AI productivity tips from a librarian with too much to read – part 1

By Tim Graves

Late in 2022, Artificial Intelligence headline-grabbed its way into the mainstream, most unmissably in the form of ChatGPT. By playing around with Artificial Intelligence throughout 2023, I am convinced that AI has saved me days or weeks of working time. Because it has had such an impact on my year, I thought I should share a few of the productivity tips that have leapfrogged me over workaday tasks. 

Relaxed person typing on Mac book pro
Photo by Christin Hume on Unsplash

Words, word, words 
In subsequent blog posts I plan to talk about AI in data analysis/visualisation, image generation, speech, and maybe even music and video.  But for now, I want to focus on how AI has helped me deal with words. The daily fusillade of emails, articles, reports, newsletters, web pages. 

ChatGPT, Claude and Perplexity 
I’ll be talking about free versions of three tools I’ve found most useful over the last year when it comes to handling long emails and documents. 

At face value all three provide a similar output. All three generate eloquent responses in English and other world languages. 

AI document tactics 
To illustrate how AI has changed my tactics, let’s imagine that someone (malevolent) asks me to read the blockbusting government publication:  “Levelling Up the United Kingdom: Executive Summary” 
My 2022 methodology would have involved gloom at the prospect of dragging my eyes from cover to cover. Maybe I would skim read a few bits, or perhaps I would add it to a queue of other ‘long-and-important-but-boring’ documents. 

In 2023, though, lets walk though my approach. 
I would copy and paste the contents into a ChatGPT session (or upload the file to Perplexity or Claude). 

I would then begin a dialogue with the text by typing in my questions: 
Me: “What is the main point of this document, in 20 words?” 
AI then provides me a one-liner that summarises the whole thing. 
I decide to find out a bit more. 
Me: “In 100 words of bullet points summarise the document.” 
AI duly delivers me a list of the main points. 
Me: “What are the five key topics of the document?” 
Me: “What is the main conclusion in no more than 30 words?” 

Immediately my relationship to the document has changed into a conversation. #
I have a quick sense of what the text covers. Perhaps that is all I need in order to prioritise whether to read it straight away or put it aside. 
If I were to be hauled into a surprise meeting I would already be ahead of the game: able to fend off basic questions, and in a much happier position than everyone else who couldn’t face even looking at it. 

One of the AI bullet points reads
“Key Focus: Boosting productivity, innovation, economic growth in less prosperous areas.” 
Let’s suppose this sparks my interest, so I decide to probe further:

Me: “What does it say about innovation? No more than 20 words.” 

So, already AI has summarised a long document for me, listed the key topics and allowed me to identify a thread I want to pursue. 

Imagine this same approach applied to
A long email,
A newsletter,
A product release notice,
A research paper,
The transcript of a meeting that I missed,
I get immediate insight without having to plough through every detail.

I can go further and adapt the tone to suit my personal requirements:

Me: “Reword this without jargon for a non-technical reader.”
Me: “Explain this to a 10-year-old.”

I can pull in information from outside the document, to clarify certain points:
Me: “Explain what SME means.”
Perhaps English is not my first language, or the document is in a language unfamiliar to me?
Me: “Translate the document into Mandarin/Welsh.”
Or I can get silly:
Me: “Rewrite the main points as a Shakespearean sonnet.”
Me: “Write me a rap based on the primary conclusion.”

Take-home message 
The next time you face a long e-mail, document, or you need to catch up on meeting notes, I encourage you to experiment with one or more of the AI models. You should be able to win back a few minutes or even hours of your day. 

Chat GPT3.5 
Claude 2.1 
If you are asked to sign into any of them, you should be able to choose Google as your account and use your standard

Direct comparison of key features
Each of the three tools has different strengths/super-powers/limitations: so, each might prove useful in different situations. The following table might help you decide which to try with any specific text. 

Model Message limit (approximate) How long a document can it handle? Can it search the Web? Can I upload a file? 
Chat GPT 3.5 30 messages per hour 10 pages Only has knowledge from the Web up to January 2022. No 
Claude 2.1 100 messages every 8 hours 500 pages No Yes 25 messages per day 500 pages Yes Yes 

In the next post 
In my next blog post I will move on to how AI has saved me time and drudge with data analysis and visualisation. If you want to win back a chunk of your day from spreadsheets, join me next time. 

Tagged with:
Posted in The Library

Leave a Reply

Your email address will not be published. Required fields are marked *