Mahatma Gandhi using laptop

A year spent in artificial intelligence is enough to make one believe in God


From judging the crypto bros on LinkedIn that became AI connoisseurs overnight to writing about this topic, life has come a full 360. After all, there were pretty bold statements being thrown around how Google’s ad model has come to an end and how hundereds odd jobs are at risk. Before I could succumb to the FOMO and join hands with the machines, I decided to experiment with AI before jumping to some conclusion.

So one fine morning, I read an article by Neil and tried to apply that framework to GPT. What would I miss if AI ceases to exist from tomorrow? Turns out that I have been thinking of rejigging my Linkedin content strategy and existance of GPT is one instance where I could do it for free (at least from a money perspective).

Now before we proceed, I want to make certain things clear. I am not an influencer (someone who is actively looking for paid collaborations for mentioning certain brands). Neither do I write with the aim of increasing my followership/engagement. I write for fun and if the by-product could be some form of credibility, be my guest.

Off late, I have been reading about how aimless shitposting like my own would lead to absolutely nothing. That is when I decided to pay some attention to it. But then I never ran out of topics to rant on, so I let it be. Until..

Introducing My Content Strategist: ChatGPT

Think of this post as an attempt to document my experience as a power user of GPT.

So a beautiful Saturday morning (most creative day of the week for me) and I share the following prompt with GPT. In response, it (yes, I am not referring to GPT as he/him or she/shlem) agrees to take up the challenge and spews out a detailed content plan for the week.

Something on these lines ๐Ÿ‘‡

And thus began the experiment of a lifetime. For the first time ever, I am not being given some links that could help me with the strategy, but an actual content plan. Let us not discount the fact that it is pretty generic, but we are off to a good start.

Also, please note that I am fully conscious of the steep target of hitting 100,000 impressions from a measly 10,000 per week.

At this point, I did expect some questions from GPT before spewing out a strategy but it did not. I have always been a fan of setting the right context, so I took the lead.๐Ÿ‘‡

Here is the first glimpse of why I call GPT as human. Despite prompting it to ask questions, it decided to simply complete the task as soon as possible. Aren’t we all like this in some way or the other? We want to do a quick shabby draft to outsource the process of iterative thinking to the consumer/manager etc. I felt that GPT treated me exactly like that.

But I was pretty clear in my head of what I wanted this to be ๐Ÿ‘‡

Aaaand just like a little human intern, it complied. This time, it asked me to post a video introducing myself. It was kind enough to give me the content of the video.

Now all of us at one point have dealt with a manager/client that has all kinds of weird constraints. So why not give GPT a flavour of humans right in the beginning. ๐Ÿ‘‡

How considerate at a ripe age of 3 months, right?

Just to be sure, I asked GPT again that if this does not align with the north star metric of 100K impressions it can ask me to create a video. But GPT was confident in its content game as it clearly mentioned that we are okay with the written content as well.

Next, it gave me the post and I asked it to describe the image it wants to go with it. GPT described the <cough> image <cough> beautifully.

Turning into a Cringe Influenzah:

Wait a minute. Isn’t it exactly what we all hated about LinkedIn? Generic words of wisdom with abstract images or random selfies at worst? Is that how LinkedIn works?

Since GPT is an aggregation of human wisdom, are we all cringy content creating influencers? Is this why these posts get so much attention because people like them?

Marketer inside my was torn apart.

So I decided to add a little more of myself. And once again, GPT was just okay with everything. I tried reminding him about his authority to say know if it thinks this would come in the way of 100K impressions.

Post 1: Up and running. Interested in the outcome? (Dig right in)

It is pretty straight forward from here on. Every night I would share something like this with GPT and ask it to share the post for the next day.

As an additional input to the model, I also shared my top performing posts in the past 365 days so that GPT could learn what resonates with my audience (GPT’s words, not mine).

Why Fire GPT?

Today, it has been exactly 8 days since I am posting content curated by my master, AI. Dystopian vision aside, I would talk about a few reasons that someone should stop worrying about AI taking their jobs. Yes. Even if they are a content creator.

1. Non-Conversational Conversationalist:

The biggest USP of GPT over other AI models in the past is its ability to retain the context and hold conversations for a longer duration (It is not a ‘stateless’ bot like its peers).

Turns out, that is quite not the case. It would often forget what our intent was and had to be reminded of a few things.

Look at this amazing conversation between man and the machine ๐Ÿ‘‡

While I am saving some rant by quoting a single example, the situation was quite worse on some occasions.

2. Go Woke, Go Broke:

A quick prediction; this is soon going to be one of the biggest AI problems. The constant war between left and right. Between liberal woke and conservatives.

