According to the Financial Association on January 6 (Editor: Xiaoxiang), if you are confused about the development of AI in 2025, then $Microsoft (MSFT.US)$ an article recently published on the official website may provide guidance for you.

Microsoft believes that by 2025, artificial intelligence will evolve from being a tool for work and home into an indispensable part of both.

This will be reflected in the following six major trends:

AI models will become more powerful and useful, AI agents will fundamentally change the way we work, AI companions will provide support in daily life, utilization of AI resources will become more efficient, testing and customization will become key to building AI, AI will accelerate scientific breakthroughs...

In Microsoft's view, AI-driven agents will accomplish more work with greater autonomy and help simplify your home and work life. On the global stage, AI will also assist us in finding new ways to tackle significant challenges such as the climate crisis and medical care. This advancement will benefit from innovative progress in AI's capabilities in areas like memory and reasoning.

Microsoft has committed to continuing its efforts to help people use and build AI safely and reliably. Chris Young, Microsoft’s Executive Vice President of Business Development, Global Strategy and Investment, stated, "AI has made the impossible possible. Over the past year, we have seen a large number of individuals and organizations transition from AI experiments to more meaningful applications, marking the beginning of this technology's comprehensive transformation of all aspects of our lives."

In the report, Microsoft noted that in just the past year, the usage rate of generative AI among business leaders and AI decision-makers has jumped from 55% to 75%. New AI tools will bring even greater potential. Want to know what the future holds? Here are six major AI trends to watch for in 2025, and how Microsoft will innovate in each trend.

AI models will become more efficient and practical.

Over the past year, AI models have become faster and more efficient. Today, large 'frontier models' can perform a wide range of tasks from writing to coding, while highly specialized models can be tailored for specific tasks or industries.

By 2025, AI models are expected to do even more and perform better.

Models with advanced reasoning abilities, such as OpenAI's o1, are already capable of solving complex problems through human-like logical steps. These capabilities will continue to play an important role in fields such as science, programming, mathematics, law, and medical care, enabling models to compare contracts, generate code, and execute multi-step workflows.

These advancements are crucial for model innovation, and progress in data organization and post-training is equally important. For example, Microsoft's small Phi model series demonstrates that organizing high-quality data can enhance model performance and reasoning ability. Meanwhile, Microsoft's Orca and Orca 2 showcase the power of synthetic data in post-training small language models, allowing these models to achieve performance levels previously attainable only by large language models, and outperforming in specialized tasks.

In 2025, improving model speed, quality, and specialization will create new, more useful AI experiences, including interactions with AI Agents.

AI Agents will fundamentally change the way we work.

Nearly 70% of Fortune 500 employees are currently using Microsoft 365 Copilot to handle repetitive and monotonous tasks, such as filtering emails and taking notes during Teams meetings. By 2025, a new generation of AI-driven Agents will take on more work and even handle certain tasks on your behalf.

"Agents can be seen as applications of the AI era," said Charles Lamanna, Vice President of Copilot Business and Industry. "Just as we use different applications to accomplish different tasks, agents will begin to change every business process, fundamentally transforming the way we work and manage our organizations."

With advancements in memory, reasoning, and multimodal capabilities, agents will handle more complex tasks, acquiring new skills and modes of interaction.

Businesses can reimagine processes, such as creating reports and human resources tasks, allowing employees to focus on higher-value work. Companies can also set up multiple agents to alert supply chain managers about inventory disruptions, recommend new suppliers, and execute sales orders, handling everyday challenges to help maintain sales.

All of this will lay the groundwork for the future, when businesses will have a range of agents (from simple prompts and responses to fully autonomous agents) that will independently or collaboratively represent individuals, teams, or functional departments to execute and coordinate processes.

AI companions will provide support in your daily life.

In addition to work, AI is expected to make parts of your life easier by 2025. Microsoft's Copilot can serve as your AI companion, providing support in your daily life. This means it will help simplify and prioritize tasks, such as dealing with daily information overload, freeing up more of your time while protecting your privacy, data, and security.

As Copilot evolves over the next year, it will help you stay more closely connected and will possess new capabilities. For example, Copilot Daily will read the relevant news and weather summaries to you every morning in a familiar voice.

