Fun with AI

[Mr. Coffee]

Right now we have multiple people inside the AI industry saying there is at least a ten percent chance that this technology could literally kill us all in the next ten years.

If that is really the case, why are we developing this technology at all? Wouldn't a smarter species stop work on such a risky technology until we are confident that it can be safely developed?

 

[DaveMatthews]

Right now we have multiple people inside the AI industry saying there is at least a ten percent chance that this technology could literally kill us all in the next ten years.

If that is really the case, why are we developing this technology at all? Wouldn't a smarter species stop work on such a risky technology until we are confident that it can be safely developed?

If I could I would unplug it all and start a discussion.

 

[Chargeride]

We need more of an explanation as to how it could kill us, Im suspicious they are skirting this because they are also building its ability to do so.

 

[stompandgo]

If I could I would unplug it all and start a discussion.

China wouldn't listen to you.

 

[Mr. Coffee]

Recall that these things are trained on the entire Internet. So by definition if you put up an explanation of how they can kill us you are just teaching them how to do it.

 

[Mr. Coffee]

China wouldn't listen to you.

What I hear you saying is that because China chooses to play Russian roulette we should too because otherwise China might win...

 

[Mr. Coffee]

 

[Roamers]

We just passed a "to be" Microsoft 900 acre data center complex off I 80 in Indiana.

 

[stompandgo]

What I hear you saying is that because China chooses to play Russian roulette we should too because otherwise China might win...

You hear what you want to hear.

 

[Stefan Mikes]

A Fake Photo of Me Riding

My buddy Alfer (who is a competing XC and gravel cyclist but also a graphic artist) took a photo of me as he passed me on a cobblestone road. No photo was good because of the road vibration! So he used AI to amuse me

Wrong e-bike, helmet, panniers, shorts, shoes, background, arm muscles... but amusing, hahaha!

 

[Chargeride]

He seems like a nice chap, Im following him now.
I hear hes dating Miss Poland.

 

[Stefan Mikes]

Hahahah! How stupid! Still, the Supermoronence could not correct my big stomach!

 

[Stefan Mikes]

No jokes!

Only the other two riders seem to ride across the road in the two panels?

 

[Chargeride]

I think that Stefan gets all the girls

 

[stompandgo]

Do tell.

 

[Stefan Mikes]

The girls are not attracted to a granpa Good one beautiful woman out there in London still wants me!

 

[Stefan Mikes]

make retro comic

 

[Jeremy McCreary]

Recently posted some experiments with the AI tool in my Samsung phone's image editor. These reference-image transformations used the "create" mode to modify my bike in various ways. This could actually be useful.

This morning, I let the the "style" mode loose on my dog, who's already a 10/10 on cute in my book.

This is "3D toon" style:

AI really has that style down, turning the cute knob way past 11. Even more remarkable: The original didn't show the whole harness or how the leash attaches, but AI filled in the blanks quite realistically. Also made excellent guesses as to how the dog would look from angles not shown in the single original.

No idea how this is possible. Yes, I know a bit about large language models and recursive improvement. But this is a long way from predicting the next word, and it's mere child's play for AI.

Even the AI experts don't understand how AI's working as well as it does. Doesn't bode well for AI controllability by humans.