AI offers tremendous potential for companies to boost efficiency and create innovative business models. However, it’s crucial that we guide AI development responsibly. The challenge lies in how to achieve this. Let’s approach this step by step.
What is AI actually?
AI can be likened to a learning program that becomes increasingly intelligent as it’s fed vast amounts of data. It’s akin to a child who learns by observing their surroundings and engaging in play.
AI breakthrough through the power of language
A major breakthrough was the development of large language models such as ChatGPT. These can understand, summarise and even create texts themselves. This made AI suitable for everyday use and also interesting for areas such as law or marketing.
Why large language models like ChatGPT are not the solution
While large language models possess impressive capabilities, they are also highly complex and error-prone. They can generate „hallucinations“ — plausible-sounding but false information. Moreover, these models are challenging to control and often reflect the interests of major tech corporations.
Small, modular models are the future
A promising alternative is smaller, specialised models that are networked together. These are more flexible, easier to control and can be better customised to the specific needs of companies.
The importance of control
A significant challenge in current AI development is the lack of oversight. Who determines which data sets are utilized, what objectives are pursued, and how the quality of outcomes is verified? It’s crucial for companies and users to have more influence in shaping the development and application of AI technologies.
AI and its risks
Risks emerge when AI is deployed in inappropriate domains or pushed beyond its capabilities. The military sector serves as a prime example, where AI might be considered for rapid battlefield decisions, despite the uncertain consequences of such applications.
The role of universities
Universities play a crucial role in the development of new AI technologies. By working on open and flexible systems that enable organisations to create AI solutions that meet their specific needs, they can make a valuable contribution towards a positive direction of travel.
How users can take responsibility
Users can actively shape AI development by voicing concerns and demanding input. This is vital for those directly involved, like scientists and universities. They need their own AI systems to represent their interests and engage with centralized AIs. Tools to influence AI data, goals, and quality are also crucial.
Currently, a handful of companies provide the foundational models, leading to a concentration of power. A shift towards decentralized AI systems with distributed control is necessary. While this process may be time-consuming, it’s vital to comprehend and address the impact of monolithic software systems.
AI and its potential
AI’s potential is vast, with applications spanning search engines, healthcare, and office productivity. For years, AI has powered search and recommendation systems, with Google leveraging it to serve billions. Language models are revolutionizing document-heavy fields like healthcare, automating tasks through language analysis. In office settings, AI aims to boost efficiency in tasks like email writing.
The opportunities for Europe
While Europe currently lags behind the USA and China in AI development and implementation, it possesses the potential to become a leader in this field. The key lies in prioritizing investment in research and development, while fostering an environment conducive to the growth of innovative companies.
About EuroCC Austria
EuroCC Austria is the national competence center for supercomputing, big data, and artificial intelligence, founded in 2020 as part of the EuroCC initiative. This international project supports researchers, startups, and SMEs by providing expertise, networks, and access to HPC infrastructure (supercomputers). The Austrian competence center is operated by Advanced Computing Austria GmbH (ACA), the University of Vienna, the Vienna University of Technology, and the University of Natural Resources and Life Sciences Vienna (as part of the Vienna Scientific Cluster consortium)—in close cooperation with the business incubator INiTS.
This article is written by Natascha Trzepizur based on Bettina Bensch interviewing Endri Deliu.
Find the full interview (in German) here: https://eurocc-austria.at/blog/llms-in-ein-paar-jahren-endri-deliu
