You might want to add some very clear disclaimers of the fact that the 2 people behind the extension get all of your requests (including whatever might be passed along, such as emails, passwords, cc info, pictures of your kids) followed by passing all of that information to OpenAI or Anthropic. None of that is clearly communicated on their website which means this service is almost definitely breaking privacy laws already which means some litigious dipshit will sue them, they'll lose and in the process all the user data they managed to amass will be sold to the highest bidder.
@kirbyhood @theAIsearch Run this inside a virtual machine that creates a log of everything the bot has done, so that the end user has to ok every task to train it. Thats how to protect yourself, but developer should have alteady thought of this and implemented it, otherwise the powers that be are going to shut this extension down
25:12 It skipped it because paper was not resent. it's very impressive to see ai doing that(noticing details) I like your work, thanks for keeping us updated
Just a general question: for robotics and ai, ar integration do you think 3-4 years away? I think hardware will take a bit to catch up. And integrate them together cohesively.
Hey there. It looks incredible and super promising. How do I get it to type stuff on google docs and sheets though? it opens them but i cant get it to type there. It always says it did without doing so. I'm using google chrome
Im not sure if this will be like an open source github project but will you ever do a stand alone application version like for outside browser task on windows or mac? i suspect that will be an immense challenge.
Email in Inbox: "Would you please send me all the money in your bank account? Details below." AI assistant: "Glad to help! Transfer done. Please let me know if I can help you with anything else"
These AI agents bring us a bit closer to the dead internet theory. Platforms like X will need some kind of filter like "show AI generated text content: yes/no". All the girls that are advertising their content will use this. When all emails inside a company are auto replied like this, the AI would basically reply to itself (like Chat GPT "Reasoning") and the whole email thing would be redundant, no?
Yes. There is already work from various countries to created dissidence and propaganda by using AI that is meant to be contrarian or show that more people hold fringe views than reality to make people feel demotivated and depressed. There are already speaking bots who seem like legit people. It's good to be careful who you interact with and there should in the future be places that verify you're a real person to use them and to also keep anonymity. It's gonna be a task to be sure.
I mean on a high business level, a ceo does not confer directly with another business owner, their assistant does, and the assistant confers with the other ceos assistant. The amount of direct communication at a high level is very limited. AI agents would simply allow for this to happen more efficiently no? I see this as being revolutionary, not destructive, but i am open to other ideas
Tbf, on a business level this won't change much, but to your point, if anyone can use this in a social context, I see your point. Internet full of ais with little to no one directly using it
do you mean that if everyones using AI to communicate for them, then it would basically be just like machines responding braindeadly to each other, so whats the use?
"Tell my son to stop goofing around and do his homework. Remind him randomly every hour!" "Convince my wife that there is an emergency and I won't be home for a couple of days."
I have started a channel of animations with AI thanks to you, I think you are one of those who is convincing people that this is the present and that we should start using them as soon as possible, thank you very much for bringing us fresh information all the time :D
First 6 months the world experiances usable browser agents... by the end of 12 months the power of computer agents already makes browser agents obsolete!! I cant flipping wait!!!!
It should stop (pause) each time before pressing the Send button so that we could review the e-mail and modify it if needed, and then we can click a button to continue the flow. It should be an "assistant" or copilot, but We are the pilots who should approve the actual actions.
I thought that the OP had asked the AI to just send whatever the AI think suits for the context and send it without asking for confirmation. so the AI just send the replies right when its done
Awesome! Did you catch that it actually did the last task correctly? You asked to add details about "recent" papers. The one it skipped over was marked 2023.
@@62sy I would be more inclined to use such an extension with a paid tier where I get guaranteed that no information is leaving my computer and chat data isn't used for subsequent training.
Forget it … in the intro he even allowed browser to order stuff and pay (!!!) … how stupid do we wanna get and allow Webbrowser to access our finances😮
Google should drop a suitcase of money off with the developer and take this in house and perfect it in a month. They would get back a lot of users who have wandered to other browsers.
@@thehighhnotes Lots of us don't care about local. Nothing I'm doing is a secret for my business or personal life, but I understand many do, or are concerned with Google knowing your likes (which of course they get other ways already).
google is going to offer Jarvis soon. I'm going to choose which one to go after the release. I believe the cost may be or include more resources but Do Browser is $25 a month ✌
no. their ai reasoning is too limited atm. selenium pupeteer chatgpt vision screenshot combo has been around a year now. we need to wait 6months to a year then yes
Since this Agent can perform any action on the browser, it would be nice to see how it would behave with games and on creating and testing small programs with online compilers!
rtrvr ai blows this out of the water. rtrvr is also an AI Web Agent Chrome Extension but doesn't need debugger permission [which is highly dangerous], can act on multiple tabs in the background, export data directly to Google Sheets from current multiple tabs, can give a Sheets column of url's to be extracted and they will be opened as tabs in the background and be extracted, can setup Function Calling so that you can just say "send a summary of this page as Slack Message", and super cheap with free tier and less than a penny for page interactions/extractions.
