Microsoft Seeing AI now free and released to Android

You know, I'm definitely on record as not being the worlds biggest fan of AI, but stuff like this is really cool, and a positive for the world.

I just wish it didn't have baggage.
 
Just downloaded to test the currency identification and scene description functions, and in cursory tests, it does what it is supposed to. I pointed it at a Lego box but without the Lego logo to see if it could tell the car on the box was fake, and it did identify it as a "toy car". It wrongly identified the bottom of the box as a table and kind of struggled with describing the background, but that's probably cause the backgrounds are usually abstract anyway.

Currency ID worked on the few bills I had on me, which is real good - a buddy of mine who is completely blind has told me that when he was younger, he had been swindled more than once by people saying "yeah, I gave you a 20 back" as they had him a 1. Prevents him from using cash anywhere where honest people don't exist.

I also tried the short text mode where it reads whatever you point your camera at. Pointed it at the DELL logo, and it thought the sideways E was a G....somewhat understandable. Pointed it at the windows timestamp in the bottom right and it correctly read it like someone would read a date and time. Tried to get it to read the article, but....it tries to read from left to right on the entire screen, so it tried to read one or two characters that were from an adjacent paragraph or in an ad to the left.

Actually good use of AI! Though, I do wonder, like Zarathustra said, how much telemetry is involved in this...
 
I can't wait to see Microsoft adding a cost to it once they think the userbase is big enough....
 
scottypippin said:
Just downloaded to test the currency identification and scene description functions, and in cursory tests, it does what it is supposed to. I pointed it at a Lego box but without the Lego logo to see if it could tell the car on the box was fake, and it did identify it as a "toy car". It wrongly identified the bottom of the box as a table and kind of struggled with describing the background, but that's probably cause the backgrounds are usually abstract anyway.

Currency ID worked on the few bills I had on me, which is real good - a buddy of mine who is completely blind has told me that when he was younger, he had been swindled more than once by people saying "yeah, I gave you a 20 back" as they had him a 1. Prevents him from using cash anywhere where honest people don't exist.

I also tried the short text mode where it reads whatever you point your camera at. Pointed it at the DELL logo, and it thought the sideways E was a G....somewhat understandable. Pointed it at the windows timestamp in the bottom right and it correctly read it like someone would read a date and time. Tried to get it to read the article, but....it tries to read from left to right on the entire screen, so it tried to read one or two characters that were from an adjacent paragraph or in an ad to the left.

Actually good use of AI! Though, I do wonder, like Zarathustra said, how much telemetry is involved in this...
Click to expand...
Lots the image is fed back to one of their datacenter’s. Which is further used to train more image recognition AI stuff.
So it’s basically community use training to make it better so more can use it.

Once it’s good enough they probably plan on a wearable version, maybe glasses, but for that they’d want a version that works offline and it likely isn’t there yet but with more training and a few more years of tech advancement who knows.
 
lRaphl said:
I can't wait to see Microsoft adding a cost to it once they think the userbase is big enough....
Click to expand...
Well it qualifies as a form of disability assistance so I imagine if Microsoft can get it to a good enough place the options for wearable hardware and medical research grants are huge. Then maybe get a subscription tie in for Teams, Zoom, etc for business use.

Microsoft also has one going for real time language translation and sign language interpretation.
 
You must log in or register to reply here.
Back
Top