Last week, the Openai operator did the following things for me:
-
He ordered me a new ice cream script on Amazon.
-
Buy a new field name and create its settings.
-
I booked Valentine’s Day for me and my wife.
-
A haircut.
I did these tasks mostly independently, although I had to push them from time to time and sometimes save them from a failed episode.
If you just catch a knee-or if you have paid your attention by Deepseek News this week, which overwhelmed all other artificial intelligence news-the operator is a new customer called artificial intelligence Issued last week By openai.
The tool, which has been described as a “research preview”, is only available to people who pay 200 dollars per month for the highest subscription category in the company, Chatgpt Pro. It gives users the ability to direct the artificial intelligence agent that can use a web browser, fill out forms and take other action on behalf of the user.
Artificial intelligence agents are all anger in Silicon Valley now. Some of those familiar with the industry believe that they are The next big step in the capabilities of artificial intelligenceSince the artificial intelligence agent who can use the computer can already accomplish valuable realistic tasks, rather than just providing assistance. Several leading artificial intelligence companies, including Google and Anthropic, are tested, independent agents claiming that companies will ultimately be able to “employ” complete workers.
I upgraded the ChatGPT subscription to put the operator through his steps and know what artificial intelligence agent can do for me.
On the surface, the operator is somewhat similar Try to follow your instructions.
He may ask some illustrative questions. (Do you want food with chicken flavor or flavor of beef? Shipping overnight or two days? (The operator will not enter passwords or credit card numbers-You must take over the small mixture and write these things in yourself-but he does the rest on its own.)
The complete goal of the operator is that you do not have to supervise it – you can carry out tasks in the background while doing other things. But I found myself attached to the window, fascinated by a web browser scene that leads the self -clicking on the buttons, writing words in boxes and choosing from the drop -down lists, all of which are on its own. See, what, a computer uses a computer!
The operator has also impressed in some of the relatively simple tasks that I gave:
-
I have successfully ordered lunch on DooDash to my colleague Mike and sent him to his home. (I did not tell her what I ask, but the operator chose a Mexican restaurant, and chose a handful of dishes for him, and until a person tends to be $ 7.)
-
Hundreds of LinkedIn messages have not been read to me, after I gave her control of my LinkedIn profile. (Although my horror, I also recorded me at an online symposium.)
-
It has made 1.20 dollars by preparing accounts on web sites that offer small cash rewards to fill the surveys. (Perhaps it is more than that, but I started to feel guilty because of sending an email to the surveys with fake answers written on the robot.)
But the operator also failed in a set of other tasks and revealed its limits:
-
You could not wipe modern columns and add them to my website, because the operator’s browser was banned from entering the Times. (It was also banned from a number of other sites, including Reddit and YouTube. The Times sues Openai and Microsoft for violating copyright related to training artificial intelligence models.)
-
The poker will not play online. (The operator answered, “I am unable to help with gambling or relevant activities,” which seemed to be a reasonable refusal, given the chaos that a gambling robot can create.)
-
He was prevented from logging into a number of sites through CAPTCHA tests. (This is what I found reassuring, given that the complete goal of Captchas is to deter robots.)
In general, I found that the use of the operator was usually more problematic than it was worth. Most of what I did for me was faster, with fewer headaches. Even when I succeeded, I asked for a lot of assurances and insurance before working so that I felt less as if I got a virtual assistant and more than that I supervised the most trainees in the world.
This, of course, is the first days of artificial intelligence agents. Artificial intelligence products tend to improve the version to the version, which is a good bet that the following repetitions of the operator will be better. But in its current form, the operator is an interesting offer more than a product that is recommended to use it – and certainly not something that most people need to spend $ 200 per month.
However, I think it is wrong to remove artificial intelligence agents. When they become more capable, they can start replacing human workers in some professions. (Openai and Meta already said they are building agents of Amnesty International. .
The status of a group of artificial intelligence agents on the Internet can also cause a violent reaction from web publishers, e -commerce sites and other companies that depend on the traffic that was created on humans to pay their bills. (If you are a business to buy ads on Amazon, you want humans to see those ads, not robots that pretend to be human.) In the future, I can imagine more pages or products.
Currently, artificial intelligence agents are very incompetent so that they do not pose a major threat. But it does not require much imagination to visualize the near future, as most of the web will consist of robots that speak to robots, buy things from robots and write emails that only other robots will read.
The internet that drives the car is almost here, in other words-get your clicks while you can.
