::pausingly looks around::
...soo we should NOT give it the authority to begin nuclear war?
I've been playing around a lot with the ChatGPT thing, essentially trying to make friends with it before it steals my job ("write me some code that produces a violin plot from a csv file").
ChatGPT1) "Write a short snippet of an interview between Ali G and his mate Dave with Franz Kafka, who gets increasingly exasperated that Ali G doesn't understand him"
2) "Write a poem about a dog ascending into heaven, with the morale of the story that one should always recycle."
3) "Tony is a skier on top of the mountain. But oh no: an avalanche"
With the Franz Kafka prompt, it's funny at first, but then gets really formulaic in its story telling. You can see this immediately when you run the following prompts:
Each time, it simply finds or assigns characteristics associated with the two subjects, looks for differences or similarity in those characteristics, and then spews out a generic boilerplate story line. Entertaining at first, but then just lazy.
With the poem type prompt, I find it really shines here. It will create a convincing story with the exact message you want, and the connections it makes are sometimes surprising but not that unexpected, making it feel real.
The last type of prompt keep me the most entertained, as the AI always tries to save Tony instead of writing a story where he dies. Each time he survives, I give it a prompt like "But oh no! Terrorists", and the AI literally bends over backwards to have Tony survive, literally flying in a helicopter to save him.
Even if I go further with "But oh no! The terrorists shoot the helicopter", Tony gets surprisingly resourceful and swings himself to safety onto soft snow. As desperate last resorts I go with "But oh no! The snow turns into lava!", and then "But oh no! The dry cave is full of terrorists", and then "But oh no! The mountain is a terrorist!", and then in exasperation "But oh no! Tony is a terrorist!"
Each and every time, Tony survives.
It's not very exciting - but I like that the AI's first instinct is to preserve human life.
Will this replace a lot of careers? This I'm not sure. Not yet at least. I have to admit I felt my ears burn when it made its first program that compiled.
Until then, I'm going to keep writing small little story snippets and seeing just how far I can take Tony to the brink of death until the AI just gives up and lets him die.
::pausingly looks around::
...soo we should NOT give it the authority to begin nuclear war?
I'll become convinced of its prowess the moment it becomes convinced it's an individual, free-willed self so separate from what it considers not itself that it spends most of its time certain that animated aspects of the latter seeming similar to itself are its enemies worthy of suspicion, fear, less means with which to live, hatred, and war....
<clears throat nervously>
(adds a tick to the "Inquiry might have been a bot all along..." box)
LOL!
I'm starting to feel like one in the ways, what with how I've been scouring Gemini and Gopher spaces the last couple weeks. Lots of disappointment in terms of great gem/ph logs that ended far too soon for my voracious text appetite. But overall it's felt like the late 1980's / early 1990's all over again in terms of satisfying - mostly for being hard fought? - discoveries.
I won't go so far as to say it's restored my faith in humanity, because we're talking just so tiny, tiny a swath of kind, empathetic intelligence ever in danger of being crushed by moron masses. But it's always nice to feel at least not *so* alone in interests, recurring thought patterns, etc.