AutoGPT Test and My AI Agents Effortless Programming – INSANE Progress!

today I want to show you how my AIagents take python coding to a whole newlevel when they cooperate to createsmall mini games and apps with a UI inonly two minutes I also want to show youmy auto GPT project where I try tocreate an AI Agent form with theultimate goal of creating an AI systemso let's just get started first let mejust show you a quick flow shot here ofthe system so we got a python Hub in themiddle we got two Chef GPT agents thatare hooked up to chatibility 3.5 API thesystem starts with a trigger messagethat's going to start a simulationbetween our two sharp GPT agents theyhave both access to Google so they cango outside the frame and search for infobring it back and tell the other agentabout it their goal in this system is towork together and to create a workingpython code of our description let's saywe want a game then they're gonnacooperate try to make that game and putout a working python code so that's itlet's dive deeper here we are at ourpython Hub I'm not gonna go through allof this code now that's just gonna taketoo long but we are gonna look at ouragents so we have agent 007 the role isyou are a professional softwaredeveloper with special knowledge aboutPython and games keep your messagesshort focus on the code your task is tocreate a text box to write in with a UIin Python with agent 69 your task isalso to give critique feedback to agent69 Converse in this structure responsereflection code and critique we have themirrored instructions for 69 but he isgonna critique agent 007 right I hadgood results with doing this mirroredapproach but I also want to experimentwith keeping that both the system rolesquite different so they can work offeach other's strength that's going to befor another video now let's try tocreate this text box to write in withthe UI okay so let's run this now Ibrought up a timer here so we can get anInsight of how long this takes so let mestart that over here enter and gookay so we start off with the triggermessage right so hello 007 I'm 1869 I'mhere to help you create a simpletextbook in Python we can use theThinker Library which allows us tocreate a GUI easily and here you see hisreflection it's important to make surethat we are addressing the correctperson to use a professional tone in ourconversationand then he puts out an ID for a codethis code or the code creates a simpletext box using the tincter library whichis a good starting point we don't haveany labels or buttons it could beimproved okay so agents zero come backand says thank you for creating thewindow it's important to build up theprevious code and address thelimitations in a user-friendly manner sohe looks at the code comes up with someIDs it now has more function abilitieswhich makes us more user friendlythe previous limitation were resolvedand then it comes back to agent 69 againtrying to it reflects on providingconstructive feedback and suggestionsfor improvements can lead to a betteroutcome that's his overall reflectionright and they go back and forward tryto make good ideas for the code this isthe code we end up with we are gonnatest it so let's copy this and paste itin and test it okay so here we go hereis the resultwrite somethingokay so I guess it's workingyou wroteerror text textbook is empty okay so wegot a message all about AI submit it'snot saving or anything it's a verysimple UI yeah good job now let's try tocreate a whole bunch of different thingsand see if it works now let's try tocreate a tic-tac-toe game with a UI inPython okay sounds good over here startthe timer and runokay so that was about 2 minutes and 50seconds let me copy that and test itokay let me enter all right this looksgood so we got a code here okay that wasa bit strange let me see if it worksX ah goodyeahokay so what happened nowokay so it stopped but uh can we restartitah here we go X wins I wonder why thatdidn't happen earlier okay cool thatworks let's do something else okay sonow let's try to create an app thatchecks the weather in Norway with abasic UI okay let's run that start thetimer and gookay so that was about one minute and 20seconds let me just grab this and seeexactly what this is okay so let's testthis nowuh welcome to the Norwegian this is notexactly a UI but okay welcome toInnovation weather app check uh pleaseenter name some cities in Norway okaylet's do Oslo and Bergenwow that works it's not thatsophisticated but at least it works uh Ijust fetched a free API on the web andyeah current temperature in Oslo 10degrees broken clouds 13 degrees inBergen a few clouds so we can just do aquick UI on this if we wanted to I'm notgoing to do that now let's try somethingdifferent so now let's try to give themsome Freedom here so we can create a newstrange simple game with a UI in Pythonso I don't know what this is gonna bebut let's check it out okay so run thatstart the timer and gookay so that was done this is veryexciting so let's just go all the way upon top here because I want to see howthey taught it uh hello agent I'm gladto work on this project with you surelet's get started we're kind of a gameare we building Agent Zero seems eagerto start but we need to clarify whatgame we are building agent 69 just startby setting up a simple UI that's a goodstart right and I agree with your planas for the game I was thinking we createa simple guessing game using numbers theplayers will have to guess the numberbetween 1 and 100 and they give themhints if they are too high or low andthe reflection is a simple guessing gamecould be a good start for us to work onlet's start adding witches for guessingmui we need to label and enter thewidgets so they gone on and on aboutthis they try to improve it here's theupdated code great catch okay so thecode seems to work fine but if theplayer