When ChatGPT was announced last year, I was knees deep with the team, inside the arduous final touches of writing an NDA.
To be specific- on the day of the announcement, I had draft section 3.2.P.2.3 on my desk- which tells the story of DP process development- from the early phases all the way to the proposed commercial manufacturing process.
When you write it, the story unfolds… then refolds…. then reshuffled… then redacted, then expanded, then justified, then…. you get the drift!
In those sections, we are communicating in the realms of risk and mitigation strategies, which are not always right and wrong, and although we aspire to base everything on data, “clear-cut” facts, and solid science, the reality of Pharma is still very grey, Human-centric and…. communicated as Text.
So there I am drowning in my own text, when ChatGPT comes along and spits out pages of beautifully written text in mere seconds.
Was all this a huge waste of time?
Surely ChatGPT could write it much faster than us- had it all the data at hand**
(** Do NOT use ChatGPT or any other LLM with your confidential company data! The T&C are not on your side.)
But writing is the easy part.
The real challenge is the thinking part- WHAT to say and HOW to say it. What is the reality we are trying to convey? Is the product “sensitive to temperature” or “temperature has an influence on attribute X which could impact Y”? What are the consequences of this conclusion?
There is a common saying in Biotech- “The process is the product”.
I think that holds for writing as well.
In the process of writing, we not only generate descriptions of data (which AI can certainly do much better than us)- we connect the dots in different ways, each representing a slightly different version of reality. It is the aggregate sum of these iterations we do as a team, from which wisdom and intelligence emerge- and this is how we eventually make decisions.
ChatGPT and other generative engines are so amazing because they aggregate the past intelligence of humanity.
But it is we, the current humans, who would need to pave the way and craft future intelligence with the right decisions, the correct balance of risk-benefit, and the hardest of all: telling the true story behind the infinite amounts of data.
======
This week in The Tree Club webinar we will discuss where the enormous amounts of data come from in CMC Development, and how it all maps into CTD Module 3. We will discuss how we currently handle this data (spoiler: not very well), and what the future holds when looking at initiatives from FDA and industry leaders in preparation for the future of a fully Digital CMC portfolio.
See links below.