Autonomous AI Agent Development Live
Materials — open to everyone, no sign-in
Topic: Autonomous AI Agent Development Live
Presenter: Coach Ken
Additional Resources:
Description
Join us on Saturday, March 15th, 2025, from 7 pm to 9 pm PST for an enlightening workshop titled “Autonomous AI Agent Development Live” presented by Coach Ken. This event promises a deep dive into the fascinating world of autonomous AI agent development. It’s an excellent opportunity for both beginners and experienced developers wanting to broaden their knowledge in this cutting-edge field.
This workshop will cover essential aspects of AI agent development, including designing, optimizing, and implementing autonomous agents. Coach Ken, a renowned figure in the AI field, will share insights and practical techniques based on his vast experience. Attendees will gain an understanding of how autonomous AI agents can influence various sectors, including gaming, automation, and robotics.
Don’t miss out on this valuable opportunity to upgrade your skills and stay ahead in the ever-evolving technology landscape. The knowledge and practical skills you’ll acquire at this event will empower you to develop your own AI agents and open up new avenues in your tech career. Whether you’re a student, a professional, or an AI enthusiast, this workshop is designed with you in mind. Come, learn, and be part of the future of technology!
Video Transcript
all right Uh welcome welcome to the meeting again Um yeah uh we are going to show you some of the latest development on AI agent front uh in this meeting and also tell you uh exciting course that is coming uh from Mingdaw school Uh so uh as everybody knows that uh AI uh is everywhere now and a lot of the software and hardware are adding AI features to help them to uh interact with the user uh in a more friendly way Um so people uh can just use voice or gesture to uh express their uh needs and then uh the the AI will uh translate that into actions that they can take Um so uh at the beginning we have like these chat bots uh which can uh have interesting conversations uh and uh now uh we are uh this year uh end of last year and beginning of this year uh we are seeing new development on the AI agent front Um so what’s difference between AI agent and like a uh AI enabled chatbot Uh the AI agent um can usually uh take input from uh the real world Uh it can use AI to process the information uh and then it can uh take action in the real world as well Um so uh for example it can uh like look up the today’s weather information or the stock information Uh and then in terms of uh taking actions uh they may be able to uh buy the ticket uh or uh buy the stock or uh do something in the real world to help the help the user to do the next step Uh so they are no longer limited to um just uh doing the doing the chat uh and they can do a lot more uh than just uh text input and output uh and they can do um graphics uh video sound and a lot of things uh that uh they are uh we are making the AI agents more and more powerful these days uh to handle more uh of the customer needs Uh yeah So uh so uh last week uh maybe like a bit maybe like a 10 days ago uh there is this team in China that uh released a autonomous AI agent uh that uh can do a lot of things uh within their chat It’s like a chatbot uh but it’s very capable uh it’s uh featured on Forbes and a bunch of other larger media uh and also on social media So uh it gained a lot of attention Um we we don’t have um so for people who have not tried this yet uh you can go to uh manners.im Uh right now it is in a a closed beta So uh only people they invite can actually use the product uh but they do have some uh demos that uh people can see like the capabilities of um of this chatbot So uh and they claim the difference is um it’s uh autonomous and um it can make decisions and can do more complex things uh than just a chatbot Okay Okay So there may be some internet Yeah So uh this one is one of their demos uh that they are uh for this one they are trying to do an analysis for the Tesla stock Uh normally if you are user you would be able to type in your request in this uh window Uh but right now we are just watching their demo Uh so what’s unique about uh the demo uh is they um there are a few things one is uh at the beginning So if we look at this um okay not retrieving So uh so what is uh one uh they have a high level plan on uh what to do when they want to do this research um uh in terms of uh looking up the Tesla stock performance their fundamental uh performance uh the comparable uh industry how other companies do um and then uh they can do execute different steps in to look up the internet uh there also some steps where they uh they actually write the Python I am back All right Yeah there are some steps uh where they also write the Python code uh to do the analysis So uh what uh this is different from other agent is uh they uh can generate code that did not exist uh when the bot was first built So for most of the chatbots uh the the limitation is usually when the developer writes the code and ship the product uh the capabilities already built in Um so if we need some new capability um then you will not get that on the fly So that would need to you would need to usually wait for the next revision uh of the software Uh but in terms of um menus one of the innovation is they uh also write Python code on the fly uh to handle like new requests that the uh that uh they the the