The editorial staff of Hi-Tech Mail.ru uses three popular neural networks I tried to evaluate The ripeness of the watermelons was examined from the photographs and then the taste of the selected samples was tested.
The experiment used ChatGPT 4.0, Claude 3.5, and Google Gemini. Each subject was shown a photograph of a stand with 18 watermelons and was asked to determine the ripeness of each watermelon.
ChatGPT chose watermelon number 5 and also highlighted numbers 2, 3, 9, and 12. Claude chose watermelon number 6 and highlighted numbers 3, 9, 11, and 12. Google Gemini asked for more detailed photos but noted that number 5 looked better than the others.
The Hi-Tech Mail editor bought watermelons 5 and 6 to evaluate. Watermelon 15 was also selected using traditional methods and the seller’s advice. All three watermelons were ripe and sweet, though not too sweet.
Watermelon number 5 had a thicker rind than the others, while watermelon number 15 had the thinnest rind. The flesh of all watermelons was bright red without any white lines, indicating the absence of nitrates. The seeds of all watermelons were black and evenly distributed throughout the pulp.
The experiment showed that neural networks can compete with traditional watermelon selection methods due to their high ability to analyze visual data. However, it is too early to fully trust AI and traditional verification methods are still valid.
Previously in Russia was created The fastest Rubik’s cube solving robot in the world.