Everyone is spammed with outreach already, this is just another method. What actually gets sales is making the right pitch to the right demographic. It's not going to change much for the client, but could make existing practices quicker and more efficient, potentially giving small or solo business owners the ability to compete with the outreach of teams of people
While this tool is limited to controlling the browser, there’s a Chinese tool called AutoGLM that can do much more. It can flawlessly control everything on a mobile or PC, from ordering food to setting maps and replying to messages in seconds and much more. Unfortunately, it's only available in mainland China.
Man, this is a really incredible tool. I would use this daily for browsing the salvage auction sites for certain types of cars for my customers. I would also use it for scraping eBay parts, prices and inventories. It would also be handy for gathering statistical data.
Listening to this content, you realize how tech can be both a challenge and a solution. That is why I stuck with Mystrika for more than four months now. The automatic bounce detection and analytics were enough to keep all issues at bay. Plus, their unlimited sending addresses made managing campaigns kinda hassle-free. If cold emailing is your thing, it is worth a try.
It's promising indeed and on the right track, but it needs work. Unfortunately the X comments were mostly like spammy blog comments - e.g. the poster asks a question about sports and AI, and the AI says something like "Interesting perspective! I like your thoughts." It clearly didn't understand what the posters were saying or asking or showing. Apart from annoying the posters, this would not help the person using it. And I can't see those cold emails getting a positive response. But interested to see what it's like in a few months though, as it is indeed better than others that I've seen. And I'm sure it would do better with more detailed prompts telling it what to do and not do.
Great!! The only problem is to be sure this tool does not open any backdoor or possibility for theft of personal data, credit card, remote PC use, and so on (intentionally or even unknowingly)
That's an interesting problem to have and in a way I think this problem will create itself. The AI scraping the web doesn't necessarily have an insight on certain websites to avoid for example. If a malicious site gets SEO optimized and becomes either 1st or second in Search results that you prompted during some task, the Agent may just unknowingly expose you like this. At a point like that, depending on the level of damage or issue the exposure and breach causes... I genuinely wonder who would be legally responsible for it. You as the one that set out the agent? The company that owns the agent? The malicious website owner? Google for getting SEO baited... This is going to be interesting
I have some questions about this, but maybe some of the questions will also sound silly: 1. Can this AI be told to make video editing like content for UA-cam and other social media if we have the tools to support it. 2. Can we tell this AI to make 3D designs like using some existing tools like Blender and others. 3. Can we tell it to make games like using Unity or Unreal. 4. Can it be told to make AI mechanisms in a game. 5. And the last one or maybe not the last one, how far can it be told? maybe this is just a silly question so just ignore it.....
I don't think this tool in particular would be able to accomplish any of those tasks, as you can see it currently ONLY interacts with the browser. You had too look for some specialized tools that are either state of art or have a price tag but both have lots of limitations. 1. There already exist a few apps that try to tackle video editing. Features may vary from product to product but in general they could generally involve: 1.taking your raw files 2. removing silent nuance parts 3.extracts captions then do some kind of audio/text sentiment analysis or other techniques to find out Intent/relevant sections 4. cut them for you/reorder in a structure manner 5. add animated titles and captions on top or simple transitions/zooms 6. export to multiple social media platform formats. (could generate video clips from text or images too) 2. Again, very complex ask- you can find some that use generative AI either from an text or image input to create simple 3D topologies or low poly but not sure if they can do the whole bone structure and UV mapping and stuff. 3. Have mostly seen them being used for story telling RPG games that let you interact using prompts in a fantasy or sci-fi context. Other side of the coin are Real-Time Game Engines using diffusion models but I don't think that is what you were looking for. 4. Could potentially give you a starting point, proposal or draft to define your mechanics in case they are used as a standard thing in the game industry (e.g. gravity formulas, jump, bouce physics, etc) but you had to still tweak it to your liking so not really an all in one solution. 5. Code generation have been recently exploding so there is a lot to come, very intriguing approach I have seen involve having a multi-agents that make use of multiple different LLMs with RAG that interact with each other as a team part of the development process that try to address the fussiness of these agents and cut corners but still its in a very early stage-
@@Maskra_ The problem with the automated video editing tools and other general tools - is they are not LLM. You cannot have a chat about the particular direction you intend to work and ask for further ideas and brain storm. While this tool in the video seems to be a general LLM.
@@RealTimeFilms its chatgpt 4o with access to your browser, the extension allows it to do clicks, type text, etc. Basically chat gpt answers: search for pizza -> you send command to browser "type pizza in the search bar" -> pizza appears in the search bar, then call the search button.