When you choose to use Copilot Vision, it will be able to see what you are viewing online and discuss it with you, as it will understand the web pages you are looking at, allowing it to answer your questions and suggest next steps. Copilot will also be able to help you make decisions—it can assist you in furnishing your new apartment by searching for matching Furniture, then help you think about how to place all the Furniture in the best way.

And this is just the beginning. In the coming years, AI experiences will become increasingly accurate, and emotional intelligence will also improve for smoother interactions.

The utilization of AI resources will become more efficient.

Although AI requires resources such as energy, innovative solutions are helping to address this challenge. For instance, even though the global data center workload in 2020 was about nine times that of 2010, the actual power demands of data centers only increased by 10%.

This is partly because Microsoft is improving hardware efficiency either alone or in collaboration with Other companies such as AMD, Intel, and NVIDIA, including its custom silicon series, Azure Maia, Cobalt, and highly efficient cooling heat exchangers designed for large-scale AI systems.

In the coming years, new datacenters supporting AI will come online with zero water cooling and the company will also expand the use of ultra-efficient liquid cooling systems like cold plates. All of this is part of a broader effort to make the infrastructure that AI relies on more efficient and sustainable by 2025.

As Microsoft helps establish a more efficient AI infrastructure, the company will also continue to invest in and utilize more low-carbon building materials, such as near-zero carbon Steel, concrete alternatives, and cross-laminated timber. Microsoft will also continue to invest in and utilize carbon-free energy sources like Wind Energy, Geothermal Energy, nuclear energy, and CECEP Solar Energy. The company is making long-term investments to bring more carbon-free power into its Operation grid and continues to advocate for the promotion of clean energy solutions globally.

Mark Russinovich, Chief Technology Officer of Azure, Deputy Chief of CSI Information Security Index, and Technical Researcher, stated that in 2025 and beyond, we will increasingly gain a comprehensive understanding of datacenters, energy, and resources, thereby maximizing the efficiency of the entire infrastructure.

Testing and customization will become key in building AI.

Testing is critical for defining and assessing the risks of AI, which is essential for responsibly building AI. One of the main developments in AI by 2025 may be summed up in two words: testing and customization.

If risks and threats can be tested, it will help address or mitigate these two issues. For example, this means detecting and addressing unfounded content, known as "hallucinations," which are inaccurate responses AI frequently makes.

Sarah Bird, Chief Product Officer for AI at Microsoft, stated that part of the work being done by Microsoft to build safe AI applications includes developing rigorous and comprehensive testing. In addition to assessing internal threats like hallucinations, testing will also better identify external and increasingly complex attacks.

"Even as models become safer, we need to elevate testing and measurement to the level of the most severe threats we see—testing represents a complex adversarial user and what they are capable of doing," Bird said. "We already have a foundation, and we will continue to improve. "

People will gain greater control over how AI applications operate within their organizations. Individuals will be able to customize applications, filter content, and establish protections that suit their work. For example, a gaming company will be able to specify the types of violent content that employees involved in building games can see.

AI will accelerate scientific breakthroughs.

AI has already had a tremendous impact on the world, driving advances in various fields, from supercomputing to weather forecasting—its not only propelled historic breakthroughs in scientific research but is also expected to unleash new potential in the fields of natural sciences, sustainable materials, drug development, and human health.

For instance, in 2024, Microsoft Research achieved a breakthrough that will enable researchers to explore some of the most challenging biomolecular science questions in the world with unprecedented speed and accuracy, including discovering new life-saving drugs. Using an AI-driven protein simulation system, researchers found a new way to simulate the dynamics of biomolecules. This method, called AI2BMD, can help scientists tackle problems previously deemed unsolvable, advancing biomedical research in protein design, enzyme engineering, and drug discovery.

The impact of AI on the field of science is expected to continue to expand. In 2025, one trend is certain: AI will continue to drive innovation and unleash new potential for people and organizations around the world.

Ashley Llorens, Vice President and Managing Director of Microsoft Research, stated that one of the most noteworthy things in 2025 will be how AI applications in scientific research will drive solutions to some of the world's most pressing problems. "We will begin to see these tools have a measurable impact on the productivity of the people and institutions dedicated to solving these enormous issues, such as designing sustainable materials and accelerating the development of life-saving drugs."

编辑/jayden