chatbot has not seen before Yeah So so you see they they’re doing some debugging of their server and then um and then finally try to uh try to fix the problem um of the server and then finally the server is running Um so uh uh does so uh difference number one they can make a plan and execute it Uh number two is they uh write code on the fly Uh number three is um they have very nice uh um output uh which is not just limited to text but also uh graphics uh about the about the business Um yeah and charts Yeah charts like this Uh so uh it is very interesting Um they claim that uh it is autonomous and it’s kind of closer to general intelligence um that uh yeah so so it’s more powerful yeah I I I think it’s more powerful than the the chatbot that we have seen before uh so that’s um one of the news 10 days ago um and then um uh so uh open AI uh after after the manus was released uh soon like in just a couple days ago uh they released a a set of SDKs and agent tools to uh help people to build the agents Um so there are uh two layers they they have released One is the SDK to build the agent in agents in general Um so each agent may have multiple capabilities uh and each agent can also talk to other agents Uh so they can call Python routines and they can also call other agents to perform additional task uh using human language So uh this is the framework they have released that’s one of their release Um and then uh the other uh the other side is the capability side uh they released some uh features to allow people to search the web uh and also search local file Uh so they uh uh yeah and and then do some um local lookup for for information on your computer Uh and then uh there’s another agent which can uh drive the web browser to perform actions on the web So there are these two layers they have released uh first layer is the framework and then the second layer is some capabilities that was more difficult for normal programmers to develop So they um build it on their server and then allow us um regular programmers to call those capabilities So uh those are the uh the release uh from the open AI Um so I have walked through So this is is really new So you probably won’t find uh similar tutorial that quickly on the internet Uh but I have looked through all of their examples and I selected a few that’s uh representative So I will do some demo uh uh to get started uh about the capabilities that they have released Um with this uh demo uh uh I will try to do uh so imagine that you are a student and you want to do some research on a topic uh and then finally you need to write like your research paper Um so uh usually it’s a very time consuming process uh you try to plan your research right And then you go to the web and uh you know write down the information uh copy and paste into some document uh look at multiple sources summarize it uh and then um you need to write it in a formal document and then with the right chapter you need to format the document So it takes a lot of time uh to uh from an idea you want to research to the final kind of uh PDF presentation that you want to want to provide or the the PDF uh file that you want to provide Um uh but now with the help of the AI tools we can do it in a very short amount of time Uh that will save a lot of time Um so I built this demo using the open AI’s latest So I built on top of their prototype they have a research prototype um that will output text output of the research result and then on top of that I build additional feature to take their text output and generate a nice looking PDF file uh as the final output that you may be able to you know uh put on your website or submit to your professor Uh so let’s take a look that uh how that works Yeah I will turn off my camera here so to just to save some bandwidth and then I will share my screen All right So you’re looking at my uh at my terminal Um I have uh a few Yeah So uh tonight we we are going to have a few demos uh but just to get things started we we just see the the the demo on the search re research bot that will generate a PDF file So uh and once we have uh more time we uh I’m going to tell you some of the design patterns they that they have uh in their new in their latest release of the SDK Yeah So now uh I need to enter a topic for research Um so any any topic is fine Uh let me see what is the news today Okay So the news today is uh US tornado uh extreme weather uh leave trail of destruction Right So that’s a not uh not very good news Uh but for people who want to learn more deeply about this topic um then we can do this research Okay Uh tornado severe weather conditions Yeah So that’s uh the thing that we like to learn um uh tornadoes and other severe uh conditions uh severe weather conditions and what’s their impact So uh the uh the research agent uh actually did a uh okay so let me uh scroll up um uh made a plan to to research the using search engine uh to research the related information So it actually formed five queries Uh one is uh US tornado severe weather condition economic impact human impact US government response advancement in tornado prediction and mitigation Uh so these are very nice you know things to research right and normally that would take a long time for a human to do this research uh online So yeah so now it’s doing the search Um it’s it has completed search Yeah So let me just scroll down Um it actually wrote the report already So it’s actually pretty fast Um yeah So this is the final report