PART I It is my understanding that a bot works (like the fish bot example) by being able to quickly categorize information, be it words or pictures like in the example, based on the data it trained on. That is to say, the information the developer programmed into the bot. By inputting information to the bot, a student can better understand the way an algorithm is used in generative AI, specifically the large language models. The output is completely tied to the information it is fed. Like any programming, it can only give out what it receives in. This is why when using AI one must be sure to test and fact check information that it spews out as it can only be as correct as its sources. In fact, if the source material is bias in some way, obviously what is produced by the bot (or other AI tool) will also have that bias attached. In short, what is generated is contingent on the data on which is trained. This is important to note even when using any AI tool, l even those intended for educational purposes. Khanmigo is one of those tools. Based on the video it does seem helpful and frankly I would need much more time to see how or if this tool would be suited for the modern language classroom. That said, I do see the promise of so great learning that could develop from the use of such tools. I student could really develop their use of French based on their unique ideas and personality and use vocabulary, for example that really suits them and their need. It seems like we, world language teachers, are always discussing VOCABUALRY! Though initially seen as only an identification level in Bloom's taxonomy (yes I said it) I can see with AI, that the language level for some students, may be able to be more advanced, sooner. It could involve a much higher level vocabulary and grammatical syntax, earlier on in the development of proficiency on the ILR , CEFR or ACTFL language scales.
