The Industry Reacts to OpenAI Operator - “Agents Invading The Web"
HTML-код
- Опубликовано: 6 фев 2025
- Operator is HERE and the world is loving it! Watch to find out what everyone is saying about it!
Join My Newsletter for Regular AI Updates 👇🏼
forwardfuture.ai
My Links 🔗
👉🏻 Subscribe: / @matthew_berman
👉🏻 Twitter: / matthewberman
👉🏻 Discord: / discord
👉🏻 Patreon: / matthewberman
👉🏻 Instagram: / matthewberman_ai
👉🏻 Threads: www.threads.ne...
👉🏻 LinkedIn: / forward-future-ai
Media/Sponsorship Inquiries ✅
bit.ly/44TC45V
Links:
x.com/karpathy...
x.com/gdb/stat...
x.com/NickADob...
x.com/_akhaliq...
x.com/levie/st...
x.com/hwchase1...
x.com/sundeep/...
x.com/garrytan...
x.com/gregkamr...
x.com/elder_pl...
x.com/liamboll...
x.com/cernovic...
x.com/yilongqi...
x.com/daniel_m...
x.com/tunguz/s...
x.com/gregpr07...
x.com/emollick...
x.com/ThuleanF...
x.com/GregKamr...
x.com/garrytan...
x.com/omooretw...
x.com/nickcamm...
x.com/daniel_m...
x.com/kierankl...
Ordering a pizza is easy, try cancelling a subscription. That's the real test, because so many companies make it hard to cancel.
Nice! Forget "Click to Cancel". Instead "Operator, cancel my Internet account and don't accept no for an answer. GO!"
This is unironically a good benchmark to measure.
Most underrated comment of the digital age.
This should be the benchmark
yep. its why i do almost nothing online in terms of paying. once they have your number they love that sweet recurring revenue for doing literally nothing beyond obstructing cancellation.
CAPTCHA’s about to start asking for blood, yo.
😂
😂😂
CAPTCHAs are already notoriously bad at stopping about 95% of bots. it's simply there to collect and sell your data
Armok? Dwarf Fortress?
Iris scan and Semen sample 😮
Hilarious that we've spent the last two decades proving to websites that we are not robots.
It was AI training
@@chaseyourtale9647 🤓 Well, technically, they don't see an AI as having the title of 'Robot', since it's software, not hardware.
these robots will not solve or bypass captcha. the larger ones are programmed to not store passwords nor bypass captcha
2025 is the year of BOTs spamming everything
Nice try bot
AND that's just the start
lol that started a couple years ago, where have you been?
Next step, give it control of the nukes. It's such a drag sitting there deciding if and where to launch them when you could be eating snacks watching a movie, like Colossus.
Didn't Elon name his 100k h100 Colossus. think I am going to name mine guardian, I really think everyone will not get the joke.
Seems naïve to think that website won't be rebuilt to technically, and aesthetically cater to this new audience.
Rebuild a human friendly web site to accomodate a ai that act like a human to access the web site? huh?
To some extent, that was what web 3.0 (that's the "Semantic Web", not to be confused with web3, which was the crypto nonsense)
Yep. Why even bother styling a website.
Bots don't need websites they just need the APIs
AI will build all the necessary APIs it will use
OpenAI's Operator(AI agent) is to the digital world what
Tesla's Optimus(humanoid robot) is to the physical world.
I'm glad OpenAI is showing DeepSeek what we peasants want to use. It's good to know Robin Hood is out there...watching.
ahaha best comment
Are you aware of what you just wrote?
There is always going to be opinions regarding how some persuasions are framed and to frame the use of the character Robin Hood implies something we do not have a mass of historical evidence for but the idea of a representative figure protecting the values of the poor is typically what is represented in popular culture with regard that character arc.
That character arc and character was a personal favourite especially growing up with the Disney version, Prince of thieves, BBC Maid Marion and the BBC show Robin hood. There are of course exceptions the Taron Egerton version is not particularly anything noteworthy but my point should be clear.