Uh that is uh that was in the text format So right now you can see the text format that was uh if you download the um if you download the uh download the prototype from I mean if you download the examples uh for this SDK you will see this uh similar output Uh however I have enhanced this uh to generate a PDF file as well So let me uh take a look at which PDF file is the latest one Okay So so right now it’s uh 7:21 right So this one is the latest one See Uh okay So I let me present the other screen [Music] now Okay So uh this one is the PDF that is generated based on the search result uh from the latest open AI SDK for agent Yeah for AI agent Uh so as you can see that uh yeah so the content so it’s straightforward to actually write the final step uh to convert So what’s difficult is within their uh within their code uh we did a lot of research and then actually summarize the information um and convert that into text uh from the text to the PDF is relatively simple but uh you can see it’s a better you know much more readable format much more beautiful right so you can uh potentially post this uh on the internet um it can you generate HTML file as well uh and you can also turn it uh uh to your professor if that’s uh something that you need to write a paper about Um yeah So uh all this uh only within you know a couple minutes as we are talking that uh from an idea we want to research to a well-ritten you know sevenpage research paper uh it only took us a couple minutes So this is uh great for the for the users who want to do more research All right So let me present the the screen again Okay Okay All right Uh so next um presenter view Yeah So uh now that you know uh AI agent can be very powerful Uh it can it can do its own planning Uh it can have multiple capabilities um and then it can produce a very nice output to the uh to the user Um how do we go about learning you know how to build an AI agent Where can we get started Um so I asked my assistant today JC to help demo um how as a beginner we can build an AI agent All right So JC are you ready Thank you Uh could I have the share screen access Yeah Okay Okay Yeah you can share the screen now So uh where are the slides I have screening Uh so we’re about to build in I’m about to show you a demo on how we build a chatbot using a powerful tool known as the cursor’s code editor I’ll show you how we can use much more than just a single line of code but a list of requirements We can through the list of requirements we can generate code and then run that code through the terminal to create a chatbot which is fully functional within just a few minutes To begin we have the cursors AI here which is a powerful interface which allows us to use the claude chatbot in order to create new code So on the left this is a regular ID On the left we can see the file managing system where we have a list of files and their contents In the middle we have the contents themselves and what’s written inside of the files And on the right we have a direct place where we can talk to a AI which can help us write code I’ve already prepared a few requirements for us to use uh based on a few things which help the help guide the AI in creating a chatbot Firstly we wanted to be able to process and respond to user input through this command line interface on the bottom We wanted to be able to support the user directly talking to it We wanted to use a open AI API key which allows it to communicate with the chat GBT and we wanted to create it within this folder and have a file to run it I’ve given it an example of how to run the API itself using the last latest version and that will allow it to directly connect to chatgbt so that it can act as chatgbt in itself on the right I’ll now ask it to generate a chatbot based on these requirements and I’ll give it the context of the requirements I have written right here The chapa itself takes a while the interface itself takes a short while to run as it’s communicating directly with claude but it will produce in the end a fully functional chatbt interface as you can see here I just accept a and then it’s created a few a full file that is I believe runnable which communicates with chbt as you can see it asks for user input here getting user input and then it uses an API call to call tragedy Now I’ll show you it’s running Uh am I in the right one No I’m not CD demo origin I’m navigating to the folder where the item is stored And now it’ll run through a the chatbot which should function As this program is reasonably simple it will probably work the first time Although with more complex programs it takes quite a few lists of revisions in order to make it function As you can see it’s had a welcome message and I can ask it anything as it’s directly connected to chatt What’s the purpose of life As you can see it directly answers my queries and I can ask it anything I want Uh how do I make uh a hamburger It’s directly connected with the API So it works like this Let’s see how long it takes See it’s created a full eight list eightstep list on how to create a simple recipe Next I’ll demonstrate a blueprint of how Next I’ll demonstrate a thing I’ve worked around two hours total on which is pretty short for a development time which is specifically intended for a traveler going to Europe This travel assistant robots directly connects uh allows the user to figure out exchange rates between any any uh currency into other currencies and also allows the user to find the