Intent Driven AI for Non-Engineers
Intent Driven AI - What is the big deal?

Intent Driven AI for Non-Engineers

People speak about intent-driven AI but what is the big deal?

I try to explain for non-engineers, let me know how I do.

TL;DR: It is not about what you say, it is about what you no longer have to say.

What is intent? The word puzzle

Word Puzzle

I am playing a word game. I have one word remaining and I cannot guess what the word is. I decide to "get help" (otherwise known as "cheating"). My first emotion is to ask Google the following

Find me a 6 letter word where the 4th letter is an "I" using the following 6 letters - "A, I, S, C, N, O".

Here is the response

Google Search Results

It does not give me an answer.

Interestingly I did not do this.

Instead, I first uploaded a screenshot of the game to ChatGTP and asked the question

What is the 6 letter word?

And it answered - "CASINO".

Completing the word puzzle

A big deal!

I look at this and I am astonished. I am a software engineer by birth. How would I traditionally code for a computer to understand this?

Consider what the computer needs to understand:

  1. This is an image containing a word puzzle
  2. Letters are being filled in the boxes in the image
  3. Where letters are shared, they have to work for all words they are shared with
  4. There is a circle at the bottom
  5. In the circle, there are 6 circles with a letter in each
  6. The letters in the circle are the letters I can use to fill the boxes.
  7. There is one row of boxes that do not have letters
  8. The 4th letter is a shared box containing the letter "I".
  9. Therefore I need to find a word of length 6 letters where the 4th letter is an "I" and I can only use the letters "A, I, S, C, N, O".
  10. Return me a word that works

I did not give ChatGTP any of this. Instead, all I said was

What is the 6 letter word?

And it gave me the right answer

CASINO

That is the difference between intent-based interactions and action interactions. ChatGTP understands what it is looking at and can answer questions intelligently. To date, we have been trained to do action interactions with machines because they do not understand such a higher-level context.

Now they do and that changes everything.

My first thought was to try ChatGTP rather than Google which is very bad for Google. We use Google a huge amount to answer intent-based questions, but we do not ask Google intent-based questions, we ask Google questions after we have processed the intent into a statement we believe Google will understand.

The amount we search on Google is decreasing.

A blank puzzle, an open question

Blank word puzzle

Next, I started a new game with no words completed. I uploaded the blank puzzle to ChatGTP and asked the question:

Complete this puzzle

And it did.

Complete word puzzle answer

In addition to all the above understanding, ChatGTP needs to understand that it is looking at a word puzzle where the task (or intent) is to fill all the boxes using the letters in the circles at the bottom and where shared boxes need to have letters shared by all words involved.

Two entities exist on the planet that can do this

  1. Humans
  2. Generative AI

Five years ago there was one. This is why this is a big deal.

I asked ChatGTP some additional questions

Have you seen this puzzle before?

ChatGTP has not "seen" this specific puzzle before but it understands the general concept of grid-based word puzzle games, it can extrapolate the likely gameplay from the screenshot and it can correctly solve the puzzle.

Is ChatGTP displaying intuition?

Is ChatGTP doing what humans do? I also had not seen this game before the first I played it. Yet I used the same logic ChatGTP did.

And to complete the conversation...

How did you know?

So what...

When we work, we spend our whole time trying to solve puzzles. The difference is that these are business-related puzzles. And we translate our true intent into questions we can get answered from our traditional systems.

Example procedural questions: What is the status of base station X, what are all the alarms in region Y? Is there any maintenance work currently in region Z?

We ask these questions to try and get an answer for our true intent: Why does this customer have a problem?

Or: What do I need to do to increase our number of customers by 10% in the next quarter?

The whole conversation can change from saying what we need to do to asking what we are trying to achieve.

In conclusion

Two entities exist on the planet that can do this

  1. Humans
  2. Generative AI

Five years ago there was one. This is why this is a big deal.

Would anyone like to post how it does it?

Then it will save me from writing the next article...







Aashu Virmani

Sales & Marketing Executive | Telecoms and Wireless Expert | Deep Tech Innovator | Lifetime student of golf

1mo

This is freaky good stuff, Geoff. A bit scary perhaps. Much like you, I also grew up with procedural programming languages, and SQL being declarative (tell it "what" you want vs. "how" to fetch the answer) was a big deal at the time. Now we are onto an entirely different paradigm when interfacing with (semi) sentient computers. Perhaps soon enough, we will be able to start asking the "why" questions that impact society, and expect a reasonable answer.. e.g. "Why does half the humanity does not believe in climate change?", or "why do you think it is near impossible to create lasting peace in the middle east?" It will be an interesting exercise to observe a massive model trained on all known content to generate insights for us that will better the best single individual point of view. Thanks for sharing.

Paul Rhodes

Builder of Open vRAN, Small Cell and EdgeAI Networks

1mo

Great explanation. In this case you have a way of checking the AI performance and agree with the quality. But how do we check for the equivalent of '6-Fingers GenAI' in non-visual contexts where AI is consistently making mistakes which seem obvious when pointed out but seemed correct at first glance?.

Tamer Alaa

Strategic Technical Pre-Sales Director | Project Management & Resource Manager | Experience in 6G research Program management, Agentic AI & Automation Transformation leader| | EMEA Region Focus

1mo

Your analysis is spot on. The real revolution will start when we have a real AI OS ( operating system ) .i believe this should be next step .

Andrew E Scott GAICD

Global technology strategist and executive | CTO | Product Development | Innovation | Emerging tech | Architecture | Startups | Strategy | Advisory

1mo

Amazing stuff! Two additional points to note: i) the answers to the "how" questions are made-up by ChatGPT based on what it has learned about how a human would answer those questions, and ii) if/when web search declines, the online advertising industry is going to have the rug pulled out of it.

Vance Heredia

Client Partner @ IBM | Driving public sector innovation using cloud, GenAI, AI & ML

1mo

It is treaply fascinating what the GenAI tools can do. I did a similar activity using Claude, but asking it to describe a picture where people were wearing costumes of each other over different decades. Didn’t just identify the people and the costumes but was also able explain the potential societal reasons for this over time.

To view or add a comment, sign in

Insights from the community

Others also viewed

Explore topics