ChatGPT - Robotic Creativity

It wouldn't be 2023 without some sort of reflection on ChatGPT...

Following Shawn Curran 's article here exploring the extent to which ChatGPT could be considered creative; namely able to connect ideas across domains, I had a play with ChatGPT on the side of Joe Cohen 's desk earlier this week; mainly as I wasn't quick enough to access it myself when the internet was initially very excited about it.

I wanted to sit properly with ChatGPT and discuss my MBA dissertation; process modularity in legal service design.

Why?

Well, because my dissertation is just a little bit niche. In substance there are two unpublished manuscripts exploring the topic in detail. One is mine, and the other is Darren Mee 's. And we're the only two people that have copies of both.

Whilst it's been ~8 years since I wrote said dissertation, and quite feasibly more ink has been spilled (and published) on the topic since then; I'm still reasonably confident there isn't that much source material out there.

And also, what else am I going to do for fun on a Saturday afternoon...

So I asked ChatGPT four questions:

1. How might process modularity from a lean manufacturing context be applied to legal delivery processes?
2. What might the challenges be in applying process modularity to legal services?
3. And the benefits?
4. Can you provide references to relevant academic papers on the topic?

The full conversation is at the bottom of the article. My reflections:

• 8/10 - Good overview to question 1 - outlined what the concept is, the high level benefits and how it might be applied. The core idea; of combining standardised processes to produce customised outputs, is in there - although comes across more clearly in question 3 (point 4)
• 6/10 - The challenges in question 2 are fair - perhaps a little generic to change management/process improvement in general, but it picked up the domain specific issues around regulation (point 6). It did miss a key challenge around interfaces (i.e. how you get one process module to work effectively with the adjacent module). It hints at it on integration (7), but I read this as a macro point on integration of the overall architecture, rather than the micro point on module to module integration. Perhaps I'm being harsh...
• 7/10 - Benefits in question 3 - pretty good/comprehensive, although there's some nuance missing. For example, point 6 - easier to manage; sometimes yes, also sometimes absolutely horrendously no. And point 8 - again yes - but dependent upon the service design and the point on interfaces, as above. Modularity can actually increase organisational silos rather than decrease. Mainly if done badly.
• 3/10 - Academic references. Appreciate at this point I'm straying outside of scope of ChatGPT's capabilities and the suggested keywords and journals are a pretty decent jumping off point. However, there are more relevant authors to the topic closer to the legal/professional/business services domain than those provided (e.g. Rahhika, Pekkarinen, Ulkumiemi, Sanchez, Mahoney).

Overall, a pretty reasonable executive summary, or abstract, on the topic - and it's not obviously lifted it materially from a single source - it's patched together across a wide range of source material and presented it coherently.

This is looking a lot like creativity to me - to Shawn's article linked above - the tracing of linkages across previously unconnected domains.

Is it as good as a well researched, referenced dissertation? Does it have the scars from implementing the concept in an actual law firm?

No. Obvs.

But, and I think this is the key reflection - it took me a solid 6+ months to research and write my dissertation, and many, many hours of testing, piloting and refining the associated prototype.

Whereas the conversation below took less than one chargeable unit (6 minutes).

Perhaps for my own sanity more than any real world impact, all that effort must count for something. But does that depth of knowledge scale proportionately with the effort expended?

I'm not so sure on that...

My takeaway is that ChatGPT is going to be (more than) 'good enough' across a whole raft of different areas. The ramifications are huge.

PS - The cover image above is a DALL·E 2 creation prompted by "impressionist painting of process modularity". I'm quite fond of it. Although at this point, I suggest we all give up - the robots are very much coming...