weather at any location in the world It isn’t that much more complicated However it directly connects three different APIs First of which is the open API the open AI key for the chatbot itself the weather API key which I believe uses open weather maps to get weather location of anywhere and the exchange key which allows it to get the exchange rates of anywhere in the world I’ll now ask it a few questions along those lines We can ask for example through this chatbot which is just a short extension on top of this main one which only took a few minutes to generate and this one only took an hour or so to create We can ask it questions like what’s the weather in Moscow for example or what’s the most important thing to know when you go to school As you can see it’s generated a few things If you want to ask for any cities I can go check the weather through this to show you how it works All right Uh we can uh see if uh anyone on in the audience want to check any cities weather Uh how about we try Madrid Madrid Spain Earlier I was doing a little bit of debugging but the entire thing is just through the uh ChachiBT interface which allows a lot of flexibility The weather in Madrid is 5.39 degrees Celsius scattered clouds and the humidity is 882% That’s really high All right All right Anyone from the audience Uh okay Yeah Yeah Great Yeah Thank you JC for the demo Uh let me take over the presentation again All right Thank you I will stop screen sharing and unmute Thank you Uh currently you’re muted Is that intended Oh Uh yeah sorry Um okay So so let me come back Yeah Uh yeah thanks uh thanks JC for the demo Uh so as you can see that uh building the AI agent um is like a stepbystep process Uh we can um we can start from the very basic agent that can chat with us uh using text and then we can add more capabilities uh to the agent Um as uh JC has demonstrated it can call external systems to do to look up information or to perform actions Um and then uh based on our product requirement you know we can get add more capabilities and uh in the past it will take it would have taken us a very long time to develop such an AI agent Uh but now there are two things that help us develop it much faster uh one is uh there are a lot of companies that build the APIs um that has AI capabilities uh that uh we don’t have to code everything ourselves uh including you know natural language parsing and generating the answer um and getting the weather and all all these things right so we can call the APIs to perform uh perform these actions uh number two is uh there’s uh a wave of you know intelligent development environment uh such as the cursor AI uh that allows us to write code really fast Um in instead of uh you know remembering the all the design patterns how to write call function um you know fix the fix the compilation error etc we can chat with the uh cursor AI editor There are a couple other editors that like that Yeah So so they we can just chat with the the editor and then they will generate code for us and if uh it does not work uh it will try to debug debug the code Uh so that actually it doesn’t take the work to zero It’s not like one sentence you you will be able to you know uh generate a whole application but it makes it much faster Uh so uh you can generate you can create a lot more code with a relatively short amount of time Um so the the barrier is much much lower and I and uh I I can see that um more more and more people will just you know code their own uh product their own AI uh so uh that will and I know that a lot of companies are also pushing their employees to use the AI tools and this is one of the front that we should uh we should evolve with the with the world and try to learn as fast as we can uh how to leverage these tools and how to leverage these AI capabilities Um so uh yeah so uh back to the AI release from the open AI uh they have released the agent SDK and a number of uh AI agent capabilities Um I have studied uh most of the examples from the SDK and I uh there are a few design patterns that are uh very interesting Um out of all the design pattern they yeah out of all the design patterns I selected three design patterns that I feel are very interesting uh to showcase Uh so so let’s walk through that together Uh so the first design pattern is uh code as tool Uh so basically the AI agent can interface with humans using natural language Uh and then after they understand what the human wants they can call existing code uh to generate the actions or to look up information um to uh to to perform those actions Um so code uh becomes a tool for the AI agent Uh so let’s uh demonstrate how that works as two Uh so first uh let me show you uh how the code looks like Uh so right now uh I just created um two simple functions One is like a generate random number the other is a multiply function Um and then we I uh uh added a yeah created a simple agent uh using these uh instructions uh which are kind of based on natural language right So uh one uh yeah so it should answer like what’s the total of 50 items multiply 25 by random So essentially just uh leverage the multiply function and the get random function and create uh create an output Um so uh the agent uh has this argument tools equal to something right So and then you we can put in a couple function names uh and it can actually pass these function signatures and generate reasonable calls to these functions So try this one Okay So enter your calculation queries For