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 stage of development not only has far-reaching technological implications, but also immense strategic significance: it comes at a time of profound geopolitical shifts that have both an economic and a technological dimension. This is precisely where we come in with embraceableAI: With our visionary approach, we are opening up new scaling paths and aim to open up a new, European-oriented 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 must 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”. This further development not only opens up new application possibilities, but also offers the opportunity to define new rules of the game.
Leading players such as OpenAI are taking their first steps with large reasoning models (LRMs), where reasoning capabilities are integrated into large models through extensive 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.
The goal is similar, but the paths are different
Instead of relying on costly and resource-intensive training procedures, we increase the performance of the embraceableONE engine purely through targeted further developments in the area of software engineering. This innovative approach not only significantly improves resource efficiency – embraceableONE is around 10,000 times more efficient on the development side than the first large reasoning models – but also maximizes the reliability of the results.
At the same time, the system remains flexible as it uses standard market models and therefore operates both independently and very cost-efficiently during operation. This combination of high cognitive performance, reliability and efficiency has the potential to permanently change the rules of the game for further AI developments.
Strategic opportunity for Europe as a business location
The upcoming phase of AI development is not only technologically but also geopolitically significant. While Europe was barely competitive in phase 2 due to the high demand for capital and resources, phase 3 offers the opportunity for a strategic repositioning. Our approach 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 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 materials or drug research.
Into a new AI era with embraceableAI
While international players are investing billions in the further development of language models, we are taking a more efficient and smarter approach with embraceableAI. With the stand-alone reasoning engine embraceableOne, we are proving that cognitive performance, resource efficiency and reliability can be efficiently and effectively combined in a bionic approach. This innovative approach 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.