Artificial intelligence agents are becoming an exciting frontier in technology, with major companies like Google DeepMind, OpenAI, and Anthropic racing to enhance large language models to manage tasks autonomously. This trend, referred to as agentic AI, is capturing attention in Silicon Valley, with industry giants announcing groundbreaking developments.
Recently, Sam Altman, CEO of OpenAI, shared, ‘We believe that, in 2025, we may see the first AI agents join the workforce and materially change the output of companies.’ This assertion reflects a significant shift in how businesses might operate with AI’s assistance.
An agent in this context is defined as a software system capable of performing tasks with minimal to no supervision. The more sophisticated the task is, the more intelligent the agent must be. Current advancements in large language models (LLMs) suggest these AI systems can handle numerous effective responsibilities, from filling out forms to conducting last-minute research.
In a showcase last October, Anthropic unveiled one of its most advanced agents, an extension of its Claude model known as ‘computer use.’ This feature allows Claude to interact with a computer, mimicking human-like actions such as moving a cursor and inputting texts. Though still in the developmental stage, it is being tested by select partners, including DoorDash, Canva, and Asana.
To discuss future enhancements in agent capabilities, MIT Technology Review interviewed Anthropic’s cofounder and chief scientist, Jared Kaplan. Below are five expected advancements in AI agents by 2025.
1/ Agents Will Excel in Tool Utilization
Kaplan noted, ‘There’s a dichotomy in AI capabilities regarding task complexity and the various environments or tools the AI can access.’ He explained that past AI systems were limited to specific environments, making them less applicable. However, advancements like computer use show promise in expanding the range of agents.
2/ Improved Contextual Understanding
Kaplan emphasized that AI systems, like Claude, must learn the user’s context to enhance usefulness. He remarked, ‘Claude will soon be capable of sifting through your documents and communications to better understand your needs.’ This contextual awareness is crucial for effective interaction.
3/ Enhanced Coding Support
AI agents are also expected to improve coding assistance. Kaplan mentioned feedback from developers currently testing the computer use feature, stating, ‘As these systems evolve, they will significantly enhance the collaborative coding experience.’
4/ Addressing Safety Concerns
Reflecting on the rapid advancements in AI, Kaplan stated, ‘We founded Anthropic to tackle emergent safety issues as AI develops. Safety measures becoming increasingly necessary as AI agents integrate into workflow.’ Ensuring these systems operate without posing risks is crucial for future deployment.
In conclusion, as we approach 2025, the landscape for AI agents looks promising, potentially leading to enhanced efficiencies and capabilities in various sectors. How will businesses adapt to these innovative changes?