That is the thing about humans, their personality is a pretty wide spectrum. To strike the right balance, is going to be difficult for sure.

Sometimes I think if polarization is the only way for platforms to survive. Speculations aside, I found it really annoying that GPT won’t let me talk about current affairs.

I do understand that GPT is not trained on live internet data but then I fed it the news of CZ being sued by CFTC for violating US regulatory requirements. I gave it some context from a news article that I came across on this topic.

This is what I got in return.

I had a similar problem while using Bing Chat to generate a post themed on the TV show ‘The Office’. As I prompted the model to bring out the Michael Scott personality that cracks homophobic jokes on Oscar, it simply refused in a rather profound fashion.

Conclusion: It is futile to train AI to take up the steering wheel.

May be in future, but definitely not today. AI is power without accountability. And this is just wrong!

3. Target Vs Achievement: Realism

I have spent enough time as a part of the sales team to know how this works. Now reaching 100,000 impressions from a 10,000 in 7 days is an insanely difficult task.

It is simply not feasible under normal circumstances. But just like a sales professional, ChatGPT took it like a boss. In fact, as a employer/manager I did not have to coat it in cringe cricket motivational examples before shoving it down its throat.

Every month you would say ‘yes’ to a hefty target and towards the end of the month you would mutually agree on a ‘likely’ achievement (And then default on it too).

So as soon as I could draw parallels here, I wanted to address the elephant in the room ๐Ÿ‘‡

Aaaand that’s when the GPT spilled the beans:

Then I asked it to give me a quote on how long would it take the impression count to 100,000. And for some reason it kept on stalling the question (in hindsight, I could see the ‘person’ who knows that the target is too steep).

Pushing it to the corner, I found a really interesting response. A target revision?

And finally, GPT was relieved. After this, I told GPT that my impressions are dwindling around 8,500 and it did the math for the remainder of the days. Did someone say Math?

4. Numbers:

We have seen this in the memes trolling AI. But then experiencing it first hand was super fun. Not only it is bad at math, it is extremely good at cooking up stuff. Swipe to explore the brilliant conversation.

I am sure these issues will be weeded out as we progress further but it makes the model extremely unreliable, especially in the financial industry.

Why Hire GPT?

Okay. I may have been a little harsh. May be AI picks up my personality and becomes the best cyanic to ever exist. But then, the right answer is always in the middle. There are some things that AI can do but humans cannot. And that makes me extremely bullish.

1. Learning Curve

Of course the initial learning models are cost and data intensive, but no one talks about the fact that you only have to do it once! Just once. As a result, if you are looking to hire GPT (essentially replacing an employee), it would be the fastest to complete on-the-job training.

Once I spoon fed it the nuances, that was it.

2. Outsource Thinking:

Of course GPT can generate ideas. But then most of these ideas are generic. Rather, what you should be doing is to come up with an idea yourself and ask GPT to build on top of it. It is amazing to see a content guzzling machine with precise adherence to nuances and constraints.

As an extension of this experiment, I wrote 10 posts on ‘Blockchain explain by characters of the office as a conversation’.

The initial idea was mine.

But GPT took it to another level.

I did not use any of these ideas directly but it served as a keystone to generate relevant content, all written by GPT btw.

3. Introspection:

This is an interesting one. It is based on the hypotheses that AI is built using the collective intelligence of human race, available on a public domain.

As it turns out, the moment I started using GPT, I had a few insights about myself:

  • The content I thought would never work (Cringy Influezah), did almost as good as my ‘researched’ content pieces. So there is an audience for everything. You are just not aware.
  • I never approached content through a number and motive driven approach before this. CTA was an alien concept for me. GPT introduced me to this world.
  • Key fact is not that I was lazy or anything. But my personal learning curve was slashed down due to GPT. Traditionally, I would have read/watched videos on this topic, learnt some, retained some and then finally contextualised it to my own use case. GPT let me skip all the steps and come straight to the point.
  • Finally, it brought a structure to this aspect of my life. I think I am more open to experimentation now than before.

Did I hit the numbers?

Unfortunately, I did not. And like a good manager, I was very transparent (and candid) about it.

I think I now get the rationale behind this statement;

AI will not replace humans, but humans who do not use the AI


If I were to go back to pre-GPT era (can’t believe I am saying this), may be nothing would change. But then the quality and quantity of content I can push out now is way beyond my bandwidth. As a reference, I scheduled 12 posts on LinkedIn this weekend. My general trend is 5, in the best of the weeks.

We are heading towards an era of content saturation and only way to save yourself from this is to be as human as possible (of course with an AI by your side).

Leave a Reply

%d bloggers like this: