Artificial intelligence factors are a scientific fiction that are not yet ready for primitive time

This is it StepbackWeekly news message that breaks one basic story from the world of technology. To learn more about everything Amnesty International, Follow Hayden Field. STEPBACK reaches the subscribers ’inbox at 8 am Each time. Choose to Stepback here.

Everything began with Jarvis Yes, Jarvis is one of Marvel films.

Well, he might not do that Start With the Iron Man AI assistant, but the fictional system definitely helped the concept of artificial intelligence agent. When you met the people of artificial intelligence about Agency AI, they often refer to Jarvis as an example of the ideal artificial intelligence tool in several ways – one knows what you need even before you ask, they can analyze and find visions in large groups of data, and can provide strategic advice on certain aspects of your work. People sometimes do not agree on the careful definition of the artificial intelligence factor, but in essence, it is a step that goes beyond chatting keys in that it is a system that can perform multiple multiple tasks on your behalf without constantly needing to communicate with you. It makes the list of its sub -tasks mainly in order to reach your preferred final goal. This imagination is closer to being many facts, but when it comes to the actual benefit of the daily user, there are a lot of things that do not work – and may never work.

The term “AI Agent” was present for a long time, but it is especially She started heading in the technology industry in 2023. That was a year concept One of the factors of artificial intelligence; The term was on everyone’s lips as people tried to get out of the idea and how to make it a reality, but you did not see many successful use. The following year, 2024, was a year Publishing People were putting the code in the field and seeing what he could do. (The answer was, at the time, … not much. It is full of a set of error messages.)

I can determine the noise around artificial intelligence agents become widespread for one specific advertisement: in February 2024, Klarna, Fintech, He said One month later, his help in artificial intelligence (supported by Openai’s Tech) succeeded in making 700 customer service agencies full -time and two -thirds of the company’s customer service talks. For several months, these statistics appeared in almost every Amnesty International Industry.

The noise never retreated, and in the following months, all the CEO of the Great Tech was moving on the period in each call call. Amazon, Meta, Google and Microsoft executives and a full group of other companies began talking about their commitment to build useful and successful artificial intelligence agents – and tried to put their money as their mouths occur.

The vision was that one day, Amnesty International’s agent could do everything from your book of travel to create pictures of your job performances. You can find the perfect tool, for example, a good time and place to hang out with a group of your friends who work with all calendars, food preferences and food restrictions – then reserve dinner reservation and create a evaluation event for everyone.

Now let’s talk about “artificial intelligence coding” for everything: For years, artificial intelligence coded was carry Artificial intelligence agent industry. If you ask anyone about realistic, successful, non -exciting use cases that occur now And not from the concept in an insufficient future, they refer to the coding of artificial intelligence-and this was largely the only concrete thing that they could refer to. Many engineers use artificial intelligence agents for coding, and they are considered well objectively. Good enough, in fact, in Microsoft and Google, Up to 30 percent From the code now it is written by artificial intelligence agents. For startups such as Openai and Anthropor, which are burned through high rates, one of their largest revenue generators is AI coding tools for institutional customers.

So until recently, artificial intelligence coding was the state of using the main realistic life of artificial intelligence agents, but this is clearly not flowing to the daily consumer. Remembering the vision, it has always been a kind of artificial intelligence agent for “everyman”. We are not there – but in 2025, we approached more than we were before.

Last October, things began in Antarbur by providing “computer use”, a tool that allowed Claude to use a computer such as the Human force – browsing, search, access to various platforms, and completing complex tasks on behalf of the user. The general consensus was that the tool was a step forward for technology, but the reviews said that in practice, she left a lot of desirable. Fast forward until January 2025, the Openai operator, which was released, copied with the same thing, and classified it as a tool to fill the models, request the grocery, book travel, and create memes. Again, in practice, many users agreed that the tool that was carried by animals was slow, and not always effective. But again, it was an important step. The next month, Openai Deep Research, an AIGNIC AI tool that can collect long research reports on any user’s subject, which things also rotate forward. Some people said that research reports were more impressive than the content, but others were seriously admired. Then in July, Openai merged deep research and operator into the AI ​​agent product: Agent Chatgpt. Was it better than most AI Agenic tools facing the consumer that came before? definitely. Is it still difficult to work successfully in practice? definitely.

Therefore, there is a long way to reach that vision of the ideal AII factor, but at the same time, we are technically closer to what we were before. For this reason, technology companies put more and more money in AIGEC AI, by investing in an additional account, research, development or talent. Google recently rented the CEO of Windsurf, founder, and some members of the research and development team, specifically to help Google pushing AI’s agent’s projects forward. Companies like anthropor and Openai races with each other over the ladder, which is included by Rung, to introduce gradual features to put these factors in the hands of consumers. (Anthropor, for example, I just announced The Chrome Chrome extension allows it to work in your browser.)

In fact, what happens after that is that we will see the coding of artificial intelligence that continues to improve (and unfortunately, it may replace the functions of many Software engineers for beginners). We will also see that the agent’s products facing the consumer are improving, most likely slowly but steadily. We will see the agents who are increasingly used for institutional and government applications, especially since all anthropoes, Openai and Xai have appeared in government AI platforms in recent months.

In general, expect to see more wrong beginnings, beginnings and stopping, integration and acquisitions with the selection of the artificial intelligence agent (the noise bubble in the balloon). One question we all have to ask ourselves with the continuation of the months: What do we really want the conceptual “artificial intelligence agent” to be able to do it for us? Do we want them to replace logistical services or the personal human aspects of life only (that is, help in writing wedding elites or a note for the delivery of flowers)? What is the quality of logistical services against personal things? (Answer the last: not very good at the moment.)