Beyond the Hype: How I Learned to Maximize Generative AI in Everyday Life

I'm often surprised by how many super smart, capable people are not yet consistently using generative AI in their day-to-day workflows, but I think I know the reason why.

Most of them have tried it, and been impressed by it, but they're not sure really how much it can do - nor how to get it to behave as they would want. I get this. At the end of last year, when I started using ChatGPT, I used to ask a friend and colleague of mine "can it do X?" or "how can I get it to do Y?".

He used to always give me the same answer: "Ask ChatGPT." Although at the time this was simultaneously both amusing and frustrating, it was actually the best possible advice he could have given me.

Because very early on, it made me accustomed to being highly explorative with ChatGPT, to constantly experiment, try things, push to find the edges of its capabilities - and basically develop my own unique 'personal' relationship with it.

I learnt its quirks and foibles, when and how it's likely to hallucinate, and how to mitigate that. It gave me the confidence and enthusiasm to constantly venture into uncharted territory, with sophisticated multi-step prompts, detailed context and skill setting, and techniques to bring out the best of its capabilities.

My key learning? A good starting point is to assume it can do anything you want. And then confirm with ChatGPT (or your LLM of choice) what exactly it can do, and how it would approach it.

Often I tell it "I have very low bandwidth for these tasks, so please shape your approach based on what you, as an LLM can take on for me." It's really happy to do that. It doesn't mean every detail will be perfect, and sometimes it takes a fairly long, iterative process to get to a satisfactory point, but the productivity this unlocks in the medium term is completely worth the effort.

It’s also worth sharing another tip - if you’re not happy with what it has generated, then tell it, as clearly and directly as possible, explaining what isn’t right and how it should be improved. (This is part of my ‘personal’ relationship with it, very transparent and based on mutual respect!)

So why the Scandi-chic photo with soft grey walls? Well, one of my recent use cases was getting ChatGPT's advice as a DIY expert.

Me:
/interior_decorating_expert
I painted a wall in an apartment.
I sanded it first, and used a primer.
Then made two coats, dark colour going onto white wall.
I think I made the first coat too thin, and now I'm not sure if the second coat will give a 100% even finish.
I have a bit of paint left, but not so much.
Should I dilute the paint and put a really thin third coat?
Should I just touch up any parts that look under-covered (or will that just make more streaks)?
Or should I buy more paint and make a proper third coat?

ChatGPT:
As an interior decorating expert, here are my suggestions for your situation:

Avoid Diluting the Paint: Diluting the paint, especially when it's a dark color going onto a white wall, is not advisable. It can affect the color consistency and opacity, leading to a less even finish.

Touching Up: Touching up can be tricky, especially with dark colors. It's often hard to blend the touched-up spots seamlessly with the rest of the wall. If the areas that need more coverage are small and not very noticeable, you could try this method with a small brush. However, be cautious as it can indeed create streaks or noticeable patches.

Third Coat: This is generally the best option for a uniform and professional finish, especially since you're working with a dark color over a light base. A third coat will help ensure that the color is rich and even across the entire wall. Since you don't have enough paint left for a full third coat, purchasing more paint is advisable.

Application Tips: When applying the third coat, make sure to:
Stir the paint well before use.
Start from the top of the wall and work your way down.
Use a roller for larger areas to get an even coat.
Be consistent with the direction of your strokes.
Allow sufficient drying time between coats.
Remember, the end goal is to achieve a uniform and smooth finish, and sometimes that requires an additional coat, especially when transitioning from a light to a dark color. It’s better to invest a bit more in the short term for a long-lasting, quality finish.

In the last year I have had ChatGPT plan a motorcycle trip for me (with consideration of scenic roads, great food and nice affordable hotels). It did amazingly well and saved hours of web searching. It did equally well with advice on sofa fabric, real estate investment, German tax law - and of course a whole load of work applications.

But experimenting with personal use cases pays dividends, because they often lead to creative approaches that translate well to work tasks - and they encourage that playful mindset that is so important in getting the most value from generative AI.

Remember, this is still the beginning of one of the biggest shifts in living memory. So even if you haven’t really used it yet, start now and you’ll still be an early adopter!

Image: pure-julia-O_4ZEWRLyl8-unsplash