intelligently interviewing your consumers to creating automated reports and dashboards for you.
- Not a silver bullet. Although LLMs might seem ‘magic’ in their ability to recognise and generate content, these models have no ‘knowledge of knowledge’ and can therefore get things wrong. As we explore the use cases for LLMs in the market research space, it’s vital that we approach this responsibly with a ‘test and learn’ mindset.
AI and Machine Learning is at the heart of what we do here at Streetbees and the introduction of these new generative models will allow us to get even closer to consumers and their everyday choices. We’re currently exploring new ways to incorporate the power of Large Language Models into our solution and our pilots are coming on leaps and bounds.
We’ll continue to keep you posted on the exciting new developments we’re working on to bring you research that’s quicker, more accurate and more actionable. If you have any questions about LLMs and their implications for market research, get in touch!