example I want to you know 60 random So oh uh 60 by 57 Okay It looks like uh it has some error like not calling the right function But uh let’s try again And randoms Oh okay Multiply 10 by 50 Oh okay Now I understand what it uh tried to do So uh my my language is ambiguous When I say 10 randoms I wanted to do like generate one random number and multiply by 10 However due to I say 10 randoms it actually generated uh 10 random numbers Uh and then he also did the multiply 10 by the first random number 85 Um so that’s a sort of reasonable interpretation of what I want Maybe I can be more specific Let’s see if that uh actually works better Generate a random number and multiply by 10 Uh what Yeah So the random number is one and then multiply by 10 equal to Okay Yeah 73 I multiply 10 by 73 and get to it gets to 730 So that’s a reasonable uh output Uh so this one demonstrates uh that uh the bots can call call APIs Um and that uh would allow us uh you can imagine that it can be more flexible than just hard coding uh what to do uh it can actually interpret uh the user’s input and then uh perform uh corresponding actions So this the first pattern okay and then the second pattern that is interesting in their uh SDK uh is manager and team Uh so that means we can have uh one main agent that takes care of interacting with the user and then that agent can actually call different agents um to perform the right action So imagine a team where you can uh have experts on in different areas uh and then uh you have a manager who takes customers request and then dispatch to these experts Um so we can see how that works uh in in their code Uh so this uh so this one uh is uh there’s an agent uh I would say maybe agency like there’s a agent team uh there’s one manager uh who is in charge of talking to the customer and then there are three experts uh in terms of translating uh language from uh I think there’s like Spanish French and there’s one more let me see which language there Uh yeah so one is Spanish one is French one is Italian Yeah So you can find this um in the release of the SDK as well Uh but uh since not all of us have a lot of time so I’m just helping you to see this demo uh since I have already installed and tried it out Um so let’s say man to Spanish Okay So let me see Yeah Okay So uh uh this uh manager agent called the Spanish expert and then translated man into Spanish Um we can see if uh try some something that’s unexpected right man translate man into English because it’s already in English Um in the in the instruction uh I asked the manager to re politely refuse any request that is not these three languages So it actually resisted this request because man is already English So it did not call any agent just says uh uh I’m sorry but I can only assist translation into Spanish French and Italian Uh so uh so that’s a pretty interesting pattern uh where we form a team of agents and we have one person uh in charge who is the manager So that’s the second pattern that I uh that that looks interesting uh in their example Uh and then uh there is a third design pattern uh which is student and coach pattern uh which means the student So in the previous pattern we can see the manager actually requests the team member to do something um and then the uh and then the team member will perform the action just uh so so the manager is like proactive and then the team member is reactive Um in real life there are other situations where we do something proactively ourselves and then we ask our co-worker for feedback or customers for feedback or our friends for feedback Right So this is kind of like a student and coach situation where the student proactively performs something and then the coach try to correct or give some suggestions to the student and then the student can actually revise uh based on what the coach uh what the coach’s advice is So we can see that pattern in the code as well Uh so that’s a the the third uh third design pattern that uh looks quite interesting in their in their set of examples Yeah So in this example um there are two parties uh the student uh so so just to take our analogy to this um example there are two uh agents in this uh in this code uh one is uh a student who can write essay uh and then write a story uh and then the other is a a judge uh who will look at the output of the student and then uh give some feedback Uh and then the student can actually uh the student can we will actually revise based on uh what the judge says uh what the uh what suggestions the judge has So uh let’s uh see what kind of story would you like to hear Sorry So let’s say real life story of Yakma My connection is a bit slow today So sometimes I see these uh error messages But uh it should it should come out Okay uh uh so you can see the first uh first output is from the student uh who writes the story Jackma and then despite a failures in school early life and then there’s a trip to US and then uh Jack started Alibaba and then there’s a kind of success story um and then the the judge says uh this story needs improvement um to better align with heroes journey framework Oh uh one one um detail that I let the judge know but I did not let the student know is uh there is a uh there’s this uh story framework uh called um a hero of a thousand faces So essentially yeah there’s a book called a hero of a thousand faces So essentially it says um all the heroes just uh in different books uh just follow the same journey Uh it’s kind of