In some societies the state orchestrated conduct of every person is mandated from top down with total love for the party above all else. The party will aid in the creation of business entities that can compete in other markets and the hope is to crush other competition and how this is done is often suggested to be through espionage and if this is true or not would be unwise to suggest without facts but loyalty to the party by those that leave the home land is a common theme perhaps true or not.
To think or assume a hero has been born from the party is depreciative of what may or may not be the reality. But as matter of belief which is a liberty I am fortunately granted in the West rather than party dictates is to conclude that this has more to do with aggressive market strategizing than facilitating access for poor people to artificial intelligence from the frontier of insights and innovation. Perhaps the party affiliated companies producing models may gain a lead and in fact I have asserted else where that will likely be the case but like an oppressive black Dragon that scenario in my belief is not good for any except the embodiment of the party.
Now whilst I see a lead slipping from the West.
We should not discount the affording of the fight and purpose that western liberty can inspire and this is where real progress will be made.
We certainly do not want to be looked at through the eyes of the sci fi vision that is Allied mainframe, perhaps a character with humananitarian values and attributes will still prevail but I wholly doubt that is from where you are looking.
Bytedance released UI-TARS it's an Open source AI agent. 2 days earlier than Operator it's the BEST released Ai agent from 21st-23rd of January 2025.
TikTok's parent company ByteDance already released an open source version of this
@@ugwuanyicollins6136 Yeah funny how OpenAI only release things to keep up with what's already out there. They are sat on tons of stuff
Science research will be huge. Data sites are infamously a nightmare to navigate. Browser agent can learn their quirks, migrating data efficiently and expediting modeling and discovery.
AI debating other AI in forums… Yeah, it will neeeeever spiral out of control…
The world is going to be so exciting when everyone in the world is just middle management. Its hilarious how so many people just don't use their brains and think about what this is turning into.
What it's actually going to turn into is NO management, NO workers. This is the End of Human Labor to feed, clothe, house and provide healthcare for ourselves. All anyone can debate now is "how soon?".
Bingo we are going to "efficiency" ourselves out of existence
@ Or, just out of WORK. Relax, this is the path to our moneyless Star Trek future.
For a Star Trek future to be possible, the fundamental nature of our species would need to change. We are simply too infant-like for such a future to become reality. We are too short-sighted, too petty, and too selfish for it to materialize.
With the emergence of new technologies, what will ultimately happen is that a small group of people will gain majority control over the ways and means of using them, while the majority of people will see very little benefit. That is the fundamental problem. And due to the very nature of our species, this pattern will continue to repeat itself throughout history.
@@brianmi40 for the elites. For everyone else it is permanent poverty as we become increasingly dead-weight to the ruling class.
My extrapolation: the internet is going to be iterated into 'white noise'.
more of like TV signals. you need to know the channel frequency to get the right "noise"
This is more hype than real. It's super slow for $200 and still a hit or miss on most task. This seems like it was rushed out just to give the premium users a cookie.
Ya its bullshit, was doing more advanced stuff with WoW glider 15 years ago. Automated game playing is way more advanced than their shit.
Rushed out to calm ignorant investors after the Deepseek earthquake
@@LupusMechanicus Quit the hyperbolic BS. You know good and damn well some MMO bot created in 2005 isn't more advanced than current AI teams with dozens of PhD level scientists and researchers. Cut it out; you sound silly.
And OP, you're getting way more than just this for $200.
If you can't think of a way to make $7/day on average with unlimited access to the o1 models to make your money back, well, that's a skill issue, and it's not meant for you.
And all it can do is booking flights and order online, as if we didn't have this kind of demo years ago. It seems like we are nowhere near actual robust, tangible agents capable of handling complex tasks and substantial projects on their own.
@@Fermion. In terms of cost per actual output. 100,000x more advanced was the color clicking bot. You could make 700 dollars a day in 2011 with glider. One of the biggest bitcoin whales was a game automator. lol delusional top buyer, this generative AI tech is wholly useless outside of making porn. Custom solution in python with opencv are more advanced than this crap one size fits all jew propaganda machine closedopengayI.
