AI Trends 2024: Prediction and Reality

At the beginning of 2024, Datahouse highlighted the key trends in artificial intelligence for the coming year in a blog post. Three central developments were the focus: Retrieval-Augmented Generation (RAG), multimodal AI models, and the miniaturization of AI models for smartphones.

Now, as 2024 comes to an end, it’s time to take stock. Which of our predictions came true, which developments exceeded expectations, and what did we fail to anticipate?

Developments in the Three Areas

Retrieval-Augmented Generation (RAG)
RAG technology has indeed established itself as a key innovation to supply large language models with current and specific information. Companies like Microsoft and Google have integrated RAG into various applications, and the topic continues to attract significant interest in research and business by the end of 2024.

Multimodal AI Models
The progress in multimodal models has also lived up to expectations. Systems like GPT-4, Gemini, or Llama enable the seamless processing and combination of various data types—text, images, and audio—with increasing quality. Particularly noteworthy is that in 2024, two Nobel Prizes were awarded to scientists whose work is directly based on AI: the Nobel Prize in Physics honored groundbreaking contributions to artificial neural networks, while the Nobel Prize in Chemistry recognized work on the “AlphaFold” system, which revolutionized protein structure prediction. This strikingly demonstrates that AI shapes not only technical and economic but also scientific disciplines—truly a “multimodal” influence.

Miniaturization of AI Models for Smartphones
In 2024, AI models not only became smaller (requiring less memory) but also benefited from AI-optimized hardware, significantly enhancing the use of AI applications on mobile devices. Smartphones like the Samsung Galaxy S24 and new chips from Apple now enable offline functions such as translations and image editing powered by AI.

In summary, the three AI trends anticipated for 2024 have all proven to be important and continue to be influential. Additionally, there were other developments that, while not predicted in our blog post, have significantly shaped the dynamics of the AI field in 2024.

Other Key Developments in 2024

Regulation: Der EU AI Act
The EU AI Act, which also affects Swiss companies operating in the EU market, created the first comprehensive legal framework for AI. The legislation defines clear guidelines and risk categories for AI applications, impacting both companies and end users. While this brings greater transparency and safety, there is ongoing debate about whether the strict regulations could stifle innovation.

Konvergenz: Rückgang der Dominanz proprietärer Modelle
As analyzed in the renowned “State of AI Report 2024”, the performance of leading AI models is increasingly converging. This means that the capabilities of different models are becoming more similar, and the dominance of proprietary (non-open-source) models like OpenAI’s ChatGPT is diminishing. Although OpenAI regained the top spot with its “o1” model, this trend underscores the growing competition and the importance of open-source initiatives.

Hardware: Der „War for GPUs“
2024 was also the year of the GPU boom. GPUs, or Graphics Processing Units, are the backbone of modern AI, as they are essential for developing and deploying complex models. Nvidia (now the most valuable company in the world!) established itself as the market leader in GPU manufacturing while other giants like Meta and Microsoft also heavily invested in data centers and hardware.

The year 2024 has demonstrated just how dynamic and multifaceted the AI field remains. While some trends unfolded as expected, others were less predictable but proved to be just as significant, leaving us eager to see what 2025 will bring.

What Awaits Us in 2025?

In January, Datahouse will host an “AI lunch” (in German) to analyze and discuss the AI trends for 2025. Join us to stay at the forefront of this transformation.