like the hero learns something uh the hero leaves the comfort zone and then the hero faces a challenge Um and then the hero fails and then the hero uh learned something um uh something from like a old person uh a secret uh secret path to success Uh so essentially advisor or more senior person uh and then apply the knowledge to the challenge and then finally the hero succeeds uh and then the hero goes back to the hometown uh where the story starts again So it’s kind of like um uh a a cycle of ups and downs in our life Uh that many many books actually use that framework So I tell the judge to use the hero of a thousand faces as the framework but I did not tell the student so the student did not know ahead of time Um so actually when you see these uh feedback from the evaluator which is the judge um you see uh some some like call to adventure meeting the mentor these are looks a little bit strange it’s just because my my prompt to the judge is using that uh using that framework right so it gives some feedback to the student um and then uh uh okay so Jackma hero’s journey and then uh the the student actually start to realize Oh the judge is actually using that framework I actually don’t know where is uh oh yeah to better align with heroes journey framework So the student actually start to realize the judge doesn’t want this uh format but they really want to align with this hero’s journey framework and then and then so so the student try to change this to uh like fall to adventure using this structure uh transforming yeah trial So so meeting the mentor and then they use the uh so so Jack has a failure which is the China pages failed uh and then uh Alibaba also face significant obstacles uh and uh uh but uh eventually he persever persevered and then went out in that in that journey and then uh became a from a entrepreneur into a leader And then this uh praises to to Jack Correct So so you use use this uh feedback from the judge and then let’s see if the judge Yeah the judge says still need more uh more improvement So the judge is still not happy with the students uh output and then see if uh yeah the student so essentially there’s this loop uh transition into yeah so uh the judge I think asked for more details here Um and then the student actually start to put in some more detail like the year num uh year number and so forth So you can see they just go back and forth uh uh to improve this essay Um so this is uh and then in the end the judge was happy with the the output Um you can see the output is actually quite uh it’s much better than the initial version uh which is a uh kind of a like a plain version of the uh of the story kind of like a news a news article or just a Wikipedia article Uh but uh in the end it uh it looks more like a novel something that’s interesting to read uh something like a a story that you you could tell to a to a kid Um so um so this is the the the pattern that uh we can see uh that the that we can write code to to to let the AI to use uh which is the student and coach pattern um where the student performs an action and then coach gives feedback and then the student improves and then coach gives a little bit more feedback uh and then eventually get get to a better product Um so uh yeah so so you can see that uh in humans life we have this um patterns of teams we have the pattern of coaches uh all of these can actually be implemented using the AI agent framework Yeah So that’s the news from the you know the industry uh where we are uh up to date to uh for this AI agent development Of course um OpenAI is not the first company to introduce this concept Uh there are uh quite a few other AI agent framework as well like L lang chain and there are a few others that could uh allow us to uh one let the agent interact with the real world Second uh agents can talk to each other and interact with each other uh similar to how uh real life human being would interact with uh our co-workers our friends our our customers Yeah So this is the the third design pattern uh for uh that’s inside the open AI release Uh there there is a uh there is a user or yeah there’s an audience member who said uh research project how about research subject of molecular mechanism uh of aging uh I can actually initiate that uh in the research using the research agent and then we can come back because uh sometimes it takes a couple minutes so I don’t want to let everybody wait but we can just initiate this uh request like to research uh more mechanism of aging Okay Okay Yeah So so it’s doing this search and then uh I can yeah five searches uh I can uh come back later to this Okay Uh yeah since all of you are here yeah thank thank you for coming to the event and uh we can all learn you know the latest development of AI AI agent together Um we uh we are planning to have a course uh starting end of this month to uh uh to help beginners uh or people who have not been into the you know the AI area uh to learn how to code AI agents So the target audience yeah as I mentioned target audience is beginner developers uh uh people who are enthusiastic about this field uh who may have some so probably people should have some coding background uh any language will be fine uh but they want to learn how to develop AI and AI agents Uh we plan to have uh five uh five evenings uh to go over this course uh uh during this course we are going to create a uh AI agent together uh we are going to yeah as outline says uh you know the we’re going to