Innovation Through Connectivity

You Build The Context

One of the cool places I see ChatGPT and other AI tools heading, that will catapult us into even more innovation, is in their ability to create commands for software further down a pipeline.

We know that building a good prompt in ChatGPT is critical to getting a desirable result. You go from mundane and vanilla results to scary good results just by giving the AI something more for context. That is what has made GPT-4 so much better with the results...it goes from 4k tokens max in GPT-3 to 8k and even 32k in GPT-4!

It isn't something that doubles or octuples the size of your blog posts!

If you look at it as the ability for bigger questions with the same-sized results, you will see where the power is. Adding additional tokens isn't about giving you a bigger result, but about allowing you a bigger question -- more context for the AI to work with.

Context is like our life experience. Remember going away to college or on a trip, and coming back to tell others about the experience? You have to tell them a little about the location, the history, the people you met, the atmosphere of the local pubs, etc. You have to paint the picture enough so the meat of what you really want to share with them can be understood!

If you just say "I went to Europe and there was this cool hike." Then it will have a limited impact on your listeners. They may be happy for you, but won't be excited to go there one day. And they won't understand why you are so jazzed about it.

But if you explain the environment, the weather, the people, the gondola you took to get to the START of the hike, the view at the top, the friendly barista you met at the end, the smell of the bakery and coffee after the tiring hike...

Context To Questions

Come on...I got you thinking about the smells, didn't I?

From that backdrop, when you then say "My shoes sort of hurt my heels. What would you have worn?", the person can converse about it better. Anything you then talk about or ask about has that backdrop, that context, in play. Far easier for others to communicate with you, and you with them.

You aren't boosting into GPT-4 mode in your story looking for a much longer explanation of the right shoes for next time, or to compare 800 different shoes. You want to be able to give the person enough information to have the best answer for you. Might be long, might be short, might be excited or they might flake out on you and ignore the question. But you have to give them information to work with.

Ok, you get the idea. So where am I going with this around AI and connectivity and innovation?

Questions to Results That Are Contexts

You talk to ChatGPT to get a result. You want the best result. So feed it the most context you can. It will take what you give it and compare it with its whole dataset and come up with a better answer for you.

Now...the secret sauce...use that output as INPUT to another software! Get ChatGPT to build the context necessary for the next software in a chain! Have ChatGPT come out with a big output that gives another software everything it needs to make its best output!

You don't have to know how to talk to Software B, but ChatGPT will. So you give ChatGPT all kinds of context and then ask it to build out the right context for Software B. Part of those instructions will be to ask Software B to create context for Software C and then connect to it.

Connectivity like this is coming about through mediums like ChatGPT Plugins. These plugins use ChatGPT to come up with results, but also to come up with the right prompts for other software! These plugins can take your output and use it as input to another software, then return to you Software B's results.

Let's say that again with clarity. You plug your question into ChatGPT and get results from Software B. With the right plugins you can chain them together. Or find a plugin that uses multiple softwares so you end up with results from Software D!

ChatGPT --> Plugin (--> Software B --> Software C --> Software D) --> Results        

Your Results From A Web of Contexts and Results

Now imagine a web of connectivity in those parentheses above! Who knows how many softwares each one hits! You plug a context and question into ChatGPT, and it comes back with results pulled from all sorts of areas.

Think about all the data sources out there. Stocks, weather, movie times, Amazon prices, sports scores, etc. How would you pull them together to get current and meaningful data? What sort of context do you want to give ChatGPT and come back with that data?

ChatGPT goes back to 2021 with its dataset, but you can get live stock quotes from it by connecting it to other things.

I will be prototyping this soon. I am building common hooks into datasets that can be used in plugins without having to know all the coding geekiness.

For now, let your brain open up to some of the innovations that are now possible. Things never done before. Things that were never connected before but now can be, with the power of AI between them to provide much much better results than ever before!

If you are interested in building some of these connected pieces with me, or in using the conencted web of contexts in your own solutions, just DM me and we can see how it would look to work together to build something really cool!