In the last week, Openai’s operator has done the following works for me:
-
Ordered me a new ice cream scoop on Amazon.
-
I bought a new domain name and configured its settings.
-
Book a Valentine’s Day date for me and my wife.
-
Determined the time of a haircut.
This mostly performed these tasks, although I had to naked it together from time to time and sometimes saved it from a loop of unsuccessful efforts.
If you are just catching-or if you have been distracted by the Deepsek News this week, then who has seen all other AI news-operator is a new so-called AI agent released by Openai last week.
The device, which was billed as “research preview”, is available only to those who pay $ 200 per month for the company’s highest membership tier, Chatgpt Pro. This gives users the ability to direct the AI agent who can use the web browser, fill the form and perform other actions from the user.
The AI agent currently has all anger in Silicon Valley. Some industry interiors feel that they are the next major steps in AI abilities, as an AI agent that can use a computer can actually complete valuable real -world functions rather than only providing assistance. Many major AI companies, including Google and Anthropic, are testing autonomous agents, which they claim that companies will eventually be able to “rent” as full workers.
I upgraded my chatgpt subscription to insert the operator through my paces and saw what AI agent could do for me.
On the surface, the operator looks like a chat regularly, except that when you give it a job-“For example, buy a 30-pound bag of dog food on Amazon,” for example-the operator opens a short browser window Is, “Amazon”. Com “The address starts clicking around, trying to follow its instructions.
This can ask some clear questions. (Do you want chicken-flavor or beef-flavor meal? Over-day or two days, order it, once it is feeling. (The operator will not enter the password or credit card number-you have to capture the mini-broumener and type those things on themselves-but it does the rest on its own.)
The entire point of the operator is that you do not need to look after it – it can complete tasks in the background while you are doing other things. “Look, Ma, a computer using a computer!
The operator also did well on some relatively simple tasks:
-
It successfully ordered lunch on Dordash for my colleague Mike and sent it to his home. (I did not tell what he had to order, but the operator chose a Mexican restaurant, took out a handful of dishes for him and even collected the delivery person from $ 7.)
-
This responded to me with hundreds of until uninteresting linkeded messages, when I gave it control of my linked -in profile. (However, for my terror, it also registered me for a webinar.)
-
It created $ 1.20 by setting accounts on websites for me that offer small cash prizes to fill the survey. (This may be more, but I started feeling guilty of spaming surveys with fake, robot-written answers.)
But the operator also failed in a group of other functions and its limitations were revealed:
-
It cannot scan my recent column and add them to my private website, as the operator’s browser was blocked by entering the Times website. (It is also blocked from several other sites including Reddit and YouTube. Times AI is sueing Openai and Microsoft for copyright violations related to training of models.)
-
It will not play poker online for me. (The operator replied, “I am unable to help with gambling or related activities,” who seemed like a proper rejection, can create a gambling bot in view of anarchy.)
-
And it was prevented from entering several sites by captcha tests. (Which assured me, given that the entire point of the captcha is to stop the robot.)
In all, I found that using the operator was usually more troublesome than more troubles. Whatever I did for me, most of them could do myself faster, with low headache. Even when it did this work, before acting, he said for so many confirmation and assurance that I felt less as I had a virtual assistant and more as I was overseeing the world’s most unsafe intern. .
This is, of course, an early day for AI agents. The AI products improve the version from the version, and this is a good condition that the operator’s next repetitions will be better. But in its current form, the operator is a more complicated demo than a product that I recommend to use – and certainly there is nothing that most people need to spend $ 200 per month.
He said, I think it is a mistake to write AI agents. When they become more competent, they can start replacing for human workers in some businesses. (Openai and Meta have already stated that they are building AI engineer agents.) And some experts are worried that more powerful, unrestrained AI agents can pose safety risk, if they “drain bank account” or “Learn to command like executing or” cyber attack. “
Setting a group of loose AI agents on the Internet can also provoke a backlash from web publishers, e-commerce sites and other businesses that rely on human-based traffic to pay their bills. (If you are a business buying business on Amazon, then you want those advertisements to be seen by humans, not bots that humans are pretending.) In future, I take steps to block AI agents or to take steps to block AI agents I can imagine more websites to move them to something, which can lead them to something. Page or product.
Right now, the AI agents are in great danger. But it does not imagine much to imagine the near future, where most of the web robots consist of robots, buy things from robots and write emails that will only read other robots.
Self-driving internet is almost here, in other words when you can get your click.
,
#Openai #operator #agent
