The Growth Of Agentic AI
There is a lot of excitement on Agentic AI these days and it's barely been a couple of years since AI became mainstream. While we are still getting over our amazement for AI and all that "we" can do with it, we are now presented with Agentic AI that makes AI sound like old technology and now it is vying for our amazement, for all that "it" can do without us! But is this the next best thing for us after Sliced AI? Will it help transform our business like the AI has? Is our good old AI already outdated? And who is it for? Let's find out!
As the father of a 11yr old boy, I have been watching my son grow and mature, and one indispensable attribute of this growth is, "Learning". We Learn as we Grow, and it's a simple fact! The progress of our AI technology too is not anything different. At it's core this progress is similar to our own growth, albeit a lot faster.
It's not hard to imagine computers growing bigger and faster, just like our kids, as their capability grows. But what I am really interested is in the growth of the AI Intelligence. I see a strong similarity in how AI grows in its intelligence with how we humans grow in our intelligence.
Intelligence (Human, Non human and Artificial) grows through 3 stages,
The (Infant) Observation Stage:
This stage is characterized by a voracious appetite for data and learning. Human Babies spend most of their waking hours looking at the world around them, observing faces, colors, shapes etc. As they grow older and start exploring their world, they are still gathering more and more information about the world and how it works.
Drawing this similarity to AI, there was a time not too long ago, in the early days of AI infancy when it too started accumulating data. Our ability to generate data and capture all that data is what created the union to gave birth to our AI concepts. AI in this stage was all about relatively small forecasting tasks, like say Regressions, Correlation, Classifications etc to name a few and they were all code based built on observing existing data and generating predictions. This is similar to how human babies use their data to classify faces and shapes, correlate food to taste and forecast parent behaviors based on their tantrums. In this stage our job as elders is just to ensure our babies are collecting the right data and the generation of that data doesn't put them at any risk. For our AI infant too this was all about feeding it the right unbiased data, excluding harmful content and so on.
Any organization that wants to incorporate AI into its DNA would also give birth to the AI initiative by first gathering data and building the data pipeline.
The (Young) Obedient Stage:
This stage is characterized by dependence and obedience to superior intelligence found in parents and teachers. Youngsters seek instructions on what to do and what not to do and all the data they had collected until now as infants is now coming in handy to help them go fetch the newspaper from the living room or call to mommy or even to go out to buy milk. They always need a superior to explicitly state whats to be done and seek reassurance when something is done, as right or wrong, or, as good or bad. As a parent at this stage our job is to ensure they utilize their learning to do things that we define as "right" and probably reprimand them when they do the "wrong" thing. We teach them a wide array of topics, although not in detail, so they are exposed to a wide variety of information. This is also the most important phase of growth, something that will mould the young identity.
领英推荐
Most of our AI solutions have now entered this stage, they are like obedient students waiting for a specific prompt or instruction on whats to be done next. They do not have the power and authority to take independent decisions and complete actions on their own. These AI solutions are internal only meaning they have no ability to alter the outside world. They too need a constant human oversight not only to provide instructions but also to ensure they are operating within bounds. This stage of AI is also highly adaptable and can be used across a wide range of functions making them apt for multi use scenarios, but they are not expert in anything. Along this growth trajectory techniques like Fine Tuning and Reinforcement Learning are used to nudge the AI closer to perfection.
This is also the stage where most of the AI industry currently is. We have a range of prompt based AI solutions that perform a variety of tasks based on what we tell it to do. Once an organization has set up its data pipeline they also start building such prompt based AI solutions that can help its employees gain productivity and also make it more capable along the way.
The (Mature) Obstinate Stage
This stage is characterized by the ability to perform tasks without any further supervision. We arrive at this stage only after earning the trust of our superior stakeholders. This is also where Agentic AI solutions emerge. Just like mature humans who can accomplish tasks based on their individual decisions that require no oversight, agentic AI solutions too is capable of performing tasks on their own and learn on their own. There is no one who specifies what needs to be done apart from the goal that was established while designing the AI when it was young. The AI is capable to determine by itself on whats right and whats wrong.
This mature stage is when we humans start specializing in our skills too and become more focused on improving one's core skills. Similarly Agentic AI solutions too stop being general purpose tools and start specializing in one or two specific tasks. Any attempt to cross train them on another task that is not aligned with the original goal is not possible since they are obstinate to any change similar to how a pro football player is obstinate to consider playing cricket at a professional level. They do only one thing, but they make sure it is done right!
This makes AI agents seem exciting and highly capable of getting stuff done, but this is also its biggest drawback. Taking the glamour away Agentic AI may not be for everyone. Before considering AI agents it's important we look at what these AI Agents are capable of, and seeing if they are aligned with the company's needs. We need to weight the amount of risk the company is willing to bear for this superpower before signing up for an Agentic AI solution. Unlike the other two stages of AI only businesses and leaders who are deeply entrenched in the fabric of AI, are the ones most capable of reaping its benefits. They also need to consider putting in considerable thought on the type of Agent they would want their young AI to grow up into. This is possible only after they have spent a significant amount of time observing their AI in the infant and young stages and building trust in its capability as it matures along the way.
So whats your take on AI Agents? Do you agree that good old AI is not going anywhere soon?
#AI #Agentic #Artificial #Intelligence #Agents #Implementation #Algorithm #Data #Inputs #Outputs #Solutions #Stages
PS: As an ISB certified AI consultant with 19 years experience in Marketing, Analytics & AI, I can help SMBs explore the power of AI to benefit their business. Whether you're looking for AI implementation, data analysis, or customer insights, I can tailor solutions to your specific needs. Feel free to leave a comment or send me a direct message to discuss your AI goals.