Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > alt.comp.os.windows-10 > #183382

Re: Scary AI

Show key headers only | View raw

Ed Cryer <ed@somewhere.in.the.uk> wrote:
> Newyana2 wrote:
>> On 4/8/2025 11:26 AM, Ed Cryer wrote:
>>> 
>>> I thought I'd have a laugh, so I asked ChatGPT for a sonnet in the 
>>> style of Shakespeare. What it gave me has caused me to gulp with 
>>> astonishment. It looks pretty good to me; maybe a little more modern 
>>> English than Elizabethan, but, still, it has a spark; as well as good 
>>> iambic pentameters and rhymes.
>>> 
>>> ******************
>>> 
>>> You said:
>>> write a sonnet in the style of shakespeare
>>> ChatGPT said:
>>> Certainly! Here's a Shakespearean-style sonnet for you:
>>> 
>>> When twilight falls and paints the skies with gold,
>>> The stars emerge, as whispers of the night;
>>> The moon, a silver orb, her secrets told,
>>> And soft winds sing of dreams within our sight.
>>> 
>>> Yet time, that thief, doth steal from youth's bright flame,
>>> And bids us age before the hour is due.
>>> No mortal heart can stay its endless claim,
>>> Though love may bloom and linger, ever true.
>>> 
>>> But in your eyes, I find the world anew,
>>> A spark that lights the path through shadowed days;
>>> A love that beats within my soul, so true,
>>> To guide me through life's ever-winding maze.
>>> 
>>> So let the years pass on, as they must go,
>>> For in your love, I find eternal glow.
>>> 
>>> I hope this resonates with you!
>> 
>>   It sounds like Mr. Hallmark had a few too many drinks.
>> 
>>   I'll be happy when the AI craze blows over, leaving
>> a light of blessed clarity, replacing fake intelligence's
>> winding maze, and let me just say eternal and true
>> love and all that good stuff.
>> 
> 
> How does the program do that?
> I used to earn my daily bread programming commercial mainframes. But 
> this AI "in the style of" beats me. I've done Walt Whitman, Louis 
> Carroll, William Wordsworth and others; and all win my praise.
> And that's only one facet of the AI phenomenon.
> 
> I'd have needed a vast database of examples set up just for Shakespeare 
> alone.
> 
> I'll not rest until I grab the working essentials of how this thing 
> works; call it a program, call it an algorithm, call it what the hell 
> you want, but I want to know how it produces its stuff.
> The sort of things I've read so far are; "It has all the Internet at its 
> call, and summons up vast resources". But that gives nothing away. If 
> you've ever had to write a computer program, you'll know that is true.
> 
> So, how does it do its stuff?

It learns through example in simplistic mimicry of a brain. You don't write
a whole bunch of if statements describing a cat as best you can whilst
avoiding misidentifying a fox. 

Instead, at the basic level, you provide a learning algorithm - which is
usually an existing library of statistical models- a huge catalogue of cat
pictures in different poses and contexts. You describe these pictures via a
complex set of feature descriptors using only the pixel data. When - or if
- you have the right dataset and featureset the algorithm will have
"learnt" the optimal combination to best describe a cat from a photo. This
is stored as a huge matrix of numbers which are uninterpretable by humans,
but then can be used to identify cats in new photos. 

Large language models (LLMs) are much more complex and, as you've found,
have learned some very detailed nuances of human language. i.e. know the
difference between Walt Whitmam and Shakespeare. 

The clever bit is identifying which statistical models work well and
programming the libraries to be as generic possible. 

Back to alt.comp.os.windows-10 | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Scary AI Ed Cryer <ed@somewhere.in.the.uk> - 2025-04-08 16:26 +0100
  Re: Scary AI "Alan K." <alan@invalid.com> - 2025-04-08 11:45 -0400
  Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-08 12:39 -0400
    Re: Scary AI Ed Cryer <ed@somewhere.in.the.uk> - 2025-04-08 19:07 +0100
      Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-08 14:45 -0400
        Re: Scary AI Ed Cryer <ed@somewhere.in.the.uk> - 2025-04-08 20:35 +0100
          Re: Scary AI jetjock <jetjock@unkown.com> - 2025-04-09 14:01 -0500
        Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-08 20:57 +0000
          Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-08 17:16 -0400
            Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-09 19:53 +0000
              Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-09 16:47 -0400
                Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-11 12:23 +0000
                Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-11 13:31 -0400
                Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-13 02:30 +0000
                Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-13 07:49 -0400
                Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-13 13:08 +0000
                Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-13 02:38 +0000
            Re: Scary AI Anton Shepelev <anton.txt@gmail.moc> - 2025-04-18 01:40 +0300
              Re: Scary AI Newyana2 <newyana@invalid.nospam> - 2025-04-17 20:22 -0400
                Re: Scary AI Borax Man <rotflol2@hotmail.com> - 2025-04-18 02:12 +0000
                Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-18 06:44 +0000
                Re: Scary AI Anton Shepelev <anton.txt@g{oogle}mail.com> - 2025-04-18 14:34 +0300
      Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-08 20:46 +0000
        Re: Scary AI Ed Cryer <ed@somewhere.in.the.uk> - 2025-04-08 22:48 +0100
          Re: Scary AI Chris <ithinkiam@gmail.com> - 2025-04-09 09:50 +0000
    Re: Scary AI occam <occam@nowhere.nix> - 2025-04-18 16:45 +0200
  Re: Scary AI Steve Hayes <hayesstw@telkomsa.net> - 2025-04-09 05:29 +0200
  Re: Scary AI Anton Shepelev <anton.txt@gmail.moc> - 2025-04-18 01:02 +0300

csiph-web