Large language models like GPT-4 offer numerous opportunities that are currently being economically explored. However, it is crucial to address the associated challenges to maximize the benefits of these AI drivers sustainably. This is emphasized by the Information Technology Society (ITG) in the Electrical Engineering and IT Association VDE in a recently published position paper.
Generative AI with assistants such as ChatGPT from OpenAI, Gemini from Google, or Claude from Anthropic is not just about developing algorithms and software for groundbreaking new applications. According to the ITG analysis, advancements in energy-efficient digital hardware and its integration with suitable software play a significant role.
Globally, there is talk of building more nuclear power plants to meet energy demands. Damian Dudek, ITG Managing Director and co-author of the paper, explains that the focus should be on making data processing more efficient and reducing energy consumption to use AI sustainably. If Germany adopts such approaches, it could unlock new market potentials and maintain technological sovereignty for the benefit of society.
Training large language models (LLMs) and deploying them on a large scale requires enormous computing resources, which are currently provided by dedicated multi-core systems. These systems rely on graphic processors with specialized software environments to enable highly parallel processing with high throughput and memory bandwidth.
There are several new approaches to overcome the limitations of traditional chip design methods. These include multi-level data storage in a single device, mixed-signal data processing circuits based on advanced CMOS technology, and the integration of electro-optical signal processing devices in system-on-chip architectures (SoC).
Experts see promise in neuromorphic computing approaches, inspired by biological systems, and quantum computing on various hardware platforms. Germany should remain active in these fields. Recently, ITG also highlighted photonics to reduce energy consumption, especially in data centers. The German startup Q.ant introduced an energy-efficient photonic AI chip in November.
Overall, the research community in Germany should address the challenges associated with AI with a focus on societal benefits and foresight for national and European economic development and technological sovereignty.