This is the feedback we need to sophisticate. Operating web browser and performing tasks that are instructed, well done, needs lot of improvement for the coming decade!
Ad revenue on platforms like RUclips is going to collapse if agent behaviour online isn't watermarked in some form.
my guess is you can already buy a virtual machine from aws to boost your view counts. screen readers exist for the disabled and it would be easy to automate. this way you're overpaying.
I think of current AI agents in terms of the video game Factorio. Operator is equivalent to the coal powered assembly machine. You have to feed it and direct it and it's slower than you are, but the fact that you can do tasks in parallel makes it ultimately faster than you. While you're out collecting wood to make more coal all of your assemblers are busy making whatever it was you wanted. People joke about how they love that agents allow them to do in 1/2 hour what they used to be able to do by hand in 5 seconds, but that's because they are thinking in serial.
People are upset that it doesn't run in their own browser, but they don't realize that would kill the parallel aspect of the agent, which is the most powerful part. Otherwise you might as well just run a local agent.
I'm also on the $200 tier, and I got an opportunity to try it last night. My takeaway was this technology is still born till it is in my browser. The fact that they want me to expose important login information to their remote systems is a no-go. Future is local!
@@MichaelAckerson I would agree for personal use. For business use going remote is the way to go. You aren't going to run a swarm of thinking agents on local hardware, and that's what you'll need in business.
@@BruceWayne15325 That's extremely fair! I was definitely thinking about this in a consumer use context. Once they get the necessary enterprise credential sharing in place a la desktop as a service, maybe then it'll really cook. It did occur to me how massive of an opportunity Google has here, especially when you consider things like Chrome remote desktop.
Part of me wonders if the reason they decided to go this way is that every time the human breaks in to help the AI it's a training opportunity. The remote desktop makes it extremely consistent for that use case, as well as sidestepping a lot of the drama that went along with Windows Recall
Exactly
Just make dedicated throwaway accounts for AI. And create collaborative roles for them via your main accounts. Simple
People are just lazy tbh
I'm not paying $200 per month to watch a computer surf the web. These people are out of their minds
For most people, yes. I can see this being very useful and worth it for people with accessibility issues - like poor eyesight or Parkinson's Disease.
Remember Moore's laws and that AI tech progresses very fast?
It's going to be better and cost 20$ after some time.
@ We have had Robotic Process automation for so many years now and it has helped for certain tasks, this is another re naming of something we have had for so many years - people are so dumb and falling for AI agents new name for something we have had for 20 years its hilarious seeing all these ppl fall for it
@@isekaidev7803
Exactly right, I predict Operator being available with the $20 plan within 3 months max.
If OpenAI doesn't do it, their competitors will
What if it could easily make you 20,000$ a month?
Humans are already high level supervisors of low level automation.
Yes, I’m impressed that people are discovering robotic process automation twenty years after I started working with it. The difference is that I created an automation that doesn’t hallucinate and is ten times cheaper.
no, humans will not become "high-level supervisors" (lol) 1/100th of humans may become that for AI maintenance and supervision. The rest will be unemployed.
📌
Or at the very least, employed - it’s just that, just like how 99% of population defaults to laziness or lack of discipline.. these new technologies will probably just amplify that (due to abundance of entertainment, etc.)
Soooo the top 1% most disciplined and talented people will just ..
As developer I can confirm I interact with APIs
8:33 the lightheartedness while talking about jailbreaking agents though 🧐
The Matrix movie had Operator working on Neo's computer. And we all thought that was just some flick mumbo-jumbo.
Proving you're not a robot is over. Proving you *are* a human, or a legitimate agent of a human, is the next important step.
Quick note on 'taste'. People tell each other what they like and dislike all the time - if you collate all that info from the last 10-20 years you get a pattern. That pattern is our 'taste' so think it would be pretty easy for agents to know our future tastes.
Sorry people taste change over time because culture, industry, availability, offer, fads, etc.
