GPT-4, Insights, and the Renaissance of Excellent Conversation

As we look towards a future where the art of conversation is a skeleton key for technology’s finest fruits, what should we Insights professionals be considering during this time of change?

Kids tomorrow hey?

Showing off a level of conversational acumen not seen in the Western Youth for 100 years. The practice of holding a naturally flowing, eloquent, and interested conversation has ‘gone viral’ in these future days. Down with them, the middle-aged grown-ups will say. All this harking back to the fire-side chats of our grandparents has become something of a friendly competition, a menace, infecting our entire Society! What ever happened to the good-old days of hunching over a phone, offering up an occasional grunt for more pizza?

In this author’s opinion, GPT-4 is another step towards a brighter future, where the best of our past fuses with the best of our present. A future where the art of empathetic conversation is a skeleton key for technology’s finest fruits. However, the old saying rings true, “fail to plan, plan to fail”. What should we Insights, ResTech, and MRX professionals be considering during this time of change?

Top of the shop, same as it ever was

Privacy, Permission, and Data Protection. Our sector trades on credibility. When connecting LLM resources (like GPT-4) we need to be clear on data flow, storage, and permissions. Far reaching regulations like GDPR are especially relevant. As a base example, those without an understanding of whether their open-ended feedback is defined as Personally Identifiable information or not, could find themselves in hot water. This will throw up barriers to early adoption. We may see AI-specific consent as standard. Options for hosting GPT models internally will no doubt become easier to implement. We’re not there yet.

Pics or it didn’t happen!

The way GPT-4’s Deep Learning builds on GPT-3 shines a light on the path ahead. GPT-4 can, “…accept images as inputs and generate captions, classifications, and analyses.”. Communicating via pics is second nature to many, and GPT-tech could leverage that channel to the full. Insights may soon be able to interpret, categorize and analyze thousands of primary research photos, in seconds, for very little cost. Further, in the right ResTech hands, GPT-4 could significantly improve the research experience for visually impaired participants, a boost for inclusion in our projects

Qual people get ready, for a change is coming

We’re told GPT-4 is more creative and collaborative than ever before. As the prevalence of GPT systems in business increases, so too will the value of those who can effectively engage. At Bayes Price, we’re collaborating on new research to understand if those with a background in qualitative research are best equipped to design prompts that generate high-quality responses from AI-driven GPT systems.

Good prompt writers or ‘engineers’ can add context, frame the required response, provide examples, use key terms to draw out the most relevant content. They hold the underlying objectives in mind and, importantly, engage in a much more conversational way than we type into search engines. It’s fair to say that qualitative research professionals have been doing all that for…. ever. With prompt writing salaries of $300K being talked about, it’s certainly worth investigating.

End of Days for the Quantie?

GPT-4 can solve complex mathematical equations, but can it interpret a very simple cross tabulation? Well, when asked to, “Provide a summary of the following cross tabulation, which presents the % of married people in different household types:”

 

GPT-4 offered up, “The crosstabulation shows that households in the “Young Single” and “Peer Group” categories have no married individuals. In contrast, households in the “Young Family,” “Middle Family,” and “Mature Family” categories are predominantly composed of married individuals, with more than 96% of households reporting being married. Lastly, the “Older Couple/Single” category shows slightly over half of the households reporting being married (50.6%).”

Base observations for a new starter. The true value of GPT-4 came when we followed up with, “Why do you think older couples / singles are more likely to be unmarried?”. The conversation that followed was most thought provoking.

Is GPT-4 our only option?

Not at all! We’ve experimented connecting OpenAI’s text-davinci-003 model to our in-house software with similar results. There are a plethora of open source and paid-for options to test. A key issue is transparency. Yes, we can get the top-10 themes from a huge text-based data set in seconds, e.g., “Why do you recommend brand X?”. A great start to a code frame, but without transparent respondent level tagging, it’s not something to hang our professional hats on…yet.

Generative-AI Software!

The game-changer for quant insights, to my mind, will come with conversational software development. Imagine a time when we can make a morning coffee, then simply say, “Hey GPT-10, please create an Adaptive Choice Based Conjoint application to assess how employee groups trade-off different types of renumeration options. Deploy the study among our 10,000 employees via an invite which maximises participation. Compare findings to previous internal research.

Consider our financial position and growth plans, then present me with the combinations which you think will improve workplace satisfaction, attract the skills we need, and reduce churn. I prefer reports in PowerPoint, and explain your reasoning. Oh, before launch, please detail your research plans, methodology, and approach to modelling. Share those pre-launch reports with my expert research team for review. Thank you.”

Manage Your Knowledge Today

Knowledge management and the growing importance of us humans using consistent terminology, is another area to keep in mind. The potential value of MS-365 Copilot searching and summarizing multiple internal docs is awesome. However, that will only be as good as the sources available. Therefore, consistency is likely to prove incredibly valuable down the line. Now is the time to refine and align terminology across your organization.

Old Tricks for Renewed Dogs

It’s important to acknowledge that, generationally, we may start to see a new relationship between age and technology. Older people often feel alienated from contemporary tech roles. In this new GPT-4 world, their experience of developing conversational skills in a pre-internet era may put them front and center. Interesting times lie ahead for sure.

All that said, things are moving so fast that these thoughts could be irrelevant by the time they are published. GPT-5 may have achieved Artificial General Intelligence by then, or regulatory bodies may have shut it all down.

To Conclude, Nostalgia for The Future…

Perhaps you’ll permit me to conclude by tossing another log on the figurative campfire, winding the clock back to an antiquated style of wordiness, and proffering a proposition. If it does so happen that AI tools inspire a renewed fervor for conversational excellence, a renaissance if you will, then I’ll wager the practice extends to our daily interactions with one another.

We’ll become more interesting and interested, with an advanced capacity to listen, and to understand from each other’s perspectives. If we’re cultivating these characteristics every day, dare we dream that an improved social cohesion among individuals ripples outwards, transforming our social world, into a more pleasant and fulfilling shared experience? Here’s to hoping. Cheers!