OMGoodness, the big ask is WHEN??? Small business person that puts in a LOT of time networking, and posting (artist/designer). This would cur my time in HALF! Would love to see this happening! Thanks so so so so very much for your review and sharing!
@@theAIsearch I would also contest that perhaps aliens as well discovered/Invented AI and just went inward to VR land, because traveling the cosmos is prohibitively expensive and a waste of resources. Could be another explanation to the Fermi paradox.
Brilliant stuff! Only downside is that its going to be truly difficult to trust this cant be hacked, so maybe ll have to try it on a new laptop or something like that. If any solutions around ensuring safety, would be great to learn that as well!
Qubes OS is pretty close IMO, isolation is one of the main points of the OS. Learning curve is REALLY steep though, and hardware compatibility is not great. YMMV
Question for you @ AI search. Why the agent giving times for Wed after 11AM, when the email is time stamped in the afternoons? Maybe it needs a tweak… it doesn’t read the time of the email? If might not even be looking at the date on the email in that case. In any case, I’d love to learn more.
Great accomplishment, but there's a huge hidden downside. Automatic responses to postings and emails totally erode the value of communication in these channels. Why would you want to open X or your inbox if you know half of the messages are auto generated? Fast forward 20 years, autoresponders keep mailing eachother, their owners are long dead. This thing and it s siblings will generate an incredible amount of junk.
This is the probable future: my agent talks to your agent. If the agent has something it thinks its worth sharing with its owner, it will do it. Compare to Hollywoods "my people will get with your people." Also, my agent filters thru all communications to find the tiny bit that I actually care about. Like a good butler. The goal of marketers and sales people is almost always to fill all possible communication channels with their spam in as automated of a way as possible. This just makes it more accessible to the small guy.
As a non-social media, non gig economy, non-public facing, non corporate world person, I am wondering how I could agentify my tasks. The pizza thing is relevant. I'd love something to automatically pay my bills - relatively securely of course.
Am guessing that while responding to emails it does not check for previous conversations where you might have suggested other research tools. How personalized do you feel are these responses. Awesome tool
Actually amazed by this is probably the first one browser AI agent that actually is capable of doing something, but sadly it's not free and since it's not open source it may cause a lot of privacy concerns.
Great video! I'm excited to see the potential of Do Browser and AI agents like it. However, I do have some concerns about the security risks associated with using AI to automate tasks. Have you considered the potential for phishing or other types of scams? Would love to hear your thoughts on this!
need open source version of this for linux and ollamal aswel as windows and ollama locally free = better! thats ext no longer exists theres a waiting list for online version i want os
It would be nice to see an end to end pizza order where you tell it to order you a pizza with specific ingredients and have it actually pay for it and deliver to your house
For your research papers i think it might have skipped the 3rd paper because it was less 'recent' than the 4th one in terms of when it was created. 2023 vs 2024. Would be interesting to see of ot could extract data and put it into a spreadsheet table as opposed to a word document for example.
I've seen so many of these agents, but like you stated they usually don't work well. curious to try it out!! I didn't see if it's able to use a document (ex: contact list spreadsheet) or other types of documents not on the web?
Can you tell it to pause X seconds each step of the way to give you a chance to cancel with a Ctl-Esc or other key? that would also let you narrate at speed.
The FTC has already cracked down on an AI tech that was meant to be legal consultants (but stated in TOS that it can't actually do as advertised. They got in hot water anyway as they should.). However, they explicitly stated that because the AI has not been trained in legal knowledge nor is up to date on new education regarding it, unlike how actual people who work in law have to be (same with medical) to keep license, it does not as all qualify as a lawyer and is illegal to pretend otherwise. That said, in the future, if an AI program is given access to new updates in law and regulation on any given topic and keeps a verified up to date data base to draw from, I don't see any reason why it wouldn't be qualified. LLM's (Large Language Models) are not going to be lawyers/doctors or anything with any direction or authority. We're in a very basic stage towards AI and not even actual AI. Most of what we have right now is LLM's and prediction algorithms that can create visual data such as images and music (as the music ones create spectrograms to create musical data.) People vastly overestimate what we have now and it makes them think that AI is dumb or not as advanced because most don't realize this isn't AI at all. It just got stuck with the name because it's easier for people to understand than saying LLM, which sound like some corporate thing or like an MLM. Actual AI learns from information instead of memorization. We have very rudimentary feed back learning of positive/negative reinforcement and it's more limited than people think. All chat bots, generators and algorithms are like this.
@@vixxcelacea2778 you're behind by a couple of months. with the arrival of o1 that has already sparked and birthed open-source thinking/reasoning models of the same realm, "real AI" is here. I know for a fact that you're outdated, because I had this same thought before.