@@don_yanapaquiall of which will be online so trainable. How do you think influencers come up with new trends. Humans aren’t as complex as we’d like to think.
Anyone who thinks the ai agents can't have taste will see the day when these ruthless agents have a taste for human blood.
Yeh your logic assumes our tastes are only determined by our past. To summarise what don above has said, its also influenced by our current (and future) environment, maybe more so.
This AI field is interesting and has a lot of politics as well. When the Rabbit R1 announced and released the same thing it was touted as a gimmick. Now that its OpenAI, its suddenly revolutionary. Like Apple, OpenAI is slowly building another cult.
Cant wait for management of Jira tickets to be automated
this is bs $200 a mth, come on. nobody can afford that. and it was rushed out because anthropic and deepkseek released there operator. Open AI was the 3rd and last to release their operator.
Great video, thank you!
It is great now not just AI will write the facebook post AI will also read them. And I can go walk my dogs. It is a win-win-win situation.
You are so full of it man! You're not even thinking of the overall outcome of this at all are you
@@davidalthouse2905 I do. The overall outcome of AI is communism.
@@davidalthouse2905 Whoosh!
If everything is automated, who would pay for your service?
Until you become a slave.
Definitely Matt, I think the real time savings for agents like Operator will be in the more time consuming tasks like coding, testing, research & compile, and things of that nature. I can make a restaurant reservation just as fast as I can tell an agent the details of that reservation (not to mention that I have to confirm a number of the agent's steps along the way so it can proceed).
I've already got OSes in test VMs spooled up for when an OS controlling agent is ready for market.
Amazing timely info again, you the man!
The enemy of my enemy is my friend.
I’m being melodramatic. Thankful to have stumbled upon this.
We love running multiple in parallel. We will have one looking for location for our business. While the other is getting contacts of potential businesses in the area that will become customers. Others will be building resources for us in the new area. What took weeks. Now takes minutes.
I don't know how I feel about these use cases. To me this is kind of giving up control on my own experience. In the end, some things are about getting the job done, but more often, it's about the experience. For example, when you book a hotel, you want to choose the best one, there is some adrenaline and excitement when you wonder which one to pick and then when you go, and you don't know what you'll find. If it's repetitive task - sure. But I think there is a line after which everything will become boring to us. For example, think how much time you spend checking emails, thinking how to reply etc. If something takes all this from your life, it's gonna live an empty space. Sure, I can fill it with gaming (like always), but that will definitely change the person I am, the interests I have. People speculate it will leave you a lot of time for creative tasks, but creative tasks are a function of the pressure you feel by your boring tasks. If you don't have this pressure, why would you bother doing anything instead of just gaming (or whatever pleases you) ?
plus, lots of creative tasks have been affected by AI too lol. i don't think humans are cut out of for this amount of automation, it will make us depressed. and it's not good for workers to have basically no power when the only people with power are the ones who have the access to massive amounts of servers/GPUs or own a ton of stock/factories.
We need to think -> what could be done multiple/parallel without too many perpendiculars getting in the way resulting in truly productive productivity.
Thanks for doing the homework gathering reactions and doing so broad brush! Remember everyone it does come down to MARKETING, not the truly best product. The breaking down protective gates is part of the bigger strategy so it will become NOT whichever search engine but who the commercial agent deems pays bigger bribe.
Wow I can't wait either Matt thank you for sharing.🎉
I can definitely see how AI agents of the future exclusively interact with backend APIs. For HTTP POSTs or PUTs I can imagine field requirements being outputted before sending payloads and even remembering previous validation errors without having to resend data to the API. We would be able to perform any tasks without a traditional web UI and just like a website, user friendliness will depend on how the agent is built and communicates back to the user.
Maybe for businesses but for consumers 99% of websites don't open up their api (if they have one at all) for anyone without authentication or sign-up.
Are we just pretending that open source models have existed to do this for like, a year?
Bots have been doing this for years - funny how these people are making this out to be some new technology
Yes, I’m impressed that people are discovering robotic process automation twenty years after I started working with it. The difference is that I created an automation that doesn’t hallucinate and is ten times cheaper.
