What's Next in Generative AI? A Stocktaking

“Those who stop getting better have stopped being good.”

This quote by Philip Rosenthal hits the nail on the head when it comes to the rapid development of generative AI. In a world that is constantly changing, it is therefore more important than ever to stay on the ball. “So Arno?” I am then asked, “What’s next?” Let’s take a look at current trends and developments that, in my very personal opinion, no one should miss in the near future.

The Future Is Multimodal

One of the most exciting developments in the field of generative AI is the rise of multimodal systems. These AI models are no longer limited to a single type of data, but can process and generate text, images, videos, and audio data equally.

Imagine describing a scene to an AI system and it not only creates an image from it, but also matching music or even a short video. This technology opens up completely new possibilities for creative processes and human-machine interaction. Particularly exciting: open-source models like LLaVA could soon make this technology accessible to a wider audience.

Small Models, Big Impact

While in the past the motto “bigger is better” often applied, 2024 showed a clear trend towards smaller, more efficient language models. These so-called Small Language Models (SLMs) are tailored to specific tasks and can deliver impressive results despite their smaller size.

The market for small language models was estimated at USD 7.76 billion in 2023 and is expected to grow further at a compound annual growth rate of 15.6% from 2024 to 2030.

SLMs require less computing power and storage space, making them more cost-effective and environmentally friendly. They can also often deliver more accurate and relevant results than their larger counterparts, as they are trained on higher-quality, specialized datasets. Models like Mistral 7B and Microsoft’s PHI-2 already show the potential of this technology.

AI Becomes More Independent

Another fascinating trend is autonomous AI agents. These AI systems can independently complete tasks, make decisions, and produce content - without human intervention.

Imagine having a virtual assistant that not only responds to your instructions but proactively solves problems and makes suggestions. This development could completely redefine the way we interact with AI and perhaps lead to more efficient workflows in many industries in the long to medium term.

Open Source Is Catching Up

A particularly exciting trend is the increasing performance of open-source AI models. The gap to proprietary solutions like ChatGPT is getting smaller and smaller.

This development is significant for several reasons: On the one hand, it enables more companies and developers to access advanced AI technology. On the other hand, it promotes innovation and transparency in AI development. With models like GPT-Neo for content generation, powerful tools are available that could democratize the AI landscape.

AI Becomes Industry-Specific

The future of generative AI could also lie in tailored solutions for specific industries and use cases. Whether in healthcare, finance, or retail - AI applications are emerging everywhere that are precisely tailored to the needs and challenges of the respective sector. Deutsche Telekom is also working together with Singtel, SK Telecom, and other providers within the framework of a “Global Telco AI Alliance” on its own LLM, fine-tuned to the needs of the telco industry worldwide.

AI specialization not only promises more efficient processes but also innovative approaches to industry-specific problems. From accelerating drug development to personalized financial advice to faster processing of fiber optic applications for cities and municipalities - the possibilities are almost limitless.

Ethics and Responsibility Come Into Focus

With the increasing spread and performance of AI systems, ethical questions are also gaining importance. Topics such as data protection, fairness, and transparency are at the top of the agenda. The AI EU Act plays an important role, at least for research and business in Europe.

It is more important than ever that we as a society develop a responsible approach to AI technologies. This includes clear guidelines for the use of AI, but also mechanisms for reviewing and controlling AI-generated content. Sometimes, however, no law helps, but rather the people who take the issue seriously and then consistently implement it.

A Few Ideas on This

1. Experiment with different AI tools to learn their strengths and weaknesses.
2. Stay critical and question AI-generated content - not everything that glitters is gold.
3. Stay regularly informed about the latest developments, e.g., through specialist blogs or podcasts.
4. Think about ethical implications and discuss them in your environment.
5. Combine AI tools with human creativity for the best results.

Conclusion

The development of generative AI continues to advance at a dizzying pace. From multimodal systems to efficient small models to autonomous agents - the possibilities seem limitless.

But this is exactly what is increasingly becoming a problem.

How do you know which new trend, which product, or which service is now the right choice? Who in the company has the decision-making competence and authority to provide guidance here? Where do we locate our AI experts? Does it make sense to hang AI knowledge decentrally in the company? At the same time, we face the challenge of using these technologies responsibly and ethically and reconciling short-term (business) goals with long-term goals. Everyone needs to be involved!

As futurist Alvin Toffler once said:

“The illiterates of the 21st century will not be those who cannot read and write, but those who cannot learn, unlearn, and relearn.”

In this spirit, I invite you to keep asking questions and thus curiously but also critically accompany the developments in the field of generative AI. Let’s shape the future together - a future in which AI supports us, inspires us, and opens up new horizons.