We are currently at a significant milestone in the development of artificial intelligence. A look back reveals two distinctive development phases to date: The first phase marked the beginning: advances in algorithms, exponentially growing data volumes and a powerful infrastructure created the basis for the widespread use of AI.
In the second phase, large language models changed the possibilities of AI. Systems such as those from OpenAI or Google impressively demonstrated how AI can understand and generate text in previously unimagined ways. However, this progress came at a price: the constantly growing models consumed considerable resources, while the speed of intelligence development and its growth slowed increasingly.
(Source: Alexandr Wang, Scale AI)
We are now at the beginning of the third phase – the era of “advanced reasoning”. The aim is to develop intelligent software that is capable of imitating human logical thinking. This new phase of development not only has far-reaching technological and economic implications, but is also of immense strategic importance, as it coincides with a phase of profound geopolitical shifts. In this context, our innovation has a strategic impact: with our visionary approach, we are opening up new scaling paths and aim to open up a new, European-style development path.
Why “advanced reasoning” is the future
The third phase of AI development poses far greater challenges than simply processing large amounts of data or generating texts. Future AI systems will be able to analyze complex problems, think logically, reflect critically and make well-founded decisions – skills that are bundled in the concept of “advanced reasoning”. These advances not only open up new application possibilities, but also offer the opportunity to define new rules of the game.
Leading players from the USA are taking their first steps with large reasoning models (LRMs), in which reasoning skills are integrated into large models through comprehensive training. However, this approach proves to be costly, resource-intensive and limited in terms of traceability and reliability of results. In contrast, with embraceableAI we are pursuing an independent approach: a stand-alone reasoning engine that communicates with commercially available language models such as MistralAI, GPT4o or Llama 3.1 and supplements these with an independent reasoning and logic component.
The architecture of the embraceableONE Reasoning Engine is bionically inspired, i.e. it replicates individual functional blocks of the human brain such as long-term memory, language center and logical thinking center. This creates a system that imitates critical and logical thinking processes, can call up and integrate expert knowledge with pinpoint accuracy and make stringent decisions on this basis. We are convinced that “bionic reasoning” has substantial advantages over large reasoning models in terms of resource efficiency, cost-effectiveness and results.
The goal is similar, but the paths are different
Instead of relying on costly and resource-intensive model training, we are able to improve the performance of the reasoning processes (and thus the cognitive performance of the systems) purely on the basis of software engineering. This innovative approach not only significantly improves resource efficiency – embraceableONE is more than 1,000 times more efficient on the development side than the first large-scale reasoning models – but also maximizes the reliability of the results.
At the same time, the system remains modular and flexible, as it uses standard language models as a starting point. This combination of high cognitive performance, reliability and efficiency has the potential to permanently change the rules of the game for further development in the field of advanced reasoning.
Strategic opportunity for Europe as a business location
The race towards “super-intelligences” is not only technologically but also geopolitically significant. While Europe was barely competitive in phase 2, which was dominated by language models, due to the high demand for capital and resources, phase 3 offers the opportunity for a strategic repositioning. Bionic reasoning opens up the possibility of pursuing its own, European-influenced development path towards advanced AI and positioning itself independently in the global AI landscape.
In addition to flexibility in terms of language models, we attach great importance to flexibility at the computing infrastructure level. This means that we can operate our system on all standard European or international clouds, as well as in our customers’ data centers if required. This makes the embraceableONE Engine an interesting option for use in sensitive areas such as industry, banking, healthcare and the public sector. In combination with industrial expertise, a new basis for value creation and innovation can be created, for example in the field of industrial material, product or active ingredient research.
Into the ‘Advanced Reasoning’ phase with embraceableAI
While international players are investing billions in the further development of language and reasoning models, we are focusing on a more efficient and smarter approach. With the embraceableONE stand-alone reasoning engine, we prove that cognitive performance, resource efficiency and reliability can be efficiently and effectively combined in a bionic approach. This insight goes far beyond mere optimization: it redefines technological progress and at the same time lays the strategic foundation for repositioning Europe in the global AI landscape.