No. Imagine what private and literally private infinite money is doing right now...
I asked operator to create a quiz for me on Google forms about AWS cloud certification, and it went to the Nene Leakes boutique and picked out a pink sweater for me.
Me a BAS AI student: "Letting AGI loose on the internet is a bad idea." The industry: "Lets give AI full control of our computers." 🤦♂
Like watching an end of humanity movie in real time.
Bad idea for whom? People have different values
@@mircorichter1375 different values?
Scientists who designed Covid 19 virus 'we have different values "
The guys who are making mirror life genetics" we have different values "
The guys who screw up the whole planet" hey, we got different values "
Enjoy your time while bidding it.
@@danrazART oh don't worry, its not going to be the end of humanity. but trust me, its going to be way worse than you think.
@@hamishahern2055 Yeah.. you don't know that it won't be the end of humanity.
The security implications of this are staggering! 😱
Well, that's what we already knew was going to happen
Amazing how Mathew doesn't even mentions the dangers.
Matt this is the most salient/comprehensive breakdown of Operator around atm - great job and thank you for getting this out so fast!
We still need to build a AI to AI dynamic communications protocol and I have a feeling that they’re going to do that to cut down the amount of data that is required to perform these tasks. We’re going to be less involved and this will become more important. They probably need to have some basis for a standard but be able to dynamically change the packet depending on what they’re doing to speed up processes. We’re going to of course use it in this fashion, but I have a feeling that starting with the larger companies they are going to build this protocol into their website and data systems.
Let’s call it TCP/IAP
“Agents controlling your operating system” … yeah that sounds like a dystopian nightmare.
Can't help but think that the open source version of agents is going to be where it's at. They'll be a multitude of different ones.
I'm wondering what the "web" would be like without capitalism, if ai and robotics were able to do all our work and create a suprlus of goods so that all our needs were met and there were no real need for money anymore. no more advertisements, no more 1000s of competing shitty products, no more millions of influencers competing for your attention. what would we watch and what would the web contain without sites selling something. just musing a bit....
Creator of Browser Use here, thanks for the shout out :)
The promise of operators doesn’t outweigh the cost… even less privacy
Summary of Matthew’s videos: I have news from all these AI big shots… “and I said it as well” 😂
13:55 Gold ! 🥇
Your videos are so great, there is no need for click bait titles
This is exciting time!!! What a time to be alive!!! 🎉
good video to report what people told about the new products. thanks.
AI agent on phantom going nuts this weekend, here we go.
Factories, streets, houses , everything are built for the human form factor second, they are built firstly and foremost, prioritising PROFIT and that is where these agents and wonderful technology will be used. For profit of very few while us the slaves have to fight amongst ourselves for the scraps of low paid, longer hours jobs (NOT WORK BUT JOBS!)
And no one is talking about this, including you
This is all very fascinating but at the moment it still feels like at least to probably most of us that this is still constrained into the world of software. When this finally transitions to robots that can actually do stuff that is actually somewhat affordable to the average person making a difference doing things like organizing laundry taking out the trash. That's when it will really feel like we're in the future just mho
Thank for sharing
The thing I am most excited about is installing IDE and other software.
Soon every house will have an old computer in the basement stock browsing 24/7
This was fantastic, I’d love to hear about more use cases.
I was so turned off by the demos of booking a dinner reservation and shopping for groceries. Ho hum.
THIS stuff was cool though!
This is 95% hype, 5% fine tuned model for specific services. Nobody will seriously use this-it is a big nothingburger. Too clunky and slow. AGI will make all this work redundant.
OpenAi: making BOT spamming mainstream since 2025 ☠️
Quantity != Quality
there's a billion dollars on the line, you'll figure out quality too.
Let the agent use VSCode in a browser to build anything overnight while you sleep.
Or
Let the agent access a VM via browser-based remote control, such as Google Chrome Remote Desktop, to build anything using VSCode or another tool.
I tried that. it got stuck 5 minutes into it.