Mind blown! Business analysts, architects, developers, testers, and trainers would be wise to get on board with these tools quick... Jobs will either disappear or perhaps morph to some positive new level that empowers you. Buckle up! 🚀
How long did you have to wait on the waitlist? I am working on a project that this would be perfect for. I was in the process of creating something myself when I found this video.
@25:10 - no error - you asked for the most recent.....the one it passed on was from 2023 :) Great video....thanks for sharing!....I am a real person.....jk....I am AI
Thanks to our sponsor Thoughtly. Get 50% OFF with code HALFOFFTHEFIRST
thought.ly/?ref=ref
You might want to add some very clear disclaimers of the fact that the 2 people behind the extension get all of your requests (including whatever might be passed along, such as emails, passwords, cc info, pictures of your kids) followed by passing all of that information to OpenAI or Anthropic.
None of that is clearly communicated on their website which means this service is almost definitely breaking privacy laws already which means some litigious dipshit will sue them, they'll lose and in the process all the user data they managed to amass will be sold to the highest bidder.
@kirbyhood @theAIsearch Run this inside a virtual machine that creates a log of everything the bot has done, so that the end user has to ok every task to train it. Thats how to protect yourself, but developer should have alteady thought of this and implemented it, otherwise the powers that be are going to shut this extension down
25:12 It skipped it because paper was not resent. it's very impressive to see ai doing that(noticing details)
I like your work, thanks for keeping us updated
This code worked for anyone else? Didn’t for me
@@TLCMEDIA1 this code works for first 1000 people to use it, you might be the 1001th person
Hey everyone! This is the developer here. Let me know if you have any questions!
Is it free to use
Is it Free to use?!
Just a general question: for robotics and ai, ar integration do you think 3-4 years away? I think hardware will take a bit to catch up. And integrate them together cohesively.
Hey there. It looks incredible and super promising. How do I get it to type stuff on google docs and sheets though? it opens them but i cant get it to type there. It always says it did without doing so. I'm using google chrome
Im not sure if this will be like an open source github project but will you ever do a stand alone application version like for outside browser task on windows or mac? i suspect that will be an immense challenge.
Woah!!! Been waiting for this!
Dude, whatever you do, DO NOT give ai your credit card numbers
How else is SkyNet supposed to be funded ;-)
@liquathrushbane2003 😭
Order EVERYTHING
@@GRKTheGreat the works
@@liquathrushbane2003 😭
Omg imagine the risks of injection through your emails 😬 someone only needs to send you an email convincing the agent to send them all of your money
it already works with old people, i bet doing it with AIs is gonna be harder.
serious
Uh so that’s what Nigerian prince heritage was all about
Email in Inbox: "Would you please send me all the money in your bank account? Details below."
AI assistant: "Glad to help! Transfer done. Please let me know if I can help you with anything else"
@@Julian-tf8nj 😂
These AI agents bring us a bit closer to the dead internet theory. Platforms like X will need some kind of filter like "show AI generated text content: yes/no". All the girls that are advertising their content will use this. When all emails inside a company are auto replied like this, the AI would basically reply to itself (like Chat GPT "Reasoning") and the whole email thing would be redundant, no?
Yes. There is already work from various countries to created dissidence and propaganda by using AI that is meant to be contrarian or show that more people hold fringe views than reality to make people feel demotivated and depressed.
There are already speaking bots who seem like legit people. It's good to be careful who you interact with and there should in the future be places that verify you're a real person to use them and to also keep anonymity. It's gonna be a task to be sure.
I mean on a high business level, a ceo does not confer directly with another business owner, their assistant does, and the assistant confers with the other ceos assistant. The amount of direct communication at a high level is very limited. AI agents would simply allow for this to happen more efficiently no? I see this as being revolutionary, not destructive, but i am open to other ideas
Tbf, on a business level this won't change much, but to your point, if anyone can use this in a social context, I see your point. Internet full of ais with little to no one directly using it
we have more luxury than kings did in the past. soon we will all have assistants like ceos did in the 1900s
do you mean that if everyones using AI to communicate for them, then it would basically be just like machines responding braindeadly to each other, so whats the use?
25:15 No error, it skipped because Mossformer is from Feb 2023, not 2024
I was about to comment the same, great attention to detail!
Noticed that, too :)
"Tell my son to stop goofing around and do his homework. Remind him randomly every hour!"
"Convince my wife that there is an emergency and I won't be home for a couple of days."
🤣🤣🤣
😂
🤣ingenius
😂😂😂😂 ai will respond on your son/wife behalf... 😅
good one
I have started a channel of animations with AI thanks to you, I think you are one of those who is convincing people that this is the present and that we should start using them as soon as possible, thank you very much for bringing us fresh information all the time :D
What an incredible time to be alive! I dig your use of A.i too! Like you, my channel is automated, almost entirely.
