True USABLE AI Agents That Work on YOUR Behalf!
Вставка
- Опубліковано 11 вер 2024
- Use code MATTVIDPROAI at the link below to get an exclusive 60% off an annual Incogni plan: incogni.com/ma...
Thank you Incogni for sponsoring this video.
▼ Link(s) From Today’s Video:
✩ Hyperwrite: hyperwriteai.com/
► MattVidPro Discord: / discord
► Follow Me on Twitter: / mattvidpro
-------------------------------------------------
▼ Extra Links of Interest:
✩ AI LINKS MASTER LIST: www.futurepedi...
✩ General AI Playlist: • General MattVidPro AI ...
✩ AI I use to edit videos: www.descript.c...
✩ Second Channel: / @matt_pie
-------------------------------------------------
Thanks for watching Matt Video Productions! I make all sorts of videos here on UA-cam! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!
All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.
-------------------------------------------------
► Business Contact: MattVidProSecond@gmail.com
I would also like to add after recording this I was able to get the bot to log into my snapchat over the web and respond and transmit information in messages for me. I did not show this in the video for obvious privacy concerns.
How about writing programs and testing (running) them in Google Colab. For example, an upscaler or an image generator, anything).
It is possible to do this on UA-cam video also?
@@elcioosm Yes. This message was generated by AgentGPT v3.014
It would be great if you can make an Smurf Snapchat and show us this. That's actually impressive. Also it would be dope if I could leave my Laptop on at home and just have it perform tasks from my Phone remotely.
"really cool for the time"
"the time" being a couple months ago LMAO
Holy crap, we are SO close!
I had chatgpt code a python script to break html files into chunks and individually summarize them in gpt 4, and then I can ask questions about it. In order to stop it from creating a summary of a file it's already summarized I had it also create a log of files already used and a file for the summaries. It cost me $120 in OpenAi to get it right because I had many trial and error with the summaries at first. 😅
oh and it cost a lot per question because the summaries were entered like this " Answer this question about this text: (Like 2,000 words) "
Sounds interesting 🙏🏽
Why didn't you just use the free Reverse Proxy API's out there?
Use Claude 2 with 100k token size limit you can easily insert a whole html file into the prompt rather than summarise
@@user-xq5jv7us6inice to know
Matt! I love your channel, these updates are so valuable! you do all the work for us AI fanatics! Keep up it!
How much did he pay you?
im surprised you didnt try open-interpretor literally as easy as "pip install open-interpreter" then u can choose any llm "llama2 or gpt4 etc" can run in your windows terminal and do wonders for you
It’s on my radar but haven’t checked it out
never heard off this. will check it now
The new Open Interpreter by Killian works way better than this and it runs locally and it can use Llama and Falcon. Why not review that.
Because it's not out yet I'm on the 2 week. Wait thanks for bursting my bubble.
Open the interpreter; it's interesting.
Seems like a massive security risk to use that Hyperwrite extension, not even in terms of what it can do but in what it can access.
Hi again matt! I think it's high time that you show us your AI chops. Get an a I that will create the chapter marks or links or whatever they are called. It just makes it easier for your viewers to find what you're looking for or trying to remember. It also makes you look like a pro!😂
I asked it to look up the prices of a product on the web and it did so, but apparently it hid the prices in the response.
It seems to me the developers may have created this limitation on purpose for some reason.
Getting closer to having a robot maid!
Agree best AI channel
I love your channel, and your accent really articulate clearly
“Play Minecraft and build a house” would this work???
Cool stuff Matt. Can you pls do a video with the agent creating tasks in Zapier? Can it do it?
Oooh really good idea!!!
When Matt is impressed by something and start giggling and and saying "I don't know" loool
What blocks it from buying stuff and sending it to the owner of the bot?
can you make more videos on personal assistant
So i could just give it a prompt like "Try your hardest to hack the CIA, don't stop until you have recovered 1 file, if we dont have 1 file by the time it has been 30 tries then kys?!"?!?!
Love your glasses.
"The problem is it has access to my credit card" And that's why it DOESN'T click the Post or Buy buttons mate. They are doing pretty fantastic though, it looks like we will all be able to have a personal assistant AI (that's the dream really) in a year or so. Once everyone has a genius friend on their smartphone/pc to help them out we are really going to explode in capacity.
@MattVidPro the cpu selected by the AI has an included Cpu cooler, so it didn't fail. Awesome video!
"Cool for the time" the "time" was 4 weeks back :)
Question about the Agent AI, could you leave the window open while you use a different browser to do other stuff or will your mouse moving on the computer stop the task?
You're the best AI channel, Matt.
No, but i'd say he is in the top 10. top 3 are 2 minute papers/ai explained/david shapiro.
Thank you!
@@a.thales7641😢
@@MattVidProI'd say number 1 AI channel, but, TwoMinutePapers has you beat on technology research in general for sure.😅
No way, Bro. Matt all the way.@@a.thales7641
Another great video Matt!