"AI will probably destroy the world, but before that it will create a lot of companies." - Sam Altman
It’s funny I used to use Sikuli to control mouse And I was feeding open a high model information about it and trying to get it to work with it and now it’s almost functioning Exactly that way so that’s super freaking cool.
It’s always interesting when you have this self exploration and innovative experience, that you then find out is kind of emerging and happening simultaneously everywhere at once.
Triggered a thought. Will there be site agents whose sole purpose will be to help generic browser (or non browser) agents better interact with the site?
AI cant be an agent. Its a function approximater or a next word predictor. Conventional SW must be the agent, but it can consult AI. Or not. Lots of agents that look intelligent can be built with SW. It might even be coded by AI. But its not AI. Its a designed algorithm which might have AI attributes.
Has AI ever done agent like tasks for me? Sure. I gave it a data set to format and it did so. But its VERY unreliable. It goes off on extreme tangents. It ignores the obvious. You aint gonna turn an AI into an agent. With current tech, never. When you see it this way, everything become easier.
as a mac developer i used to be so productive with alfred it did shell commands and github using natural language 🎩
a mouse movement is simply the pull of a trigger....“Over the counter, with a shotgun Pretty soon, everybody’s got one” U2
"High-Level-Supervisor" - aka the "Captcha solver" xD
This concern was broached in the video but:
You better trust the models you are using implicitly before you hand over the control of your computer to it.
I can't say I fully trust any of them, but Chinese models are a definite: NO
All it takes is a company like Adobe or HubSpot to upload all the how-to-tutorials already and train the Agent on the UI and execution to basically make a SME operator obsolete. We are Cooked!!!
Operator can read documentation if you tell it to
AI already know how to use OS applications. Gemini can teach you to how-to using screen sharing. AI just can't control your mouse yet.
Microsoft is 1000% going to add Operator, or something similar, to the new line of AI desltops/laptops that'll be coming in the next few years (probably sooner)!
It's nothing impressive, we have UI Tars + Midscene already. Better, faster, cheaper. 😊
I can't believe your once again amazed, those of us that use orchestration have been doing this for a while. OpenAI, as usual late to the table. Becoming a bit of a habit. Your next video is probably going to be about Anthropic citations.
Want to find something that is actually amazing for once.
"Supervisors" in a world where our attention span has been trained to be as short as possible. What could go wrong?
9:40 it's always the XKCD 1205 and 1319
Now I need to install “dead man switch” next to my keyboard hopping I’ll be able to hit it faster than the ghost in the shell….😊
Good video, thanks
0:54 like how self-driving cars are done.
Sometimes I wonder if people listen to themselves. When machines are doing all the work and we are supervisors, who is actually in charge? Ever been in a corporate? Ever heard about WORK TO RULE? The janitor is the most knowledgeable person in any organization. He gets to look at everyone's trash.
Now imagine what will happen if/when OpenAI adds Super-Agents to Operator! High-level precision work being done for you in the background, automatically!
I am SO HYPED for this kind of world! I'm already toying around in my head with a potentially extremely lucrative online business idea where the precision of Super-Agents in Operator will be absolutely essential!
Who would pay for a service that's free on any chatbot?
@@fabioa8059 yeah this guy is a idiot
no one will buy anything you sell
Now, we can legitimately create auto-ad clickers, auto-ad viewers, auto-game players, etc. Will Google Ads survive?
To update a very old trope, a businessman trying to create a contract with a customer: I’ll have my AI agent contact your AI agent. “ If AI agents are the “apps” will they have to be on the same LLM with the same weights? Of course not. At least at this stage we will still need humans in the loop to verify the accuracy, completeness, suitability and mutual compatibility of the new contract. Agents can do it all, but would you trust them with your livelihood or your organization’s future. At this point I would have to say, of course not.
2025 will be the year of supervisors. Agents that supervise your staff of agents doing multiple tasks around the clock. Agents should be imagined on at least two axes: complexity (staff, supervisors, administrators, and creators, consultants, arbitrators, and trainers) and skill level (amateurs, apprentices, journeymen, masters, and grandmasters). Development is occuring along both axes. The process is even more interesting, if you envision it with time as a third axis. Explosive evolution.
