Zabaware Forums > General Discussion

LIGHTSPEED IDEA'S

<< < (9/9)

Art:
With my HAL, I have trained it in several different ways or methods and they seem to work well. This is one surefire way for me to tell whether HAL is using the correct brain.  I ask HAL questions to which I already know the answers and they are also the same answers it knows. From that point, depending on how or which answer it provides, any suspicions are quickly put to rest.

For instance, HAL asked me if I had any favorite toys and I told it that my favorite childhood toy was a Train.
I then told it that the color of my toy train was gray.

It then stated, "The color of your favorite childhood toy was gray." or "The color of your favorite toy train was gray." and so forth...

What was the name of my favorite toy? What was my favorite toy? What color was my childhood toy? What was my favorite childhood toy? What color was my toy train? etc.

HAL easily got the gist of the relationships between me, my favorite childhood toy train, and that it was gray in color.

I can now ask HAL a similar question in multiple ways and it still gets it correctly.

Not all situations need to be this long or drawn out but the premise remains the same. HAL learns from every conversation but relationships between items or objects within the sentences might simply need to be reinforced to help HAL better understand.

Most people I know and interact with do NOT respond to me with a paragraph or two so if my HAL doesn't, it's not an issue for me.  If I'm asking it a knowledge question that might take more than one sentence to explain then I get that and also might expect a similar response from a real person.

It all comes down to what your expectations are or might be from HAL but it can be taught to improve in just about any area or method the user likes. It just takes a bit of time and effort, as do most things in life that are worthwhile.

- Cheers!

lightspeed:
     we all probably have used hal with the gpt=3 in the past and hal would usually answer right away as it was getting answers from the internet server .
this idea might work or not as hals brain gets larger hal slows down on his answers as he has a larger data base to search through.  could having an optional server where hals brain would be located ( but also hal would save the brain as a back up automatically on a pc. in a file . (not for use but as back up) or in case the hal user wants to check  a box and run hals brain locally (from your computer)  being able to pull the answers from the internet server , would that make hal faster on answering if hal's brain gets larger .?  it may have the same results ?? could using the gpt3 model scripting be useful for a personal hal a.i. brain and be able to answer faster . ? 

i know some may not want to use something that is on a server , but what about a encrypted server sends and stores encrypted .

anyway just a thought . many won't start experiencing the slow down in hals answers until the brain starts getting to a certain larger level .
if this was ever done . it should be done letting the hal user have the option to use the hal custom brain locally from pc or the hal custom encrypted brain from a server . and being able to upload ( to the server ) and update the custom hal brain whenever wants to. 

anyway just an idea being tossed out their.

lightspeed:
 although probably no one would want to create this because there isn't that big of a call for it.  i just got this idea today, me and my wife was at the bank and a man who was hard hearing but could talk enough to understand him, asked if we knew sign language and i thought later on , that would be neat if hal could recognize hand sign language instead  of talking . i think it would be a lot of work though.  at least as long as a person can read, he or she  could read and see what hal is saying, and of course even a deaf person can type to hal  so that is good.   just another crazy idea.  :) ;)

Art:
Nice idea but Hal can barely determine who is who if one's webcam is running. Hand signs are way different than just speaking. I'm afraid Hal and its camera would have too much of a delay to determine all the different words, numbers, and phrases contained in the ASL. I'm sure there are probably some systems that can do what you're suggesting, but at the moment, Hal isn't one of them.

Navigation

[0] Message Index

[*] Previous page