Intelligent man is making rapid progress in many areas up until that point, as he is also the ability of the human in the field, such as medicine. However, when it comes to dealing with the issues of the simple and logical, but in these systems, the sophisticated, yet it seems to fail so miserably.
The paradox of Alice in Wonderland
In a study carried out by the organization LAION tested the effectiveness of many of the language, such as the GPT-3, GPT-4 & GPT-4o of OpenAI, Claude 3rd Opus of the Anthropic, Gemini and Google's Called the Metës, and the Mistral and Mistral. The protocol of the test was very simple: the answer to the problem of the so-called “Alice in Wonderland”.
The question at issue was, “She is [X] my brethren, and also [His] sister. How many sisters has he Alisë?“. In spite of the simplicity of its notes, almost all of the models tested failed to give a correct answer, thus demonstrating the gaps of a sudden, in the capacity of their reasoning, logical.
It came as a surprise to many scholars, it was not only the inability of the model to solve the puzzle, but also the trust of over a show on their answers to be incorrect. Some of the models, artificial intelligence, as well as Called for 3 of the Metës, they have given an explanation of the details, but it's absurd to justify the choice of the spot, and making them out to be unreliable.
Need new tests to assess the capabilities of IT
These results are in stark contrast to the findings of the good to be obtained by the same model on tests such as the MMLU (Multi-task Language Understanding), which evaluates the ability of an intelligence man to solve the problems. This has led to researchers to also emphasize the need to review the use metrics to measure the ability of the current system of artificial intelligence.
A word of warning to trust too much to IT
Even though the intelligent man is making great strides in many areas, and this study serves as a warning about the importance of not trusting too much in the ability of it to date. Despite the progress of technology, the issues, the basic logic can be still an obstacle for these systems, highlighting the need for further improvement and an approach to the balance of the application of IT.
Discussion about this post