The low cost of artificial intelligence will not solve all problems

Artificial intelligence suffers from a major problem as it is very expensive for technology that helps businesses reduce it. The laws of expanding artificial intelligence, which emphasizes the need to increase computer power to make stronger models, has placed technology companies in a race to spend billions of dollars to build large data centers and buy strong electronic slides, which are the costs that businesses cannot download to their clients. The Google instrument that helps office employees create documents or email with artificial intelligence is not cheap. It adds $ 20 to the bill that the employer carries monthly, and its value is $ 6 for the company’s work area (workspace) for each employee. The artificial intelligence assistant, “copilot”, food of “Microsoft” $ 30 per month per worker. Meanwhile, the distribution of artificial intelligence could cost directly in the company’s systems between $ 5 million and $ 20 million, according to the Gartner Research Company, which expects to repay 30% of the IQ projects by the end of 2025, partly as a result of all these expenses. The good news of these clients is that the cost of artificial intelligence has been reduced, which helps to fill the gap between the return and investment. The bad news is that it does not address the biggest problem of its use, which is expected to be resolved for a few years. Capital spending is the wisdom of the Silicon Valley, the wise opinion that prevails in the Silicon Valley, is to continue to gain a foothold in the future. On Tuesday, Microsoft announced that its capital expenditure was a record 19 billion in the last quarter, or more than 80% higher than last year. CEO Satia Nadala said all these investments would continue to “exploit the opportunity of artificial intelligence.” Sandar Bishy, ​​CEO of “Alphabet”, repeated the same opinion at a recent phone conference on Google’s business results and explained that “the risk of investment deficit is much greater than the risk of excessive investment for us.” Investors were not entirely convinced, as the shares of “Microsoft” have fallen by 2% since the announcement of its latest profits, and Google shares fell by 5%. But even with the high cost of artificial intelligence training over the years, the artificial intelligence services of both technology giants appear to be moving in a cheaper direction. A spokesman for “Goeni”, which can use businesses to automate or summarize customer service, is stronger than its predecessor, although it is close to half its price. The most recent models of “Oben Ai”, known as “GPT-4O”, faster than the previous model “GPT-4 Turbo”, but it’s also 50%cheaper than that. A spokesman for the company told me that the cost of access to its models, measured by the treatment of symbols (which are basically words by the linguistic model), has fallen by 99% since 2022. She added: “We are committed to continuing this path.” Reducing the costs associated with developing models for scientists for artificial intelligence, reducing costs using techniques such as “reduction in size and scattering” and “quantitative division” was an important focus axis at recent conferences. Nairrav Kingzland, an executive officer at the competitive company, “Anthrobick”, told me that it is logical that the cost of its models will fall to 25% of its current price during the next year or two years, and that the company that collected $ 8.8 billion from investors, including ‘Google’ and ‘Amazon Dotcom’, the cost of developing a modern model have. There are other signs of low cost. In China, artificial intelligence businesses were involved in a price war that led to the low prices of using obstetric intelligence, partly due to the organizational environment, the low -thech organizational environment, low labor costs and government support. For example, a boot of artificial intelligence is called ‘Deepseek’ $ 0.14 per million symbols of business people, while a similar model of “Oben Ai” costs $ 10. Corporate management efficiency comes the efficiency of cost management by corporate employees who also use these applications. Many companies realize that they do not need the most powerful models for artificial intelligence to give their employees a productive benefit, which is why they try open models of businesses such as ‘Meta platforms’. Customer service robot may need an advanced artificial intelligence tool that can do immediate reasoning, but analyzing customer calls to improve service can be done with less advanced technology. Man Group, a global asset management company, tells me that the models I use to summarize the texts for their investment portfolios, or to reduce daily work to 30 minutes for other employees, are already declining. But there are questions about this future price road. Silicon Valley has a history of price subsidies, as broadcasting latforms, participatory transport applications and cloud services are all lost from the profit margin to increase its share in the market. The goal is to win competition, eventually raise prices and switch to profitability. But here lies the thorny point for obstetrician artificial intelligence, there is still a widespread question about the benefit of it for the profits of companies, which, according to Gartner, is the main reason for his expectations that 30% of the projects will be abandoned by the end of next year. If this technology remains in just chat robots and the summary of texts, it may not even earn low prices. This is the problem that technology companies need to address more than others, even more than the problem with costs.