Artificial intelligence is evolving at an unprecedented pace, making it difficult to predict trends in the technology. Liz Centoni, Executive Vice President and Chief Strategy Officer & GM, Applications, at Cisco, has taken on the task and summarized what she considers to be the five most important technology fields for the year 2024 in a recent blog post. The findings are based on the 'Cisco AI Readiness Index'. This is based on a double-blind survey of 8161 business and IT executives from the private sector in 30 countries in 2023. It was conducted by an independent third party who interviewed participants from companies with 500 or more employees. The index assesses companies' AI readiness in six areas: strategy, infrastructure, data management, governance, specialist staff and corporate culture. For Germany, 300 specialists were surveyed. The five most important trends are as follows.

APIs simplify the use of AI

Companies have a growing need to use data, automation and innovation. However, according to the survey, only 17% of German companies prioritize budgets for the introduction of AI over other technology investments. One solution will be the increased use of interfaces (API). Many AI tools and services will be integrated via this level of abstraction in the coming year. Such 'API abstractions' enable a more cost-effective integration of AI into business processes without having to delve deep into the technical details of the AI introduction or develop your own Large Language Models (LLM). By accessing various AI functions via APIs, repetitive tasks can be automated and decisions can be made on a better data basis.

APIs also enable the customized implementation of AI. Companies combine interfaces from different providers to create solutions for their individual requirements. This integration also supports collaboration with external AI experts, start-ups and research institutions. The first models of such curated AI ecosystems are already emerging - models that we will see more frequently in the coming year.

Call for alliances against AI-supported cyber attacks

The number of AI-supported cyberattacks will grow this year. Companies, politics, NGOs and civil society will be increasingly at risk from AI-generated disinformation. According to Cisco's 'Cybersecurity Readiness Index 2023', only 11% of German companies are resilient enough to withstand cyberattacks - and only 29% have a good understanding of the various cyber threats posed by AI. For this reason, technology companies and governments should work together to sharpen solutions against AI-powered threats such as deepfakes, AI social bots or cloned voice recordings and implement appropriate cybersecurity measures. Investments in risk detection and the training of AI models with large data sets will also increase. In order to detect threats at an early stage, companies must therefore invest in advanced security technologies and give higher priority to data protection.

Generative AI is fully established

To remain competitive, companies must implement AI within the next year. This is why the focus in 2024 will be on natural language interfaces (NLIs) for new products that are supported by GenAI. Half of the new products will have such interfaces integrated as standard. GenAI will also improve interactions in B2B business, offer interfaces and services for data access and be used in many business applications. According to Cisco, this primarily affects business tasks that analyse and visualize data, for example in project management, software quality assessment or the analysis of compliance fields as well as HR tasks.

It is also foreseeable that specialized AI models will come more into focus. This will lead to a shift towards smaller LLMs with greater accuracy, relevance, precision and efficiency. For example, LLaMA-7B models can be used for language tasks such as writing and completing code or classifying images with just a few shots ('few-shotting'). In addition, multimodality, which combines different data types such as images, text, speech and numerical data, will expand B2B use cases in areas such as business planning, medicine and financial services, where it will continuously deliver better results.

Improving energy efficiency in the use of AI

Smaller AI models tailored to specific use cases will already reduce energy costs when using AI in 2024 compared to generic systems. These special systems are trained on high-precision data sets and perform specific tasks much more efficiently. In contrast, deep learning models require large amounts of data to be processed in order to achieve results.

Furthermore, the rapidly growing application of energy networking will contribute to better energy efficiency. This refers to the combination of software-defined networking with DC microgrids. This will help companies to measure energy consumption and emissions more accurately in 2024. Many functions in IT and smart buildings can be automated with IoT sensors and made more efficient through integrated energy management capabilities.

Ethics and frameworks increasingly important

The introduction of AI is an unprecedented technological change that requires both innovation and trust. However, according to the 'Cisco AI Readiness Index', 76% of all companies worldwide lack comprehensive guidelines that regulate the use of AI. In view of the risks of GenAI, there is a broad consensus that such guidelines and voluntary self-regulation in the AI industry are generally necessary.

It must also be ensured that consumers retain access to and control over their data - in line with the current EU Data Regulation. This is a challenge for the companies themselves: With the growing importance of AI systems, publicly available data will soon no longer be sufficient for training AI models. High-quality language data is expected to be exhausted before 2026, meaning that a switch to private or synthetic data will soon be necessary. However, this entails the risk of unauthorized access and data protection violations. Those responsible for the use of AI will therefore commit to greater transparency and trust in relation to the development, use and results of AI systems.

Technology companies in particular will have to be prepared to demonstrate a new level of openness in the coming year - for example, which governance processes control the internal development, application and use of AI. If they are not able to credibly demonstrate a trustworthy approach to AI across the board, the regulatory framework will also be tightened in the coming year.

In her blog post, Centoni emphasizes: "Trust between people and the AI systems and tools they use is fundamental and non-negotiable. This means creating clarity about what AI can and cannot do through new frameworks for data transparency and accountability, new efforts to educate people and businesses about how disruption can happen, teaching the skills needed for new AI-enabled and -assisted workplaces, and new ways of working together that serve people's interests."