In principle, one can distinguish 3 types of AI - academically not quite complete - which are not always sharply separated, but more or less overlapping:
DeepNews uses (c) statistically based methods for the most part, and (a) rule-based methods to a very small extent.
DeepNews 'is' a so-called 'argumentative AI'. This means that any decision made in DeepNews, such as that 'Text A has a semantic proximity of 83% to Text B,' can be understood and expounded in minute detail.
*On a personal note: We, the developers of DeepNews believe that only argumentative AI can be reconciled with the principles of an open and transparent society.
AI decisions and results based on neural networks, such as those of ChatGPT, are not comprehensible to humans or machines. Results and statements and facts stated therein are not justified, but delivered and are to be accepted- they are not verifiable.
In addition, neural network based methods often behave 'biased'. This always happens when not enough attention was paid to 'neutrality' in the underlying training sets, which is also difficult given the size of some training sets.
Many of the core functionalities (perspectives, extraction of topics, ...) that clearly set DeepNews apart from other news aggregators are not feasible without the use of our proprietary AI and NLP procedures - especially in terms of quality and scalability: