Until You Get In There

Until YOU muck around with these tools, you're just pushing opinions around. If you're reflecting on all kinds of articles but you haven't yet had a copilot build you out a dietary plan for the week ahead, or give you ways to update your LinkedIn profile, what do you really know? How can you make your judgment?

Sure, sometimes you don't have to get into something. There are many crimes we don't have to commit to know they're wrong. But this isn't that.

Experiment and Make Something - THEN Make Up Your Mind

But even this advice is risky. What if you don't know the best ways to prompt our little robot friends to get you the best results? Maybe you don't know that there are prompt tuning tools, and suggestion makers and the like. Maybe you haven't cracked open other people's spellbooks yet, to see what they've built into their queries to make them more effective.

I learn by making lots of mistakes and iterations. For instance, as hard as I try, even supplying visual content to reinforce my intentions, AI art wants to draw me as a much younger version of myself. So I spend a whole lot of the prompt explaining that I have brown hair, a white beard, glasses, etc. But each experiment, with different lighting, different styles, teaches me new details about the different programs I use, and I take these into other learnings.

Before you get the wrong idea, I don't mean this about drawing. I mean it about all the ways we're going to use AI copilots.

My Ultimate Use Cases Don't Even Exist Yet

First, it's downright NUTS that Google hasn't thrown out "assistant" and replaced it on all my devices with Bard (even though Bard is kinda crappy right now). OR, I yearn for a voice interface into GPT4. You have to remember, my ultimate use case is JARVIS.

I want to be able to say, "Hey Copilot. Open my DALL-E folder on Dropbox and find me that steampunk Tesla car I did, and then send it to Paul?"

Or "Hey Copilot. I want to move the less important meetings on my calendar to other days. Can we free me up a bit?"

Or "Hey Copilot. Are there any stocks tanking right now that I would consider buying? If so, can you see how much I can afford and buy the dip when it ticks back up? Maybe no more than $10K, though, please."

Break any one of those prompts down, and there's a lot of computing going on, but if you look at all the smaller use cases people have been doing with GPT-4, they all seem do-able. Add things like BloombergGPT and ChatSpot.ai and all the new siloed AI models and it's getting exciting.

But We Can't Get There If We Don't Get In the Lab

Learning this stuff is how you'll be ready when it becomes the norm. Remember when you first gave up your Blackberry for an iPhone? You had to learn a lot of things in a short time, and it didn't go as well as you remember with the fog of nostalgia swirling around you. It stunk. But you figured it all out. Then you saw an article. Maybe watched a YouTube.

There's zero formula for getting there. It requires your exploration and experimentation. And while you're entitled to your opinions, it's a lot harder to have a valuable opinion if you're not mucking around in here with us all.

And if you're not sure WHAT to muck about doing, I can always give you some ideas.

Chris...