Here there be dragons...

"I'm telling you stories. Trust me." - Winterson

I've been told ;-P

So TalkPal, the AI I’m using to practice speaking Spanish is, obviously, new tech - with all the challenges one might expect and then a few given the whole generative AI concept.

This one happened today and I’m still highly bemused by it, so thought I’d share.   I tried the “phone call” option (ie - no text shown since it’s mimicking being on a phone.

(Squirrel moment - skip this paragraph if you just want the actual story). Some notes on this one:

  • the visual display is very similar to what you’d see if someone were calling you on a generic smartphone.  It causes the same anxiety as when a real person does it ;).  I imagine if you actually enjoy your phone ringing, you will enjoy this too.  Super smart approach.
  • You can still translate what the AI has said and see the text version; you can also get her to repeat herself
  • You get feedback on what you’ve said if you click the feedback button but it wasn’t as obvious to me whether there was feedback there or not, and all comments from my original post still apply on the questionable quality of said feedback (validated, might I add, by my actual real-life Spanish teacher).
  • There doesn’t appear to be any way to “teach” the app - which seems fairly short-sighted since that’s kinda the whole point of generative AI (regardless of which approach, this doesn’t just apply in phone mode).
  • Conversation memory is still quite short.
AI Phone Screen

Now back to our regularly scheduled program.  She (AI’s name in this one was Emma) had asked me something generic about my plans for the day and I gave a pretty basic answer along the lines of relaxing by reading a fantasy novel and playing outside w my dog.

And I got this answer:

🤣 so, I mean, it does tell me my pronunciation was clearly off in a potentially fatal way somewhere because let’s be honest - that's not a challenging sentence to construct.  I *know* the words (at least the ones I was *trying* to say) were correct.  And completely innocuous.   And yet twice I got this error.   On the third time I managed to get it to understand and the conversation continued.   But because this version doesn’t script what you said, I couldn’t see what it *thought* I was saying in order to fix it.  And once you get that error, you can’t click through to where it’ll give you a translation ;)

So yeah - idk what that was about, and probably never will, but I’m still mildly amused.  If nothing else, if I can get it to understand me, people sb super easy!


Post a Comment