How should we use AI? I’m baffled as to why this question takes up so much time. We’ve been here before. I’m that cohort of Gen X that went through school on paper and were met with computers, email, and the internet during university. Lecturers, who’d earned their stripes in the hard yards of the library, were unsympathetic to our lazier search engine sourcing. What we had to learn quickly was the hierarchy of search engines. Same again with AI sources.

Many more learned and tech-minded people write cogently on this topic. This newsletter is a flashback to my university days – a blagger’s guide.

ChatGPT is not AI

It is. But it’s not the only one. If you Google (irony not lost on me) “Best Chat GPT alternatives”, you’ll be greeted with an evolving list. This piece is a good current summary (and will be obsolete soon, no doubt).

I use a blend of:

1. Claude – https://claude.ai/

2. Perplexity – https://www.perplexity.ai/

3. ChatGPT – https://chat.openai.com/

4. Bard – https://bard.google.com/chat

Here’s why and what it looks like.

Research

The European Union’s pithily-titled Corporate Sustainability Due Diligence Directive (god bless you, bureaucrats) is easy to find. As risk and compliance folks, we should know the source material. But it’s not an easy read, even the 7-page Annex summarising the main requirements.

Before we test the tools, we need some rules. Cr@p in, cr@p out. If you copy and paste the text and write something lazy like, “Please summarise this document,” the results are unhelpful to anyone but you. You’ll be left with an abridged list, like this image:

We need to use the AI tools as intended. Step up Perplexity (but Bing and YouChat may also be helpful here). Perplexity does some research along with its synthesising data. Helpfully, it footnotes the sources. I asked Perplexity:

“What are the key requirements of the EU corporate sustainability due diligence directive?”

The reply:

It’s not a bad start if your objective is to brief someone on the EUCSDD. Let’s say, the board. But not good enough for general consumption.

Summarising

Here’s where it gets complicated. I’ve used ChatGPT, Claude, and Bard for summarising, with mixed results. Just as in the early days of the internet, the search engines were not created equal. Just as Yahoo, Google, MSN, etc., started to speak to different audiences, so is the case (from what I can see) with the AI tools. ChatGPT is suitable for everyday stuff. If you’ve got an email that’s too long, or you want the Cliff Notes on that “oh so important” 45-min read of an article your single friend sent on a Sunday morning as they leisurely peruse the papers unencumbered by warring hell-hobbits [children], go to ChatGPT.

But if you want to work, I use Bard for technical stuff (e.g., with a mathematical component) or Claude for everything else. Here’s the difference when I asked, “Can you re-write [the Perplexity summary above] for an audience with no prior knowledge and a reading age of 10 years old.” I use the 10-year-old test in most work product.

Chat GPT:

Claude:

To my read, Claude treats the reader as someone worth informing. ChatGPT treats the reader the way David Walliams treats his readers – like you’re stupid (for the uninitiated, he’s a Roald Dahl rip-off using bigger font [less effort] with lots of space taken up with ellipses and extended words, “Arghhhhhhhhhhhh”, etc. ).

Arghhhhhhhhhhh, but wait

To my mind, summarising is Claude’s expertise, especially in legal or other complex documents. But what if we want to be more creative? I use AI to generate ideas – as that’s the most sapping and intense part of my job – I write between 50,000 and 350,000 words a week. Much never sees the light of day, thank god. But my writing strategy is the “throw it against a wall and see what sticks” variety. For every ten ideas, one might be passable. Coming up with stories, scenarios, or anything else engaging is made easier with AI.

I asked ChatGPT and Claude:

“Can you give me 5 scenarios where this could go wrong for these big companies, and what they should do to stop things from going wrong?”

Claude first:

ChatGPT:

Again, Claude gives me more to work on, but the gap closes again. There are some examples (from the 10 scenarios) that we can run with.

However, scenarios don’t come to life alone. We need stories. So I asked the tools:

“Could you turn these five scenarios into a short but memorable story for training?”

Chat GPT

Claude:

ChatGPT is more didactic and a touch longer, we now see. But here, that’s not altogether a bad thing. I have five mini-scenarios I can build on. At the same time, Claude is more creative and does the heavy lifting for me. Is that perfect? No. I should have first put specifiers around the industry to tailor it to a particular organisation.

So what?

Would I copy and paste this into the work product (training, etc.)? No.

Would I rely on these tools for a concept I didn’t know or hadn’t read the source material? Not yet.

Does it help me multiply ideas, crowdsource, and overcome inertia and creative blocks? Yes.

What’s the value of that? Priceless.