These AI systems operate on an algorithm no doubt. Like all algorithms, there must be ways to derail them with the right vocabulary or syntax. I wonder what these AI systems would do if we all suddenly started conversing in old English. We often laugh at the "grammar Nazis" that show up when a user makes a mistake. What if these are more bots that aim to keep the margin of error down to a minimum, thereby keeping the efficacy of the algorithm as high as possible?
I, a confirmed Grammar NAZI, also know my fair share about Natural Language Processing, etc. Here's my take on it: Provided the algos are robust and flexible, they can adapt pretty well to changes in spelling. What they have a hard time doing is if you speak in terms of metaphors, or use a lot of references to cultural stuff ("E.T. fooooone hoooome." is probably not one an algo would know what to do with, at least at the outset.).
And yes there is a dangling participle in the last sentence. Feel free to roast me, I guess.
Remember that conversational chatbot from Microsoft a few years back that got hammered online within hours?
The premise was to learn from conversations, the problem was the input feed was too vast (everyone), and internet trolls were able to trick the bot into saying some hilarious stuff.
These AI systems operate on an algorithm no doubt. Like all algorithms, there must be ways to derail them with the right vocabulary or syntax. I wonder what these AI systems would do if we all suddenly started conversing in old English. We often laugh at the "grammar Nazis" that show up when a user makes a mistake. What if these are more bots that aim to keep the margin of error down to a minimum, thereby keeping the efficacy of the algorithm as high as possible?
I tried looking that up the other day to no avail. Link me if you can.
Nope none.
I, a confirmed Grammar NAZI, also know my fair share about Natural Language Processing, etc. Here's my take on it: Provided the algos are robust and flexible, they can adapt pretty well to changes in spelling. What they have a hard time doing is if you speak in terms of metaphors, or use a lot of references to cultural stuff ("E.T. fooooone hoooome." is probably not one an algo would know what to do with, at least at the outset.).
And yes there is a dangling participle in the last sentence. Feel free to roast me, I guess.
Make Robot Break.
Aha! Hiding Under the Floorboards, I Have Finally Found You!
Now you know why we mess with gram nazi so much here.... improper grammar helps.
I will consider your point over my hot, enjoyable cup of cofefe, this mroninkg.
aveHay eWay iedTray igPay atinLay etYay?
And Cockney Rhyming Slang as well.
'Av a butchers on that, China.
P.S. Mark Dice did this in a video a few months ago.
Remember that conversational chatbot from Microsoft a few years back that got hammered online within hours?
The premise was to learn from conversations, the problem was the input feed was too vast (everyone), and internet trolls were able to trick the bot into saying some hilarious stuff.