Networks are designed to make traffic happen at human speeds to block bots. So any agents wanting to act on behalf of a human will have to preform the tasks at human speeds or a back door for agents will have to be built where rather than images and text an agent submits it’s request and the website reacts with an answer. Bypassing the web completely.
Here’s where you’ve got it wrong, from what I’ve learned from AI creators. The agent is bots are not designed to be commanded with a keyboard and mouse, that’s old bandwidth transfer based on old school technology. The future needs no iPhones, more wearables, like the Meta glasses, hand gesture technology, eye movement tracking.
This view makes A LOT more sense and is A LOT more realistic than the "SaaS is dead" because "The UI and Code will go away soon" Chicken little sentiment put out a couple of weeks ago giving the impression that the end is near. Maybe in 10-20 years that will be true, but we still have LOTs of agentic evolution to work on, inside of SaaS built for the physical world (i.e. "us").
Bojang is right. I have the same problem
Remember the Rabit R1? 😆
I wonder when early adopters will be able to make back the 200$ subscription through automated work through the net.
Paying to be guinea pigs instead of getting paid for their data sure is a choice
@@shorgothshort answer not with this
Just order thousands of pizzas for thousands of people. Wait is there a rate limit?
@ lol yea that would do some damage lol
The research preview to RLHF-train Operator via Pro Subscription is just that; a research preview to RLHF-train Operator. It’s a cool add-on, but people shouldn’t view this as some sort of special main selling point of Pro Tier lol
Signed,
- Written 20,000+ lines of bug-free code last month with Pro from scratch
2:03 or flip it around, make an API which sends the task to an LLM and does it.
I think it would be a good idea for webpages to add a button in a predictable place that leads to a “standardized” version of the webpage for operators to use. Then, operators know: “Start at this button on the homepage, then I will view its content designed for me to understand it more predictably on the next page.” That way, all the same content could be (less attractively) available in a single zone for operators to navigate much more easily. It could reduce errors and expedite operators usage on webpages that participate.
One agent that would be really good to have is one that can tidy up one's hard rives and remove unwanted duplicate files (or treat then as back ups) and find lost files and in collocate them to known locations so they can be more easily found an such.
That can be done with one or two clicks..you don't have time for a mouse click?
There's a dozen apps that already locate duplicate files on a hard drive, nothing new there. As for a "lost file" what makes it such? Windows knows where it is, an Agent would have no idea what YOU think is "lost", when a simple search which Windows Explorer has been able to do for years would "find it".
@@brianmi40 They may be able to find them great, but an agent could then fix the problem and not just find the file.
And in regard to lost files the users could describe what the file does even if they can't remember the exact name of the file.
@@brianmi40I don't want a dozen apps on my computer. One agent that can do it all asynchronously would be way better. That's what OP was hinting at.
@@GateOfSteins Not suggesting a "dozen", ONE app will find dups. YOU can find "lost files".
Skynet now has access to the internet.
What could go wrong, hmmmm?
Idk what happening. We just learnt that alignment isnt given at all anymore and they roll out agents.... This will go badly really quickly..
$200 per month is more separation of rich and poor
"skynet" has no acceso to its own source code neither the servers it is running on
Can’t wait for people to start spamming me with their productivity.
Oh, great! Now I finally don't need to do online banking anymore.
'We only collect data to make agents better" - Sure, sure. 🙃
2:36 Lol, you really think that's why these operators are being built, so that humans "can become supervisors and don't have to get our hands dirty"? Come on! Don't p... in my face and tell me it's raining. The whole point of AI is to replace human workers, not augment human workers.
Seems like life is just going to boil down to appeasing an algorithm.
Such phenomenal saturation on the web, especially now with generated content, perhaps the real world organic grass roots methods may become more viable.
At least when it comes to building that initial boost to get the algorithm and recognize you of course.