In 2024, the landscape of AI engineering has experienced significant maturation, highlighting evolving trends that define the integration of artificial intelligence into business practices. A recent analysis from The New Stack delineates five key AI engineering trends that have emerged and taken shape this year, underlining the requisite innovations and transformations within the sector.
One of the standout trends this year is the proliferation of agentic systems. These systems utilise large language models (LLMs) to perform various automated tasks, placing human operators at the core of their functionality. Notable presentations at the AI Engineer World’s Fair held in San Francisco in late June showcased the contributions of AI startups such as LangChain and LlamaIndex. LangChain introduced its LangGraph, crafted for creating AI agents, stressing the importance of maintaining human involvement in its operation. CEO Harrison Chase described LangGraph as “purpose-built for agents,” emphasising its controllable nature. In contrast, Jerry Liu of LlamaIndex conceptualised AI agents as an evolution of Retrieval-Augmented Generation (RAG), branding them as “knowledge assistants” designed for enterprise use.
AI coding tools have also risen to prominence, becoming common within the developer community. Surveys indicate that 76% of developers are either already using or intend to adopt AI tools, primarily driven by the need for increased efficiency. However, concerns linger regarding the quality of the code produced by these AI assistants, as only 23% of developers felt that AI enhanced the overall quality of their solutions. Prominent tools such as JetBrains’ integrated AI functionality are now essential components of the coding process. David Eastman, a contributor for The New Stack, stated, "To some degree, the ‘Hey! Look! We have AI’ is a current business necessity for IDEs as the environment expands and while agreed expectations are still forming." Additionally, new innovations, such as Solver—an AI tool allowing for autonomous task handling—were noted for their potential to significantly transform coding practices.
The profession of AI engineer has gained recognition and importance, evolving into a critical role amidst growing AI implementation across various organisations. As Fatih Nar and Roy Chua articulated, AI engineers are essential for the effective design, implementation, and scaling of AI solutions within enterprise software ecosystems, providing proficiency in managing data pipelines and ensuring data quality.
Emerging preferences for small models and locally hosted LLMs are also shaping the direction of AI tools for developers. Smaller language models (SLMs), such as the Gemini models launched by Google, have garnered interest due to their cost-effectiveness and ease of deployment. Director of Google DeepMind, Tris Warkentin, noted, “there are a wide variety of applications for these models”, highlighting their compatibility with many developers' project requirements. The capacity to run open-source LLMs locally further supports data security, allowing developers to manage sensitive information confidently.
Finally, the discourse surrounding open source AI has intensified. Concerns have been raised regarding the transparency and accessibility of AI models, particularly Meta’s LLaMA models, which some argue do not fully adhere to open-source principles. Mark Collier, chief operating officer at the Open Infrastructure Foundation, outlined that restrictions on usage and modifications conflict with traditional open-source definitions. In response to these issues, the Open Source Initiative has sought to clarify what constitutes open source in the AI context, with recommendations to share data and source code integral to training AI systems.
As 2024 unfolds, the integration of AI into business operations reflects a deepening reliance on sophisticated technologies, rendering roles like AI engineer pivotal. The year illustrates the ongoing evolution of development practices, primarily through innovations in automation, coding tools, smaller models, and the principles governing accessibility and usage of AI technologies.
Source: Noah Wire Services