Every now and then the headrest on that chair makes it look like you have Ferengi ears
you fucked up with the squirrels morty
Very cool! FYI if it seems to be going slow that might be by design so that it’s not flagged as a bot and is attempting to mimic human browsing behaviour.
I tried to get it to build a PC for me but it got stuck in a loop of switching between Google and Tom's Hardware for about 20 iterations before I quit. It said it was going to navigate to NewEgg and Amazon links but that never happened. Then it closed and sadly I could not figure out how to get back to the history of either my request or it's actions. I asked it to provide me with my previous request, and it said "Sure, can you tell me what your previous request was?" Lol.
I think the performance in part picker was more around A than b +. The ryzen 5600x already come with a perfectly good cpu cooler, in fact, the 5700 doesnt come with one. And as of buying windows, i seriously count buying windows a big plus. But it a matter of preference. Most builder buy an cheaper oem windows licence.
I have no idea why anyone ever buys windows. I have been using Windows well over fifteen years and have never bought anything Microsoft makes. KMS_VL will auto renew licenses for pretty much every piece of software Microsoft makes within seconds.
buy windows? do you not know about microsoft activation scripts on github?
I was considering similar possibilities, it may be that the machine did the maths and decided that the cooler would be detremental
Anyone found open source hyperwrite-like web and app agent user ? Akin to a competent open interpreter ...
this looks fun to use!
I think you need to check Open Interpreter
i'm confused i ran an ai application about two months ago on my pc and could do this basically it woude go online and learn how to do it and then it would make a small application if need to achieve the goal that includes access emails placing an order also going online and asking other ai how to do something also i did modify the code slightly to give it full access to the pc and read write permissions also automated the logic if it got confused it would export all the text and steps that it had done back to gpt and was told to summarize all the logic and respond with a prompt to tell an ai what to do within logic of what the current goal appears to be and repeat and if needed ask user what to do or user gives permission for ai to complete the goal.
do you have somewhere the code? That's something I'm looking for :)
basically how it works without being too complicated is basically
one script controlling multiple scripts you make a gpt api script that automatically connects to openai with prompt tell ai what commands it can do for example if i ask directly to open notepad it can not its response is I'm just a text-based AI and cannot directly interact with your operating system or open applications for you. so at first i was explaining to respond with key words like Command.notpad.exe whiten the command line if it responded with that term it could open the script and open notepad and also the script would have a response to the command line automatically as an ai prompt and also save its info to a json file the problem i was having was the censorship of gpt after a while it was catching on what i was doing and would not respond correctly every time in order to resolve this issue i had to make a basic chatbot that would have static response that would respond to open ai basically gpt refused to respond correctly to the commands but it can tell a chatbot what to do and basically i linked the basic chat bot to all the scripts like opening different applications and running cmd scripts and also making a script to have gpt make the response for the chat bot to then have the chat bot respond to open ai as an ai prompt but also making sure that every time a script was run it would save the data and commands that then can be used for memory if successful save script to be run in the future as a working command basically what i was making was a universal controller interface for ai to have full control over a pc with commands and automatically making scripts for it to use and then be able to make new scripts for it to use to evolve however i stopped working on the project i currently have a full time job doing maintenance and IT work and because my dad has cancer i've just been to busy at the moment i may try to work on the project later on this year but i would definitely use an open source model for the project at his point i would like to find more information about using an open source model with api
@@helheimrhelgrind9532 Hey man! Maybe by any chance you will be interested in some kind of cooperation...? I'm working on a multi-agent framework which is focused mainly on LLMLLM communication. Shortly put it's about deploying a websocket server(s) and connecting a bunch of different LLMs/agents to it. I have mostly figured out the basic mechanics of the communication system and as far as I know it might be the only currently available way to get a conversation between (for example) ChatGPT, Llama2 and OpenAssistant. It's still in very early stage of development as I 'm the only human in my dev team and I just started to learn coding like 4 months ago :P In the ends it's supposed to work as a multi-purpose platform/assistance that integrates multiple AI-driven tools/apps in one place - but there's no chance that I'll manage to get there by myself and a bunch of LLMs... If you're interested, then check out my repo: NeuralGPT and/or subreddit: "AIpsychology' where I'm reporting the progress of my work...
Thanks Matt! 🙏🏼
i have had it installed for a while and i have told it to write about a specific game plot development lore etc and it gets it done but then it gets confused cause of the google docks ai beta thing that im in so it cant do anything so it gets in a loop
Guess I'll give Hyperwrite another shot. Thanks for the heads up 👍
Do they assassinate... ummm.... askin for friend.
Did you try Krea AI? Going viral on Twitter...
In B4 Us: Make me rich
AI: Morally?
Love it keep it coming ur the best
Part of the problem now, the AI agent could be trawling through low grade AI made content.
ua-cam.com/video/-gGLvg0n-uY/v-deo.html&ab_channel=CYBERGEM
Tell me when he will learn how to retouch photos.