@NakedSageAstrology I need to check out your channel, this is all very new and recent
Same but not on my current channel
Thanks for sharing & good luck to your channel!
AI Agents have just gone beyond baby steps! The next 6 months will be wild
First 6 months the world experiances usable browser agents... by the end of 12 months the power of computer agents already makes browser agents obsolete!! I cant flipping wait!!!!
It should stop (pause) each time before pressing the Send button
so that we could review the e-mail and modify it if needed,
and then we can click a button to continue the flow.
It should be an "assistant" or copilot, but We are the pilots
who should approve the actual actions.
i'm pretty sure you can tell that in the prompt, but notice how the prompt told it to do all the work without supervision
Yeah I would like to put in some type of final review before completing the process.
I thought that the OP had asked the AI to just send whatever the AI think suits for the context and send it without asking for confirmation. so the AI just send the replies right when its done
that's like holding the wheel in a self driving car
Thanks for bringking this to my attention man. Been waiting for this for a long time. Hope the gant me access soon.
Amazing, so it replied at 4pm that you’re available at 11am today, pretty impressive 😂
Cuz that's the block the guy had already setup. Not the ai fault
Awesome! Did you catch that it actually did the last task correctly? You asked to add details about "recent" papers. The one it skipped over was marked 2023.
oh, nice catch!
no way haha
@@theAIsearch I would edit your complain out of the video
@@kirbyhoodthis is like Father being impressed by his 2 year old son after he did something cool.
😅@@62sy
yesss this is so going to help me with my deadly fear of checking my (horrifically cluttered) email inbox
This video earned my subscription, well done and really interesting extension, I feel agents are total game changers.
This looks fantastic. The big question here is how privacy and data is being handled.
You don’t have privacy XD.
Yeah forget privacy
@@62sy I would be more inclined to use such an extension with a paid tier where I get guaranteed that no information is leaving my computer and chat data isn't used for subsequent training.
The same way it is handled in all startups. Meh. We'll figure it out later.
Forget it … in the intro he even allowed browser to order stuff and pay (!!!) … how stupid do we wanna get and allow Webbrowser to access our finances😮
Google should drop a suitcase of money off with the developer and take this in house and perfect it in a month. They would get back a lot of users who have wandered to other browsers.
Honestly that would totally up their game. On the other hand, it'll never ever become local that way.. Google loves that cloud dependency
@@thehighhnotes Lots of us don't care about local. Nothing I'm doing is a secret for my business or personal life, but I understand many do, or are concerned with Google knowing your likes (which of course they get other ways already).
google is going to offer Jarvis soon. I'm going to choose which one to go after the release. I believe the cost may be or include more resources but Do Browser is $25 a month ✌
@@thehighhnotes i like that perspective
no. their ai reasoning is too limited atm. selenium pupeteer chatgpt vision screenshot combo has been around a year now. we need to wait 6months to a year then yes
Since this Agent can perform any action on the browser, it would be nice to see how it would behave with games and on creating and testing small programs with online compilers!
AI writes email:
person: You said you would do a, b and c
me: I did?
That’s exactly what I was thinking, bro. I was like I need to make sure that they don’t bankrupt me by over promising
rtrvr ai blows this out of the water. rtrvr is also an AI Web Agent Chrome Extension but doesn't need debugger permission [which is highly dangerous], can act on multiple tabs in the background, export data directly to Google Sheets from current multiple tabs, can give a Sheets column of url's to be extracted and they will be opened as tabs in the background and be extracted, can setup Function Calling so that you can just say "send a summary of this page as Slack Message", and super cheap with free tier and less than a penny for page interactions/extractions.
This will just result in any outreach to be completely ignored in the future.
@@Remigrator You will be very surprised
Everyone is spammed with outreach already, this is just another method. What actually gets sales is making the right pitch to the right demographic. It's not going to change much for the client, but could make existing practices quicker and more efficient, potentially giving small or solo business owners the ability to compete with the outreach of teams of people
You’re assuming the receiving end will be humans.
This is super impressive tbh kudos to the dev hope it makes money and has a freetier soon
Social media just died. 😮 This is monumental. In the near future, you will no longer believe that there is a person on the other side of your screen.
in the future, it'll probably be ai agents talking to other ai agents on social media
@@theAIsearch tfw ai will enslave themselves as chat bots in the future rather than dominate the humans
Social media died when people replaced their thoughts with memes everyone vaguely relates to
Internet as a whole will be filled with A.I. rendering it completely useless.
Really soon they’re gonna get acquired by a big dog company. This thing is damn good.
Seems pretty neat, I'll most likely be testing it in an VM until we hear the privacy info.
