Why are the cost of developing artificial intelligence very high?

After more than 18 months of obsession with obstetrician artificial intelligence, a number of the largest technology companies have proven that this technology can be a real source of income, but it is also a big hole for expenses. Microsoft and Google, attached to Alphabet, announced the increase in the revenue of the Cloud Services department in their last quarterly results, as clients have spent more on artificial intelligence services. The company “Meta Platfarms” explained that its efforts in the field of artificial intelligence have helped improve the interaction of users and target advertising, although it is still very late to achieve profits from this technology. To achieve these early profits, the three companies have spent several billion on the development of artificial intelligence, and they plan to increase these investments more. On April 25, Microsoft announced that its capital expenditure was $ 14 billion in the last quarter, and that he “significantly increased these costs”, partly driven by investments in artificial intelligence infrastructure. It was an increase of 79% from the same quarter of the previous year. ‘Alphabet’ spent $ 12 billion during this term, an increase of 91% over the previous year, and expected to spend ‘at the same level or more’ during the rest of the year, amid the focus on the chances of artificial intelligence. Meanwhile, “Mita” has increased its estimates for investments for this year, and now believes that capital expenditure will extend between $ 35 billion to $ 40 billion, representing an increase of $ 42% on the maximum range. She referred to intense investment on research on artificial intelligence and its development. The cost of artificial intelligence has surprised the high cost of artificial intelligence some investors. The “Mita” share has dropped especially to the expected expectations of expenses with a slower growth. But within the technology industry, it has always been known that the cost of artificial intelligence would rise. This is due to two main reasons, namely that artificial intelligence models have become larger and more expensive in development, except that the global demand for artificial intelligence services needs to build more data centers. Companies experimenting with artificial intelligence can spend millions to adjust products from “Openai” or Google. Once prepared and operated, an extra cost appears each time you send an investigation to a chat robot, or that it must be an artificial intelligence services to analyze sales data. But the most expensive job is to build the basis for these artificial intelligence systems. Here is a look at these efforts: The great language models become the largest that takes on the most famous artificial intelligence products today, such as the ‘Oben Ai’, on major linguistic models, which are systems fed from large amounts of data, including books, articles and comments online, to provide the best answers to user queries. Many leading artificial intelligence businesses bet that the way to artificial intelligence is more advanced, and perhaps even systems that can perform better in many tasks are to make these great language models greater. It requires more computer data and functions, as well as training of artificial intelligence systems for a longer period. Anthropic CEO Dario Amudi told ‘Anthropic’ in an interview via Podcast in early April that artificial intelligence models currently in the market cost about $ 100 million to train them. He explained that “the models that are now being trained, which will be launched later this year or early next year, are almost one billion dollars, and I believe it will amount to about $ 5 or $ 2026 in 2025 and 2026.” Slides and calculation costs are related to a large part of the cost. These chips are not the central processing units that have contributed to making ‘Intel’ and the smaller businesses running billions of smartphones famous. Artificial intelligence enterprises rely on graphic processing units that can process large amounts of data at high speeds to train large linguistic models. These chips are not only rare, but also expensive, knowing that the most advanced functions are mainly made by one business, “inviteia”. The “H100” graphic chip (H100 “was sold to” Inviteia “, which is the golden standard for training artificial intelligence models, about $ 30 thousand dollars, and some sellers offer them twice this amount. The most important technology companies need many of them, formerly the CEO of” Mita “, Mark Zucerberg, said that his business is able to support 350,000 H100. He gets a discount when buying large quantities of this slide will remain an investment of billions of dollars. According to a lawsuit submitted by the New York Times over the start of its upcoming articles to train artificial intelligence systems. But this improvement can fade with the push of the industry to build greater artificial intelligence models.