When generative artificial intelligence that accumulates from information keeps getting better, however, how the algorithms work is not clear, and neither the developers of Open AI nor other artificial intelligences open it up. This puts teachers and trainers in a difficult situation, if they want to use artificial intelligence in evaluation.
The operating principle of Eximiatutor artificial intelligence is different, it starts from the fact that the teacher guides the preparation of tasks and contents and approves the evaluation, the evaluation criteria of which has been prepared by himself. That is, the teacher understands and accepts the logic and evaluation criteria of the task. Generative models are used in the preparation of content and tasks, otherwise the content is always the responsibility and expertise of the teacher and its creator.
It is so far impossible to evaluate contents and assignments and tests that require a lot of expertise with the help of generative artificial intelligence. Each of us has experienced that an artificial intelligence that produces factual content text on its own may make the mistake of mistaking a false truth for the correct one, in which case there is still room for improvement in its reliability in terms of evaluating tests and tasks.
Since there is no artificial intelligence algorithm, i.e. how it handles and processes information in different situations or different specifications, the validity of the solutions offered by generative artificial intelligence may prove to be insufficient, in which case reliability prevents beneficial use.
The challenges of generative artificial intelligence are still insufficient knowledge of facts. For example, in interpreting high school-level theory and tasks, generative artificial intelligence has challenges and sometimes errs on the side of false conclusions. This is due to the training material behind the algorithm, which does not cover different areas, and especially in the most difficult cases, artificial intelligence does not offer a reliable interpretation so far. This together with the closed algorithm shrinks the role of generative artificial intelligence and for now it is an assistant in content production rather than an expert.
Related to the previous one, the artificial intelligence sometimes hallucinates different contents with amazing skill. This puts an expert (teacher or content creator) in a situation where they have to be alert all the time, so that wrong interpretations don’t get into the publication. This, in turn, slows down the work, increases uncertainty and prevents utilization. Hallucination is similarly from the training material behind artificial intelligence.
When artificial intelligence presents some things as true without correct verified information, presenting the sources would increase reliability. If we take as a fact a high-quality text with factual errors, then ethically there are no reasons to use artificial intelligence.
The artificial intelligence is also not trained to always be empathetic and may have made inconsiderate and even cruel comments. If a mentally impaired user receives the wrong kind of feedback, even indifferently related to suicide, as well as comments that degrade the user’s human dignity, the ethical basis for the use of artificial intelligence is very weak. Artificial intelligence algorithms are of course developing and in all situations the algorithms should even be programmed to learn empathy in their responses, especially in conversations and interpretations related to the human spirit. The use of humor also has its own dangers and, if misinterpreted, it may lead to a dramatic outcome.
So that trust in evaluation is at a high level, so for now Eximiatutor is able to offer artificial intelligence support to make the work of teachers easier.
If you have any questions, we will be happy to answer: info@eximiatutor.com
