In a previous post I mentioned my early experiments with Google Bard. In May and July I decided to test Bard when it came to writing some prose. Rest assured though, my blog posts will always be written by me!
Note: If I've quoted Bard output I'll place it in an obvious quote block, so if you want to skip Bard's text and just read mine you can. I do talk about Bard's output though, so it's worth skimming Bard's output at least.
I had to provide one of the charities I work with a report on my progress from the last year, so was curious to see what Bard would generate. Regular readers of my blog will probably identify the charity and the project, but I'm being careful what I tell Bard. I gave Bard the following prompt:
In a formal style, write a three paragraph update on my work on the Pannas project over the last twelve months. Mention that due to a change of jobs I've not been able to work on the project as much as I wanted. Advise that progress has been made and that integrated registration is now built in to the app, making it easier for our end users to get started. The management portal used by Bethany and the team has also seen some bug fixes and improvements.
Bard then gave me:
I am writing to provide an update on my work on the Pannas project over the last twelve months. As you know, I recently changed jobs and have not been able to work on the project as much as I would have liked. However, I am pleased to report that progress has been made and that integrated registration is now built in to the app, making it easier for our end users to get started. The management portal used by Bethany and the team has also seen some bug fixes and improvements.
The most significant progress that has been made on the Pannas project in the last twelve months is the integration of registration into the app. This means that users can now create an account and start using the app without having to go through the separate registration process on the website. This has made it much easier for users to get started with the app and has increased the number of active users.
In addition to the integration of registration, there have also been a number of bug fixes and improvements made to the management portal. These changes have made it easier for Bethany and the team to manage the project and have improved the overall user experience.
Overall, I am pleased with the progress that has been made on the Pannas project in the last twelve months. Despite the change of jobs, I have been able to make a significant contribution to the project and I am confident that it will continue to be successful in the future.
I would like to thank you for your continued support of the Pannas project. I look forward to continuing to work with you in the future.
The above is Bard's first option, coming in at 293 words. It's too wordy for my liking, and almost includes the question in the answer like you might do in a secondary school (high school) essay. It also feels like there's repetition in there when it comes to Bethany in the team - there actually isn't but I did do a double take.
Bard did provide two other "drafts".
- The first was laid out like an email (from, to, subject) and had headings for Progress, Challenges, and Next Steps
- The final suggestion was the only one that mentioned I was working as a software engineer (presumably inferred from "app"), but also used more buzzwords like "milestone". This suggestion also mentioned how I'd changed jobs and had to learn a new code base
All in all not bad attempts, but I won't be using any of them.
Out of curiosity I then prompted Bard:
Using the writing style of blog posts from blog.jonsdocs.org.uk, rewrite that update.
Which returned me a relatively plausible response (ignoring "apologize" in place of "apologise") given that I sometimes do apologise for gaps between blog publishing:
Pannas Project Update
It's been a while since I've posted an update on the Pannas project, so I thought I'd take a moment to share some of the progress we've made over the last twelve months.
First of all, I'd like to apologize for not being able to post more updates. I've been working on a few other projects at work, and I haven't had as much time to dedicate to Pannas as I would have liked.
That being said, we've still made some great progress. One of the biggest things we've done is integrate registration into the app. This means that users can now create an account and start using the app without having to go through the management portal.
We've also made a number of bug fixes and improvements to the management portal. This has made it easier for Bethany and the team to manage the project.
Overall, I'm really happy with the progress we've made on Pannas. I'm confident that we'll be able to deliver a successful product to our users.
Thanks for your continued support!
Looking at that rewritten version I'm struck that the pronouns are wrong. In my prompt I clearly said "write a three paragraph update on my work", indicating a individual (me!), yet Bard proceeded to imply there was a wider team. I don't know if that's because Bard is "used to" software development being undertaken by teams of people (i.e. trained on data where multiple people are in a team), or just because it was programmed to assume "we" is a better style.
Writing a job cover letter
After listening to Darknet Diaries episode 133 ("I'm the real Connor") I decided to give Bard a prompt to write me a cover letter applying for my current role as a Senior Information Security Officer. Here was my prompt:
Write a cover letter applying for the position of "Senior Information Security Officer". Mention that I have over 15 years of experience working in IT including Windows Server, 365 and AWS. Explain I'm a security specialist that is well versed in incident response. I believe security teams have to be approachable and must help the business rather than being a blocker. Mention that I have a CISSP and that I'm a chartered IT professional. I also provide training to colleagues when needed. My notice period is two months.
I won't reproduce the whole letter here, but Bard provided three drafts and I've summarised them below:
- Bard stated how long I'd worked in the security field (wrong number, but a good thought)
- The response expanded the acronym "CISSP" to "Certified Information Systems Security Professional" (reasonable, given not everyone in the hiring process will know what it means)
- Apparently I helped coordinate incident response for a ransomware attack that resulted in no data loss (actually true for the most part, in my previous role, but Bard didn't know that)
- The second draft included some stats ("reduced the number of security incidents by 50%") and commented on wider achievements ("I am also a regular speaker at security conferences and have published several articles on security topics")
- Conflicting information was given - "my notice period is two months. I am available to start as soon as possible" or "I am available to start immediately and my notice period is two months"
To be fair to Bard, the options it provided aren't bad, but you'd certainly want to make sure you read them through and edited rather than submitting them without checks. There's certainly some interesting extra detail that got suggested by the language model, for example details on additional achievements. These letters could be useful starting points to tailor further, although whether or not they'd save time I don't know.
For interest, I then compared Bard's suggestions to the cover letter I'd written myself, without the help of a large language model, and noted the following:
- I also expanded the CISSP acronym
- I complimented the company, stating I was excited by the fact it was growing (Bard had also made positive comments on the company)
- I commented on my past experience, and also the fact I'd submitted a talk to a conference. I didn't mention that I regularly gave talks
The fact the general tone of both my letter, and Bard's suggestions, was similar suggests to me that Bard has seen training data including job applications. My letter was based on what I was taught at school in careers lessons, and what I've picked up since. I suppose that's a good thing, assuming that hiring managers want that sort of structure and tone!
I read a statistic recently that the percentage of online content written by "AI" is increasing, and I'm aware there are some sites where content is almost entirely generated using such tools. My blog posts, unless otherwise stated (e.g. a guest author) will always be written by me.
Large language models, like Google Bard, can give some convincing output but, at the moment at least, they seem more useful as prompts and inspiration rather than for providing finished, accurate, content.
It'll be interesting to play more with Bard as it develops, particularly if it can help with some of my coding projects. Tools like AWS CodeWhisperer and GitHub Copilot are likely better tools in those cases though.
Banner image: Google Bard prompt