We’re a few weeks away from a new year, and Silicon Valley’s obsession with artificial intelligence isn’t going anywhere. Yes, the biggest trend of 2024 will continue to dominate 2025. But there will be some new wrinkles over the next 12 months, as well.
Companies will begin releasing more powerful AI models, while AI agents will gain prominence. The industry will also focus on bringing AI features to more consumers and work to make AI data centers more energy efficient.
According to Ashley Llorens, corporate vice president and managing director at Microsoft (MSFT) Research, AI models will soon be able to handle far more complex tasks.
“What we’ve got now is we’ve got AI that can reason better, that can perceive the environment in more sophisticated ways,” Llorens told Yahoo Finance. “And so what that’s going to mean is that we’ll be able to delegate a more sophisticated set of tasks to the AI to complete on our behalf.”
That’s where AI agents come in. Think of agents as autonomous or semi-autonomous apps that can perform specific tasks for you. They’re more capable than the traditional chatbots like ChatGPT that you’re likely used to, but can still be controlled via natural language inputs.
You can, for example, tell an AI agent how to triage incoming customer requests or pull information out of invoices and drop it into the appropriate spreadsheet to more easily track employee expenses without having to do it all yourself.
Llorens says that Microsoft is using AI agents to connect employees across its organization.
“We have this capability we call ‘coffee connections.’ And it’s essentially the idea that I delegate a half hour on my calendar every week to an AI system that tells me who I should have a coffee chat with from anywhere in the organization,” Llorens explained.
“We have a background process where AI is essentially simulating lots of random conversations between people, analyzing the outcome to see which conversations are most interesting … and then using that to propose … here you go, talk to this person, and by the way, you may want to hit these topics.”
AI will also become increasingly multimodal over the next year, helping it to do things like interact with text, as well as visual and audio inputs. Microsoft says you’ll begin to see this when it launches its Copilot Vision for Copilot in Windows. The feature, which will be opt-in, will be able to see what you’re looking at on a web page and then allow you to ask questions about it.
So, if you’re looking at a movie, you might be able to ask Copilot Vision who’s in it using your voice or text without naming the movie, and the app will know you’re talking about the movie poster or web page you’re looking at and provide the appropriate response.