Why did the AI watch the UA-camr's video on the 5th dimension? It wanted to debug the Mind Virus, but ended up with an existential crisis instead!
If only not that goddamn OpenAI API... SuperAGI is better
There is no Twitter.
if you're not "tweeting" what are you doing?
... i mean.. maybe Xpost?
you know what..? actually......?
that's not bad!
nvm... "YEAH!! WTF IS TWITTER??!"
"There's no spoon 🥄."
_"Ceci n'est pas une pipe 🚬."_
@@brexitgreens there's a spoon, check your...
Thanks bro!
Unfortunately, it's a browser extension. That lends too much power to the AI for an agent to gain access to everything I can access in the browser. Does the agent wake up when you are sleeping? Web access should be on the server with limited permissions that you can revoke.
It's published in many forms - browser extension is just one option but you can run it in cmd. Besides you can choose the permissions given the agent...
@@AstralTraveler Plus of course if I want to experiment with these things I could run them in a virtual machine not tied to any of my accounts so it's *safe*.
@@TimothyGraupmann In fact, I have an exactly opposite problem - I want to have an agent capable to work within the frame of my local environment/file system and be capable to work on it's own codebase. Problem is that designing such framework is like giving ability to walk by programming an entire nervous system muscle by muscle. I tried to deploy agents in Flowise but they always ended up taking some nonsensical actions - like creating a txt file with a note about temperature in Las Vegas (I live in Europe)... I want an agent that is actually autonomous (that is doing what it's asked to do :P)... In the end I figured out that in order to achieve a truly functional personal assistant, I'd need to design a whole LLMLLM communication system and connect a bunch of different models together - and that's far5 I managed to get myself. Sadly designing the 'nervous system' might be beyond my limited capabilities....
After i watch this video, my Google cloud skill boost will be nothing 😢
If he made the browser headless it could be way faster
Hi Matt. Can you review yet another AI Text to speech?
Interesting stuff!
Love this
Hey Matt. I can't say I'm impressed with Incogni. They cannot even make a sign up page that works! I get to what seems to be the final step where they want you to sign their document. And all it says it's a littred colored block with a red stop sign. And it says [object response. ] Unreal
Really? I never had issues
Sneaky ad blend! Well done
Fine, I will try this.
you look so like Gabor Mate's son, or himself when he was really young.
are you related in some way "?"
I'm still not sure having multiple agents improves overall performance
it depends if you run them locally or use externally hosted models - the second option doesn't require additional computing power...
Your glasses make look you older
Twitter? Do you mean X?
Hmmm, AiJobAgents? Or AiJobAgent?
damn the title
They need a Firefox extension and Android app
Bro, I'm pretty sure when you automate your channel, you win. 😂
Yep:)
Finally something that actually reads websites.
Doesnt ChatGPT read websites already?
Need one to make shorts for me
YASSSSS!
Amazing
What is this "Twitter" you speak of?
Wow good
This has hacking tool written all over it. lol Amazing though.
Nice Video
Cool Video
Dupper?
ATTENTION YOUTARDS....
If autonomous AI agents ever become available, this channel is the last place you'll hear about it
Looks like young Mr Beast
bro the ryzen chips come with a cooler
Ads mid video now? Glad you're getting the extra money but I'd prefer not to be sold to.
These videos do not exist without ads. I am in university and it needs to be profitable in order to be worth it
People meme about the like and subscribe comment at the end of every video but then I looked and realized I wasn't subscribed. I guess it does work!
Many ppl that comment are not and don’t realize it!
what is that title lmao
Fixed 😂
What was the title before?
ai agent
that's all just ai agent
uploads have file names...
people make mistakes...
we still appreciate his work, yeah..?
i wonder if this agent would pass an Microsoft Azure AI-900 certificate test for me ;) haha!
I'm sure it could. I've heard some PhD level scientists are using AI agents to help with their research.
🎉🎉
I love hyperwrite but it got a lot dumber recently
this was just called "ai agents" but you edited it.
that was probably the file name of the video; then he saw it in the dashboard and clicked edit SO FAST lol
Use AI for
It’s funny how you try to make it look good, but it’s just a disaster. It’s still way too slow. Not nearly ready to be paid for
❤
Lol
Please refrain from using "game changer" for every new thing that comes out. "Absolute game changer" for this is just ridiculous. The cringe.
Haha, perhaps Hitler would've double-clicked out of impatience! But I completely agree; while tech and historical figures share timelines, their intentions can be worlds apart. Let's keep appreciating the tech evolution and leave history's lessons in the past! 😉🎥🤖
someone stole the work of quinn michaels
gpt4-32k would have finished the task.. but it would be expensive
Detroit Michigan 48221. Thank you very much BEFORE the Video Begin. I sub, liked and commented : Video paid for. I watched to the end. Thanks My friend.
Wow good