set set it up and then we uh have a basic chatbot and then we add more capabilities to this uh to this AI agent uh through adding more APIs uh text based APIs data APIs uh image APIs um and then uh How to we also uh discuss a little bit how to use it in real environment where we need to keep it safe like uh have appropriate content um make sure that it’s not uh there’s no malicious out there’s no malicious user who can attack it Uh so uh we have some uh essentially like a basic outline of how to build an AI agent Um if you are interested to join us uh you are very welcome to sign up for our course Uh it should be uh pretty friendly to beginners Um and if you are more advanced uh we have some uh optional offline material such as you know how to how to uh use source control how to work with a team uh how to use the uh web how to get information or perform action on the web Uh so those are optional offline materials for more advanced learners Um we are planning to start this course uh at the end of this month and you’re welcome to join us Um yeah so uh here are the QR codes for our uh for us uh on the left hand side is the uh it’s a wechathat group uh if you use if you’re on WeChat uh you can use this uh you can use this QR code and then uh in the middle is our linking group uh so I have um Mingda AI linking group so you can join us on linking uh there’s a discord channel uh and then there is also uh the sign up sign up link for the AI agent course uh which is going to be five evenings uh we are going to build the agent together and it’s very will be very friendly to beginning developers Uh so the audience uh actually asked a question Uh okay So there was Okay So it did let me go back to this research of uh of the molecular uh effect on aging Um and then we can see the PDF file as well Okay Uh so as you see there’s a nice table of content and then it actually uh went into pretty good detail about you know how the aging works uh yeah the mechanics of aging and uh in diff from different angles uh cellular hallmarks new nutrient sensing inflammation recent development of therapeutic interventions future directions and conclusion Uh yeah so it’s a looks like a somewhat serious I don’t know it’s probably uh yeah medium serious about you know a paper that you can find using the internet uh and you can probably write a write a paper if you you have time to write a paper like that but uh with AI you can just do that within a couple minutes all right uh and then uh Dr Y asked in the code of your manager and team example do you call the manager and use if condition to decide which agent to use or the manager call AI to decide which agent to use Uh yeah so the manager calls AI to decide which agent to use Uh in the code there’s actually no uh uh there’s no if statement agent S2 So you can see uh in the code uh maybe the font is kind of small Let’s make it bigger Now it’s big Okay So there are these uh three agents and then there is this uh kind of manager agent Uh you are translation agent If you ask multiple translation you resolve uh never translate on your own You always use the provider to politely refuse So there’s actually no if statement inside this to decide whether it’s going into Spanish or French or Italian Um and uh you can see the main function Yeah there’s no there’s essentially no no if statement here It’s all based on the AI Yeah All right All right Yeah Uh let me see if we have any other audience questions Uh again this is the all the QR codes to join us Uh and oh yeah uh another thing is tomorrow we are going to have another event uh where we are uh going to discuss uh retrieval augmented generation Uh so yeah we invited um Cindy coach Cindy uh who is a director of machine learning uh to tell us how we can uh allow the AI to retrieve vast amount of information uh from the internet or from your private network and also resolve the problem of hallucination uh because in uh some cases where the AI does not have an answer um it sometimes hallucinates and gives you you know the wrong product or discount that does not even exist Uh so um tomorrow at 700 p.m we are going to have another event uh for rag retrieval augmented generation Uh you’re all welcome to join us Uh and of course um uh very welcome to join our course our group Yeah you can join our group Uh you you don’t have to commit to sign up for our course Um if you sign up for our course uh within a couple days uh let’s say by um by Tuesday end of Tuesday uh we can still give a discounted price Uh so the normal price of the the course of the five evening would be $500 But if you sign up in uh uh before the end of Tuesday uh we can still uh provide a discounted price of $400 Uh and this is the first time that we are uh we are teaching this course So uh we are going to you know learn together Um thank you for your support Yeah And uh let’s see uh does anyone want to ask any questions All right Yeah If not uh thank you everyone for joining the the meeting today Uh it’s great that uh we learned this together with uh you know the latest development on the AI agent front and I hope to see you tomorrow uh in another exciting event where we’re going to discuss retrieval augmented generation uh and uh you’re welcome to join our class and tomorrow the uh coach Cindy is also going to have another uh tell you about another uh upcoming course uh on on retrieval augmented generation Thank you everyone I will see you in uh see you tomorrow or in our next event