Instruments with generative artificial intelligence will save companies a lot of time and money if models learn to admit when they simply don't know what to do.
This was explained by Vik Singh, vice president of Microsoft, to AFP.
"To be really frank, what's really missing today is a model to raise their hands and say, 'Hey, I'm not sure, I need help,'" Singh said.
Since last year, Microsoft, Google and their competitors have been rapidly rolling out artificial intelligence generative applications like ChatGPT, which create all kinds of content on demand and give users the illusion of omniscience.
But despite the progress, they still "hallucinate" or make up answers.
It's an important problem that Copilot's CEO needs to address: Singh's corporate clients can't afford to let their AI systems go off the rails, even from time to time.
Marc Benioff, chief executive officer of Salesforce, said this week that many customers are increasingly frustrated by the ramblings of Microsoft's Copilot.
Singh insists that a "really smart crowd" will try to get the chatbot to admit "when it doesn't know the right answer, and ask for help."
"Real economies"
According to Singh, a more modest model would be no less useful. Even if the model has to turn with the help of a human in 50% of the cases, this saves "tons of money".
At one of Microsoft's clients, "every time a new request comes in, they spend $8 to have a customer service representative answer them, so there are real savings, and it's also a better experience for the customer, because he gets a faster answer".
Singh arrived at Microsoft in January and this summer led the team developing Copilot, Microsoft's assistant that specializes in sales, accounting and online services.
These applications have the critical task of generating revenue and justifying huge investments in generative II.
In the midst of the post-II craze, startup companies that develop the technology promised vastly improved systems that would "elevate humanity," according to Sam Altman, head of OpenAI, which is primarily funded by Microsoft. But for now, the new technology is mainly used to increase productivity and improve productivity.
Copilot can do surveys for merchants, freeing up time to call clients. Lumen, a telecommunications company, saves around 50 mln. dollars annually in this way.
Singh's teams are working on integrating Copilot directly into the tech giant's software and making it more autonomous.
"Let me say that I am a sales representative and I have obligations from a client," said the executive director. Two weeks later, the model can "prompt the rep to go forward, or better yet, just go and automatically send an email on behalf of the rep because he's been approved to do it."
In other words, before finding a solution to the problem of global warming, II is expected to free humanity from tedious, recurring obligations.
"We are not in the first phase," said Singh. "A lot of these things are based on productivity, but they obviously have huge benefits."
Did all these increases in productivity translate into job losses?
Managers of large companies, such as K. Kritivasan, the head of the Indian IT giant TCS, announced that Generative II will destroy all call centers.
But Singh, like many Silicon Valley executives, believes that technology will make people more creative and even create new jobs.
You tried that in Yahoo in 2008, when a number of editors were choosing articles for the main page.
"We came up with the idea of using artificial intelligence to optimize this process, and some people asked, 'Oh my God, what's going to happen to the servers?'" Singh explained.
However, the automated system made it possible to update the content more quickly, thus increasing the number of clicks on the links, as well as the need for new articles.
"In the end, I needed to hire more editors," Singh said. | BGNES
