Anyone Play With Google Dream? (AI)

[KG4CGC]

I've been messing with google dream on and off for years but they have introduced text to AI.
I've been trying to get the hang of it and some are pretty good. Some, meh, lackluster.

Anyway, here's a link and you can check out the stuff I uh, AI'd. There are several pages going back years.

https://deepdreamgenerator.com/u/112821/account?page=1

[KG4CGC]

aiqbu9_65bd1ff6053cd1f6cd0a76334f5b1f6893d46264.jpg

[KG4CGC]

m4srop_ba304ea2f561726a1468e3400d7a61303963319e.jpg

[KG4CGC]

rafewv_98520eca6468fd59a9dada722c379daae7db5e51.jpg

[KG4CGC]

In this one, I asked the AI if it dreamed about being human.


rondzq_424890ec1224a91599f5392484e80468ad68dfcd.jpg


It's gotten me into a whole series of asking the AI questions.
Imagine having consciousness but you can only see things in your mind based on everything that's on the web.
No other senses. No touch, smell, vision, breath, etc. If you gained sentience, would you begin to wonder why you weren't human?

In the pic above, the AI dreams of being human, surrounded by flowers and imbued in the scent and feel and all that is associated with that.

[koØm]

rafewv_98520eca6468fd59a9dada722c379daae7db5e51.jpg

Ask your AI to picture itself 420'ed

.

[KG4CGC]

vbrppl_b1eee0f8a905368432771db663aebb269352cf13.jpg

[n6hcm]

i haven't (not very artsy) but i'm unsure i'd use it ... i only teach a little bit now, and the ChatGPT answers to questions that I've seen have been ... informative.

[KG4CGC]

i haven't (not very artsy) but i'm unsure i'd use it ... i only teach a little bit now, and the ChatGPT answers to questions that I've seen have been ... informative.

I'd like to hear more about that.

[n6hcm]

A colleague found that ChatGPT could write a brief exposition on a direct question that would be indistinguishable from something written by a student. (An undergrad, to be sure, but ...)

Most of what I see *looks* good but sometimes the underpinnings are shaky. Anything with citations is risky (the citations look good, but the citations don't match the idea they're cited with ... or, in some cases, the citations don't actually exist).

I suspect that a longer document might expose more pain points.

There are services that will examine a text in order to discern whether that text was produced by an AI process or not ... but those services use an older version of the version that is now generating these texts, so it's unclear how successful these services are.

And then, of course, ChatGPT type services aren't particularly interested in their own plagiarism scores, so they may end up using texts that closely match what is available in scanners like TurnItIn ...

Academia is going to have a very interesting few years.