The automated future has just approached a few steps. In the past few weeks, almost all major IA – Opennai, Anthropic, Google, XAI, Amazon, Microsoft and Perplexity companies, among others – have announced new products that are not focused on questions or make their human users a little more effective, but to accomplish the tasks themselves. They are presented for their ability to “reason” as people do and serve as “agents” who will eventually do complex work from start to finish.
Humans will always push these models, of course, but they are designed to help fewer people do the work of many. Last month, Anthropic launched Code ClaudeA coding program that can do a large part of the work of a human software developer but much faster, “reducing development time and general costs”. The program actively contributes to the way a colleague would do it, writing and deploying the code, among others. Google now has a wide available “working horse modelAnd three separate AI companies have products named Deep Research, which quickly bring together and synthesize information in the name of a user. OPENAI shutters The ability of its version to “accomplish research tasks in several stages for you” and to accomplish “in tens of minutes which would take a human of many hours”.
AI companies have been built for a long time and benefit from the story that their products will ultimately be able to automate major projects for their users, move jobs and perhaps even business professions or sectors. As early as 2016, Sam Altman, who had recently co -founded Openai, wrote in a blog that “as technology continues to eliminate traditional jobs”, new economic models may be necessary, such as a universal basic income; He has warned several times since when the AI will disrupt the labor market, developer My colleague Ross Andersen in 2023 that “jobs will definitively disappear, a complete judgment”.
Despite the disturbing nature of these comments, they remained firmly in the field of speculation. Two years ago, Chatgpt could not carry out basic arithmetic, and criticisms have long hacked On prejudices and mythomania of technology. Chatbots and IMA image generators have become known to help children cheat homework and flood the web with low grade content. Significant applications quickly emerged in certain professions – coding, aligning customer service queries, writing a Plaque de Passe -Partout copy—But even the best models of AI were clearly not able to precipitate the displacement of generalized work.
Since then, however, two transformations have taken place. First, the search for AI has become standard. Chatbots have exploded in popularity because they could clearly – although frequently incorrectly—Aupte human issues. Billions of people were already used to asking questions and finding information online, making it an obvious use of AI models that could have otherwise seem to be research projects: now 300 million people use Chatgpt each week, and more than a billion use the preview of Google AI, according to companies. Further emphasize the relevance of products, media companies, especially The Atlantic–Lucrative agreements signed with OPENAI and others to add their content to AI research, bringing both legitimacy and an additional examination to technology. Hundreds of millions have been accustomed to AI, and at least part found the technology useful.
But although simple chatbots and AI research introduced a major cultural change, their commercial prospects were always small potatoes for technology giants. Compared to traditional research algorithms, AI algorithms are more expensive to execute. And research is a former commercial model that generative IA could only improve, which may cause some additional clicks on paid advertisements or producing a little more user data To target future advertisements.
The refining and expanding of the generative AI to do more for the professional class – not only students who rush into the long -term documents – this is where technological companies see the real financial opportunity. And they built to grasp it. The second transformation that led to this new phase of the AI era is simply that technology, while being riddled with bias and inaccuracies, has legitimately improved. The slate of reasoning models known in recent months, such as O3-Mini of Openai and Grok 3 from Xai, has in particular impressed. These AI products can be truly useful and their applications Advance scientific research could prove the rescue. Economists,, doctors,, codersand other professionals are largely comment On how these new models can speed up their work; A quarter of this year’s cohort technological start-ups at the prestigious incubator Y Combinator said that 95% of their code was generated with ai. Large companies – McKinsey, Moderna and Salesforce, to name a handful – now use it in essentially all aspects of their businesses. And the models continue to become cheaper and faster.
The leaders of technology, in turn, have become blurred on their hopes that AI will become good enough to make the work of a human. In a meta winning call at the end of January, CEO Mark Zuckerberg said“2025 will be the year when it is possible to build an IA engineering agent” who is also qualified as “a good intermediate level engineer”. Dario Amodei, CEO of Anthropic, recently said In a conference with the Council of Foreign Relations, the AI ”will write 90% of the code” in a few months, although still with human specifications, he noted. But he continued: “We will end up reaching the point where AIS can do everything that humans can”, in each industry. (Amodei, we must mention, is the ultimate techno-optimist; in October, he published a sprawling manifestEntitled “Machines of Loving Grace”, which posed the development of AI could lead to “defeat of most diseases, the growth of biological and cognitive freedom, the lifting of billions of poverty to share new technologies, a rebirth of liberal democracy and human rights.”) Altman used Altman. Language of the same large Recently, imagine countless workers in virtual knowledge moving in all industries.
These brilliant visions have decreased considerably when put into practice: Elon Musk and the Ministry of Government efficiency efforts Replacing human civil servants with AI can be the clearest and most dramatic execution of this game book, with massive loss of employment and a little more than chaos to show so far. Meanwhile, all generative problems with AI models with biases, inaccuracy and bad quotes remain, even if technology has progressed. OPENAI image generating technology is still struggling to produce people with the many appendages. Salesforce would be in difficulty To sell your AI agent, agentForce, to customers due to precision problems and concerns about the high cost of the product, among others. Nevertheless, the company continued its argument, just like other AI companies have continued to iterate and promote products with known problems. (In a recent profit call, the CEO of Salesforce, Marc Benioff, said that the company had “3,000 agent customers for forces who experience unprecedented productivity levels”.) In other words, defective products will not stop the push of technological companies to automate everything – the natural future of AI will be at best imperfect, but it happens anyway.
The motivations of the industry are clear: Cloud companies of Google and Microsoft, for example, increased rapidly in 2024, driven considerably by their AI offers. Meta business tent, Clara Shih, recently said CNBC that the company expects that “each company” uses AI agents, “the way companies today have websites and email addresses”. Openai is would have Considering invoicing $ 20,000 per month for access to what he describes as research agents at the doctorate level.
Google and Perplexity did not respond to a request for comments, and a Microsoft spokesperson refused to comment. An Openai spokesperson pointed me to a essay From September in which Altman wrote: “I have no fear of lacking things to do.” He may well be right; The Labor Statistics Office projects The AI considerably increases the request for computer and commercial occupations until 2033. An anthropic spokesperson referred to me to the initiative of the start-up to study and prepare the effect of AI on the labor market. The first article of effort research analyzed millions of conversations with the Claude of Anthropic model and revealed that it was used to “automate” human work in 43% of cases, such as identification and repair of a software bug.
Technological companies reveal, more clearly than ever, their vision of a post-work future. Chatgpt started the generative-Ai Boom not with incredible commercial success, but with psychological success. The chatbot was and is always possible losing Company money, but he has exhibited Internet users from around the world to the first popular computer program that could hold an intelligent conversation on any subject. The advent of the research on AI may have played a similar role, presenting limited chances for immediate benefits, but living – or perhaps inoculating – millions of people to robots that may think, write and live for you.