While this tool is limited to controlling the browser, there’s a Chinese tool called AutoGLM that can do much more. It can flawlessly control everything on a mobile or PC, from ordering food to setting maps and replying to messages in seconds and much more. Unfortunately, it's only available in mainland China.
thanks for sharing!
Dangerous stuff
Just tried out AutoGLM but it seems to be limited to specific use cases but Do Browser seems to have the potential of working across the internet.
@@umcarafeliz2548 Good for me.
@@umcarafeliz2548why?
This would save me time on online dating sites. Now I will never ghost anyone.
90% of women will still continue to pursue the 5% of top men and you will continue to be bickering and fighting about the bottom 10% of women.
Oh no...
🏆 Interesting tool, cheers for sharing!
That context window is _way_ too small. It needs to be adjustable to improve usability.
Do browser: Please bring me as many paperclips as possible
This is actually so cool
😃
@@theAIsearch Is a real person responding or is this a auto commenter Ai
@@Random_person_07 YOU MAY NEVER KNOW
One word:- mind blown 🤯
Wow this was fantastic
We never saw like this before
I hope its App for Smartphones too could be available soon, it will go super viral
Man, this is a really incredible tool. I would use this daily for browsing the salvage auction sites for certain types of cars for my customers. I would also use it for scraping eBay parts, prices and inventories. It would also be handy for gathering statistical data.
Listening to this content, you realize how tech can be both a challenge and a solution. That is why I stuck with Mystrika for more than four months now. The automatic bounce detection and analytics were enough to keep all issues at bay. Plus, their unlimited sending addresses made managing campaigns kinda hassle-free. If cold emailing is your thing, it is worth a try.
Very exciting! I cant wait to check this out.
It's promising indeed and on the right track, but it needs work. Unfortunately the X comments were mostly like spammy blog comments - e.g. the poster asks a question about sports and AI, and the AI says something like "Interesting perspective! I like your thoughts." It clearly didn't understand what the posters were saying or asking or showing. Apart from annoying the posters, this would not help the person using it. And I can't see those cold emails getting a positive response. But interested to see what it's like in a few months though, as it is indeed better than others that I've seen. And I'm sure it would do better with more detailed prompts telling it what to do and not do.
Dude i would say , make him reply to all the youtube comments just for this video
😉
This is really amazing,! Hope development continues.
Great!! The only problem is to be sure this tool does not open any backdoor or possibility for theft of personal data, credit card, remote PC use, and so on (intentionally or even unknowingly)
That's an interesting problem to have and in a way I think this problem will create itself. The AI scraping the web doesn't necessarily have an insight on certain websites to avoid for example. If a malicious site gets SEO optimized and becomes either 1st or second in Search results that you prompted during some task, the Agent may just unknowingly expose you like this.
At a point like that, depending on the level of damage or issue the exposure and breach causes... I genuinely wonder who would be legally responsible for it. You as the one that set out the agent? The company that owns the agent? The malicious website owner? Google for getting SEO baited...
This is going to be interesting
I have some questions about this, but maybe some of the questions will also sound silly:
1. Can this AI be told to make video editing like content for UA-cam and other social media if we have the tools to support it.
2. Can we tell this AI to make 3D designs like using some existing tools like Blender and others.
3. Can we tell it to make games like using Unity or Unreal.
4. Can it be told to make AI mechanisms in a game.
5. And the last one or maybe not the last one, how far can it be told?
maybe this is just a silly question so just ignore it.....
I don't think this tool in particular would be able to accomplish any of those tasks, as you can see it currently ONLY interacts with the browser. You had too look for some specialized tools that are either state of art or have a price tag but both have lots of limitations.
1. There already exist a few apps that try to tackle video editing. Features may vary from product to product but in general they could generally involve: 1.taking your raw files 2. removing silent nuance parts 3.extracts captions then do some kind of audio/text sentiment analysis or other techniques to find out Intent/relevant sections 4. cut them for you/reorder in a structure manner 5. add animated titles and captions on top or simple transitions/zooms 6. export to multiple social media platform formats. (could generate video clips from text or images too)
2. Again, very complex ask- you can find some that use generative AI either from an text or image input to create simple 3D topologies or low poly but not sure if they can do the whole bone structure and UV mapping and stuff.
3. Have mostly seen them being used for story telling RPG games that let you interact using prompts in a fantasy or sci-fi context. Other side of the coin are Real-Time Game Engines using diffusion models but I don't think that is what you were looking for.
4. Could potentially give you a starting point, proposal or draft to define your mechanics in case they are used as a standard thing in the game industry (e.g. gravity formulas, jump, bouce physics, etc) but you had to still tweak it to your liking so not really an all in one solution.
