so I think we're going to live in a world where there are going to be hundreds of millions of billions of different AI agents eventually probably more AI agents than there are people in the world from generative AI to what's known as agentic AI that agent force is what he's calling it agent force will be AI agents has become the Talk of the Town and today I'm going to explain this concept in a way that even a high school student can understand it easily let's start with analogy and then we will go into technical details
let's say you have a restaurant you hire two waiters moan and mad moan is intelligent and accurate he will go to a customer who is coming to a restaurant and then he will offer the food Manu the customer will uh order few things he will accurately KN it down and deliver it in a nice manner while moan is accurate in his work he doesn't provide extra suggestions or recommendation M on the other hand is also intelligent and accurate but he provides extra suggestions for example if a customer is ordering Nan and Indian curry and if
the weather is cold outside a mad will say why don't you order this hot tomato soup the weather is cold outside and it will really go well with this dish if the customer is coming again and again Mad will recognize their favorite dish and they will say oh last time you order harabara kbab would you like to repeat that so this way madav is going one step ahead he's an independent thinker he's providing this suggestions and he's autonomous in the way that he works the first waiter moan on the other hand is not autonomous he
will do things as directed in this analogy the second waiter mad is an AI agent the first viter moan is a traditional AI system okay so both are AI system the first one is a traditional AI system the second one is an AI system which is based on AI agent here I have the example of a regular chatboard where you ask questions such as what are your store hours what P options do you have you can also place an order this kind of chatboard can be easily built using Frameworks such as dialog flow rasa there
are llm based Frameworks too AI agent based chatboard system on the other hand will be autonomous in certain decision making for example you are visiting this p P store pandai Pizza Store every Friday evening and you're ordering same large veggie pizza with olive topping this chatboard can learn from it and when you say I want to place an order it can look at your history and say hey would you like to reorder your usual large pizza you'll say yes it will also have awareness about environment so here it is saying hey it's cold out there
do you want to add hot chocolate so all these suggestions that it is providing they are not coded into your Python program this is something that this agent is coming up on its own it can also check weather and traffic condition and say hey you want to order a pizza but there is a snowstorm and deliveries might be delayed this is like human if there is a human he will be aware about environment weather all of that and they will provide extra suggestions now let's understand a little more in technical details how regular AI chatboard
works whenever you are asking any question it will first identify an intent so let's say in my chatboard I have three type of intent General inquiry placing order refund based on the question it will say hey this is a general inquiry intent now you will be like okay you can just do exit sentence matching and you can say this is a software program where is AI but this is an AI based system because because your question might be different instead of saying what are your store hours you can say hey can you tell me when
the store is open so there is no exit word to word matching these two sentences are different but the meaning is same so if you're using any llm such as GPT or Claude Etc it will be able to match all these sentences okay you can just say that hey llm if someone ask questions which is what are your store hours or a similar question in English language then the in ENT is General inquiry once you tell that to llm next time if a person asks questions differently then also it will be able to match that
intent a second question you can ask is can you place an order with large vpa all and spinach topping it will match the intent say place order not only that it will extract the meaningful information which is what is my order size my type toppings Etc and it will call an appropriate code or API or let's say python function which you can run and place the order and insert a record in database and so on so this is a traditional chatboard this doesn't use any agent Etc if you want to learn more in detail I
have this free YouTube video where I buil the exact same chatboard using dialog flow framework now let's see how AI agent chatboard will work so first of all AI agent chatboard will have exact same capabilities as the regular chat board but in addition it will have extra capabilities so when you give a sentence it will be able to identify the intent it will extract the information called the python function API and so on but in addition it will have access to something called tools now when I say tools let's say weather API is one type
of tool so it will go check weather API and say there is a snowstorm and you know there is a traffic deliveries are delayed then it will give you this kind of response it will say hey due to snowstorm deliveries might be delayed it may have access to another tool such as a database database contains all the past records from the same customer now you can figure out that this customer is ordering same pza every Friday at 700 p.m. or evening then you will say hey would you like to reorder your usual large pizza with
olive topping so see this is intelligent and autonomous in your code you have not return return that uh when person says place an order you should uh give this as a response okay in your code you have not written that but the AI agent figures this thing out on his own and provide suggestion it can have access to web search it can have access to variety of apis so these tools are something that you can Define when you are writing code for AI agent so AI agent um solution is something for which you have to
write code obviously but you don't type all this instructions that if person places an order and if person has a same history Friday 700 p.m. large V then just repeat that okay you are just providing all these tools everything and this suggestion is something that this AI agent comes up on its own I gave an example of an AI chatboard but there are variety of AI Solutions there can be a recommendation engine there can be a document search you know you can build variety of AI Solutions chatboard is just one category of AI solutions to
summarize the main thing in AI agent is autonomy independent thinking obviously there will be some limitations on autonomy your chatbot cannot say hey I will place your order for free and your family and friends I will give everything for free that kind of autonomy you can't give to the agents so when you are building AI agents by writing code you will have some kind of control over it but within that that control the AI agent will be autonomous it's like you have a dog and you have a lease right so dog can roam around but
you have a Lee so the area where it can go that area is limited right let's say you have a 2 m long leaves or the string you know that you tie at dog's neck the dog can roam around in that 2 m radius so it is autonomous but on top of that there is some control that you're imposing there are Frameworks such as langra Microsoft autogen crew AI Etc that you can use to build AI agents I'm going to be publishing more videos on AI agents in the coming time so please keep an eye
and if you like this video then please give it a thumbs up share it with your friends if you have any comments there is a comment box below [Music]