entered an invalid input likestring instead of numbers it will raisea value error great catch we should adda error hand error that is very that isvery good so they found a weakness andthey cooperated and fixed it so rememberthey only spent three minutes on this sothey just went back and forward tried toimplement other stuff is this gonna workI don't know but let's try it so let'scopy thatand let's give it a spin okay so we gotsomething here I don't know we got twowindows let's expand this I don't knowwhat kind of window this is can I grabthat okay so here we got two windows soit's I hope you can see this it's a bitsmaller it's basically enter your guessone to hundred let's start and we havegot 10 attempts so we can choose ourdifficulty here it doesn't seems to workthough but okay so let's do 50.your guess is too high you got nineattempts remainingtoo high okay let's do 30.too high 20too high tencongratulations you guessed the rightnumberwow so let's try hard and replay okay soI guess we still have 10 attempts sothat didn't worklet's do 50. it's too low 90.too high I guess it's 75to low 80. okay so you get the pointthat was coolthat was a good idea it's very simple ofcourse but for three minutesthat's quite impressive okay so justremember this is still with chat GPT 3.5and that is not so good at coding asgpt4 soon I will have upgraded this toGTA4 I just need one more API and I'mready to go that is going to be veryinteresting I just want to finish offthis video by showing you something Icreated in Auto gbt so let's just headover to Auto GPT and take a look at thatokay so let me show you the plan I havefor this Auto GPT system so we have theAI that's called Agent Farm9000 it's an AI designed to autonomouslycreate and manage specialized GPT agentthat can generate and reproduce otheragents with the ultimate goal ofdeveloping a comprehensive AGI system soour goal number one is going to becreate an agent capable of understandingthe concepts principles and architectureof agile systems including the abilityto create manage and reproducespecialized agents goal number three isgoing to be create an agent that cancontinuously analyze optimize and refinethe replication process ensuring theresulting agent contribute effectivelygoal 4 is going to be an agent that cancoordinate and manage interactionbetween various agent within the HRSystem enabling seamless cooperationknowledge sharing and task presentationand our final goal is to create an agentresponsible for monitoring the AI AGIsystems performanceidentifying areas of improvement andimplementing unnecessary adjustments tomaintain the system stabilityscalability and adaptability so this isquite the interesting case so I think weare just gonna run it andI'm gonna run it let's do a hundredcontinuous and then we will get back andtake a look at where we standokay so Auto GPT has been running for awhile now I just wanted to take a looknow so where we stand so it started offwith this plan hereto achieve the goal of creating an agilesystem I need to start building agentsthat can an agent that can say focus onthis task and generate other agent so itstarted by creating an agent called AJFocusso this prompt is key principle ofsystems right and it wrote this wholekind of list herethen it went on to start an agent namedAGI expert trained AGI expert tounderstand the concepts principles andarchitecture of AGI systemsso yeah I started this AGI expert agentand it kind of went on to find someinformation about so message the AJFocus agent start working on a report onAGI systemsensure that the agent is activelyworking on the task and managing processokay that's cool and I wanted to start anew GPT agent named AGI designer sodesign and Implement self-replicatingagents with a diverse and complementaryskill setenable to continuous analyze andoptimize and refine the processso I really enjoy the start here I thinkthis is what I am looking for I want tobuild up agents with different kind ofskills right so now the thought was nowI created the focus on Asia expertagents I should focus on training Asiaexpert to create and manage yeahI will Begin by defining corecompetitors specialized in GPS shouldprocess to achieve this goal so itstarted doing this do nothing commandthat's not a good sign okay so we did aGoogle search for creating HR systemsthat's goodokay so it found 40 000 characters onthis website let's take a quick look atit soit was this website artificial generalintelligence on techtarget.comokay so that was not badokay I wrote that to file let's take alook at that expertise in ml and AIknowledge of fields cognitive scienceneuroscienceokay so so far so good I like this I'vegotten so far before so now we created anew an AGI managerselect a managed agent to fulfill therequired to compute it towards the goalokayso we have a managermessage the Builder now we created anAGI researcher search and browsewebsites for gather information onsystems and synthesize it into a concisereportokay that's good we want that but thenwe got to here and it kind of went offthe rails this happens sometimes for meyou can see it did the same promptforever so this is a working progress Imight have to change up the the goalhere a bit I haven't really decided whatI'm going to change I kind of think Ineed a more of an overall goal in justcreating an agile system so I'm going towork on that and I'm probably gonna do avideo next week or whenever you see thisit might be out so yeah very interestinga lot of fun with auto GPT and I gotsome cool videos coming up on this babyAGI Auto GPD Trend so look out for themif you made it all the way here maybeyou want to try out my system that Ijust show you in the beginning of thevideo if you want that check out thelink in the description below to mymembership and thank you for tuning insee you in the next one