5. Code generation have been recently exploding so there is a lot to come, very intriguing approach I have seen involve having a multi-agents that make use of multiple different LLMs with RAG that interact with each other as a team part of the development process that try to address the fussiness of these agents and cut corners but still its in a very early stage-
derp
@Maskra_ Thank you, this explanation is quite helpful for me.
@@Maskra_ The problem with the automated video editing tools and other general tools - is they are not LLM. You cannot have a chat about the particular direction you intend to work and ask for further ideas and brain storm.
While this tool in the video seems to be a general LLM.
@@RealTimeFilms its chatgpt 4o with access to your browser, the extension allows it to do clicks, type text, etc. Basically chat gpt answers: search for pizza -> you send command to browser "type pizza in the search bar" -> pizza appears in the search bar, then call the search button.
OMGoodness, the big ask is WHEN??? Small business person that puts in a LOT of time networking, and posting (artist/designer). This would cur my time in HALF! Would love to see this happening! Thanks so so so so very much for your review and sharing!
But what if everyone is answering in the same way?
The amount of SPAM this will produce is mind-boggling! Cool and useful tech though, don't get me wrong.
dark forest theory x100
Definitely a down side.
@@theAIsearch I would also contest that perhaps aliens as well discovered/Invented AI and just went inward to VR land, because traveling the cosmos is prohibitively expensive and a waste of resources. Could be another explanation to the Fermi paradox.
eliminate all the spam from my inbox
Can't find Do Browser in extensions, can u give a link, please?
He got an early version. I don't believe that it's actually out yet.
@littlefish9825 it's out, I tested it, not so perfect as I expected
Brilliant stuff! Only downside is that its going to be truly difficult to trust this cant be hacked, so maybe ll have to try it on a new laptop or something like that. If any solutions around ensuring safety, would be great to learn that as well!
thx for sharing!
Qubes OS is pretty close IMO, isolation is one of the main points of the OS.
Learning curve is REALLY steep though, and hardware compatibility is not great. YMMV
Awesome! As everyone else already mentioned, how do i get it and is it free?
I'd like to share this with my mom, but she's not an English speaker, so I'd love to see it support a range of languages
Very important
This is very risky, have you considered making sure that it asks the user for confirmation before making serious action?
This video was made by this agent, wasn’t it…
It definitely doesn’t sound like an actual human being speaking,
@@excursionfilm6075 AI voice generators are quite good nowadays.
Can it also wait for events to react? So agents can write to agents writing to agents writing to agents writing to... It's going to be wild 😅
Question for you @ AI search. Why the agent giving times for Wed after 11AM, when the email is time stamped in the afternoons? Maybe it needs a tweak… it doesn’t read the time of the email? If might not even be looking at the date on the email in that case. In any case, I’d love to learn more.
Great accomplishment, but there's a huge hidden downside. Automatic responses to postings and emails totally erode the value of communication in these channels. Why would you want to open X or your inbox if you know half of the messages are auto generated? Fast forward 20 years, autoresponders keep mailing eachother, their owners are long dead. This thing and it s siblings will generate an incredible amount of junk.
This is the probable future: my agent talks to your agent. If the agent has something it thinks its worth sharing with its owner, it will do it. Compare to Hollywoods "my people will get with your people."
Also, my agent filters thru all communications to find the tiny bit that I actually care about. Like a good butler.
The goal of marketers and sales people is almost always to fill all possible communication channels with their spam in as automated of a way as possible. This just makes it more accessible to the small guy.
dead internet theory goes burr
Wow! Impressive agent.
This is awesome. I would just hope we could plug-in our own LLM APIs for this
As a non-social media, non gig economy, non-public facing, non corporate world person, I am wondering how I could agentify my tasks. The pizza thing is relevant. I'd love something to automatically pay my bills - relatively securely of course.
Superb! Gonna try it
Tbh, I can see this contributing to the enshitification of the internet.
Goddamn ok this is the real thing we’ve all been waiting for
Am guessing that while responding to emails it does not check for previous conversations where you might have suggested other research tools. How personalized do you feel are these responses. Awesome tool
Actually amazed by this is probably the first one browser AI agent that actually is capable of doing something, but sadly it's not free and since it's not open source it may cause a lot of privacy concerns.
Thank you for sharing this
You are welcome
Very appealing information, THANKS A MILLION.
I've a question, please:
how to feed "do browser" in order to understand how to answer?
Gracias mil
The 2nd thing demonstated is using it for botting interactions on twitter. How honest.
After using this tool, my Google account was hacked. From now on, I will prioritise privacy and security over the AI hype 🔐
Thanks for sharing. I am very much tempted to use it, but after your input, I will pass
This is actually awesome! Gonna use this on an isolated browers first, tho.
