© Unplash/Markus Spiske
Since the start of the year, the tech world has been obsessed with generative artificial intelligence. This new advance, popularized by ChatGPT, makes it possible to entrust certain tasks to AI. But since generative AI is very powerful, it also consumes a lot of resources. And if tech giants like Microsoft or Google are investing massively in technology, the Wall Street Journal explains, in an article, that for the moment, we do not know exactly how these investments will be profitable in the short term.
The WSJ cites the example of GitHub Copilot, an AI offered by Microsoft that helps developers generate computer code. Despite the fact that this tool is paid, 10 dollars per month, and that it already has 1.5 million users, it would not always be profitable. In fact, the average cost per user of GitHub Copilot would be $20. And here, we are talking about a paid tool, while Microsoft also offers services based on artificial intelligence for free, such as GPT-4 which is integrated into Bing Chat, or DALL-E 3 which is integrated into its generator images.
So, if for the moment, the tech giants are very enthusiastic about offering new features based on generative AI to their users, next year, they could adopt more pragmatic attitudes that take into account the profitability. “Next year, I think, will be the year where the slush fund for generative AI disappears,” explains May Habib, CEO of the generative AI company Writer, to the Wall Street Journal, according to our colleagues at 'Ars Technica.
In other words, it is possible that tech giants will stop spending lavishly, which will impact users, one way or another. In order to reduce the costs of generative AI, while offering this technology to their consumers, some players are already using two models for the same functionality: one for greedy queries and a smaller model for simple queries.
It is also possible that on certain services, tech players decide to apply restrictions in order to limit costs. The other solution would be to execute certain tasks locally, on the user's device, so that the servers are less busy. Moreover, according to our colleagues at Thurrott, Microsoft is already pushing the PC industry to put chips dedicated to AI on their machines. And at the same time, it would develop its own processors dedicated to AI to reduce costs.
- If functionalities based on generative artificial intelligence have become very popular, companies do not yet know how to make them profitable, according to an article in the Wall Street Journal
- The American newspaper cites the example of the GitHub Copilot tool (an assistant for coders) which is offered at $10 per month to developers, but whose average cost per user would be $20
- Next year, tech giants could stop spending lavishly and think about profitability