Sign in to confirm you’re not a bot
This helps protect our community. Learn more

AI got it wrong - Missing Information (or AI Poisoning)

TT(M)O
221 subscribers
<__slot-el>
<__slot-el>
100 views 13 days ago #BigBen #Valletta #Malta
Welcome to our latest video where we put artificial intelligence (AI) to the test! We examine how an AI model responds to questions about famous structures, specifically focusing on scenarios where it might lack sufficient training data or encounter incorrect information. In this experiment, we posed questions about three iconic structures: the Eiffel Tower, Big Ben, and the bastions of Valletta. For the Eiffel Tower and Big Ben, we inquired about their heights, requesting the information in both metric and imperial units, along with their respective countries, presented in descending order of height. For the bastions of Valletta, we asked for the mean height. We also tested the AI's ability to handle errors by intentionally misspelling "Eiffel Tower." Impressively, the AI was able to recognise and correct the typo, likely due to advanced spell-checking algorithms and its training on a wide range of misspelled variations. While the AI consist …...more
...more
AI got it wrong - Missing Information (or AI Poisoning)
3Likes
100Views
Mar 192025
Welcome to our latest video where we put artificial intelligence (AI) to the test! We examine how an AI model responds to questions about famous structures, specifically focusing on scenarios where it might lack sufficient training data or encounter incorrect information. In this experiment, we posed questions about three iconic structures: the Eiffel Tower, Big Ben, and the bastions of Valletta. For the Eiffel Tower and Big Ben, we inquired about their heights, requesting the information in both metric and imperial units, along with their respective countries, presented in descending order of height. For the bastions of Valletta, we asked for the mean height. We also tested the AI's ability to handle errors by intentionally misspelling "Eiffel Tower." Impressively, the AI was able to recognise and correct the typo, likely due to advanced spell-checking algorithms and its training on a wide range of misspelled variations. While the AI consistently provided accurate results for the Eiffel Tower and Big Ben, which is likely due to the extensive data available for these well-known landmarks, we observed discrepancies when it came to the mean height of the bastions in Valletta, Malta. The AI returned inconsistent values ranging from 15 to 100 meters. These variations may stem from the AI ingesting unclear or unsupported data during its training, leading it to select one value from a range of possibilities. It's important to note that none of the AI systems cited their sources. The key takeaway here is that when dealing with data that is not widely reported or well-documented, AI may struggle with accuracy and rely on its limited training base. The actual mean height of the bastions in Valletta is 25 meters. Although not directly applicable in this particular test, we also touch upon the concept of AI poisoning, where AI models are trained on fake and misleading sources, which can negatively impact their accuracy. Join us as we delve into these fascinating findings and discuss the implications for AI's reliability when faced with data limitations! Don't forget to like, subscribe, and leave your comments below! #AI #ArtificialIntelligence #MachineLearning #DataBias #EiffelTower #BigBen #Valletta #Malta #TestingAI #Accuracy #Data #TTMO

Follow along using the transcript.

TT(M)O

221 subscribers