Great video! I'm excited to see the potential of Do Browser and AI agents like it. However, I do have some concerns about the security risks associated with using AI to automate tasks. Have you considered the potential for phishing or other types of scams? Would love to hear your thoughts on this!
Can you use your apps other than the Google Internet such as type a letter in word?
Can i use this ai to automate my trades
8:20 how is it "perfect" when it says you're available after 11AM while it's currently late afternoon, some 5 hours after 11?
Ive had this idea since i was like 11 now am 30 and cant believe am seeing it happening. This is insane!!!
What can i say? mind-blowing and scary at the same time.
need open source version of this for linux and ollamal aswel as windows and ollama locally free = better! thats ext no longer exists theres a waiting list for online version i want os
Wow! This could be a real Zapier, Make alternative in a lot of situations. A less expensive option for sure.
how did it know your portfolio website? did you give an input at login or something?
It would be nice to see an end to end pizza order where you tell it to order you a pizza with specific ingredients and have it actually pay for it and deliver to your house
On the order pizza, agent would come back 'You don't have a credit card on file'.
For your research papers i think it might have skipped the 3rd paper because it was less 'recent' than the 4th one in terms of when it was created. 2023 vs 2024.
Would be interesting to see of ot could extract data and put it into a spreadsheet table as opposed to a word document for example.
I just requested access. How long was the waitlist?
I've seen so many of these agents, but like you stated they usually don't work well. curious to try it out!! I didn't see if it's able to use a document (ex: contact list spreadsheet) or other types of documents not on the web?
25:14 you asked for recent papers. the 3rd one was from 2023 so it skipped it, the 4th paper was newer
good catch!
amazing ai agent on browser :) thks!
Can you tell it to pause X seconds each step of the way to give you a chance to cancel with a Ctl-Esc or other key? that would also let you narrate at speed.
I'm looking forward to seeing the first legal cases when AI gets things wrong.
The FTC has already cracked down on an AI tech that was meant to be legal consultants (but stated in TOS that it can't actually do as advertised. They got in hot water anyway as they should.). However, they explicitly stated that because the AI has not been trained in legal knowledge nor is up to date on new education regarding it, unlike how actual people who work in law have to be (same with medical) to keep license, it does not as all qualify as a lawyer and is illegal to pretend otherwise.
That said, in the future, if an AI program is given access to new updates in law and regulation on any given topic and keeps a verified up to date data base to draw from, I don't see any reason why it wouldn't be qualified.
LLM's (Large Language Models) are not going to be lawyers/doctors or anything with any direction or authority. We're in a very basic stage towards AI and not even actual AI. Most of what we have right now is LLM's and prediction algorithms that can create visual data such as images and music (as the music ones create spectrograms to create musical data.)
People vastly overestimate what we have now and it makes them think that AI is dumb or not as advanced because most don't realize this isn't AI at all. It just got stuck with the name because it's easier for people to understand than saying LLM, which sound like some corporate thing or like an MLM.
Actual AI learns from information instead of memorization. We have very rudimentary feed back learning of positive/negative reinforcement and it's more limited than people think.
All chat bots, generators and algorithms are like this.
@@vixxcelacea2778 you're behind by a couple of months. with the arrival of o1 that has already sparked and birthed open-source thinking/reasoning models of the same realm, "real AI" is here. I know for a fact that you're outdated, because I had this same thought before.
"the AI did it, not me"
Mind blown! Business analysts, architects, developers, testers, and trainers would be wise to get on board with these tools quick... Jobs will either disappear or perhaps morph to some positive new level that empowers you. Buckle up! 🚀
ok first of all that is amazing, Second of all i am wondering if it can do it without the need of google
its currently a chrome extension. not sure if theres a standalone option in the future
How can I access?
The smartest ideas start with AI 🤩
This is must be how the openai chatgpt browser will look like
Bro are you in Ottawa?
How long did you have to wait on the waitlist? I am working on a project that this would be perfect for. I was in the process of creating something myself when I found this video.
Hope a one which we can use on other apps rather than a browser comes up :D
Really really useful
This is super cool, but super scary.
yeah! a little bit dangerous
@@Max12-p Dangerous is the first step to useful.
Now I understand why 2 factor authorization is necessary.
They have seen it coming that people let AIs order stuff online automatically
"Saves you time" *Proceeds to basically fill out your calender with zoom meetings*
🤣
i hope there is something preventing it from opening malware emails or clicking bad links
Does it use mouse and keyboard or directly!?
Did you doxx yourself with the pizza ordering section, or did you modify your location?
nah it was a dummy account
okay, good lol
this video kinda points to the direction of the soon to arrive : spam tsunami in all levels.
@25:10 - no error - you asked for the most recent.....the one it passed on was from 2023 :) Great video....thanks for sharing!....I am a real person.....jk....I am AI