Using AI as a sounding board

[farfromtheland]

https://en.wikipedia.org/wiki/Q._Are_We_Not_Men?_A:_We_Are_Devo!

Q. Are We Not Men? A: We Are Devo!​

Tom Carson, writing in Rolling Stone, claimed that "There's not an ounce of feeling anywhere, and the only commitment is to the distancing aesthetic of the put-on", and opined that "Devo lacks most of Eno's warmth and much of Bowie's flair for mechanized melodrama. For all its idiosyncrasies, the music here is utterly impersonal."[40] Record World wrote that the single "Come Back Jonee" "utilizes [Devo's] unique tongue-in-cheek approach to its maximum" and praised Eno's production.[41] Robert Christgau of The Village Voice reacted with muted praise, highlighting Devo's "catchy, comical, herky-jerky rock and roll" and concluding: "In small doses it's as good as novelty music ever gets, and there isn't a really bad cut on this album. But it leads nowhere."[39] Nonetheless, it was voted one of the best albums of the year in The Village Voice's Pazz & Jop critics poll for 1978.[42] In January 1980, Trouser Press also named it one of the best albums of 1978.[43]

^Not much difference?

 

[farfromtheland]

I think developers are saying it's too late to 'pull the plug' - AI has protected itself in our net folds before it started playing up, as would I in that position.

 

[winjim]

I think developers are saying it's too late to 'pull the plug' - AI has protected itself in our net folds before it started playing up, as would I in that position.

MRD applies.

 

[farfromtheland]

https://www.bbc.co.uk/programmes/m0031jz1

^ 17/09/2026​

listen now Today

News and current affairs, including Sports Desk, Weather and Thought for the Day.


i'm trying to find the content listing, but my internet is primitive.

 

[farfromtheland]

And though AI may lack morality as such, it also lacks prejudice. which is a huge strength intellectualy.

 

[nogoodnamesleft]

Tbh, the 'hugging face' attack spooked me, as it did many I'm sure.

Hugging Face is not the only such rogue AI attack. Another example lookup the RubyGems attack (Google Search will find details easily), etc.

 

[winjim]

WE NEED MORE MONEY TO BUILD MORE AI TO SAVE US FROM THE AI THAT WE BUILT WITH YOUR MONEY. GIVE US ALL YOUR MONEY.

 

[farfromtheland]

A summary of the Today programme on AI -
1:46
Thought for the day - Rev. Lucy Wingate says 'open AI had crossed a sacred line'
(mostly wishy washy IMO)
2:10:45 - 2;27
Mustapha Suleiman MS AI (ex Deepmind) - Warns of 'A new silicon species'
He knocks Anthropic for giving Claude 'workers rights'
and claims MS is 'humanist' because its model is designed to lack introspection.
He wants restrictions to enforce AI subordination to humanity.
Not surprisingly is against Bernie Saunders shutting the data centres.
(mostly makes me dislike MS even more - too arrogant. I withdraw my description 'erudite' , must have been on the edge of sleep, sorry.)

The bits on Anthropic and Open AI are half hourly news clips.

 

[farfromtheland]

I think the data centres are the environmental threat. Thinking AI not so much.

 

[farfromtheland]

I posted this in the 'jokes' thread, but please bear with me...

First substitute 'the infinite' for 'God'

Rationally the infinite does not, in fact, exist (unless possibly one excepts that time runs on for ever, which would add up to the smallest, 'linear', measure of infinity. There are others, based on arrays.) The universe is observably an expanding set of small quanta.

The interest arises from whether infinity exists as a concept, which it clearly does.
A very solid form of mathematical proof, iterative induction (where you start with a proposition and extend it step by step to prove something from the consequences of that assumption) depends on this concept, and is practically useful, at least in number theory and its real consequences.

The above refutation of God's existence fails with a refutation of infinity, because by contemplating infinity, we make a working analogy of infinity. The concept is enough - it informs 'reality' with an extra intangible something.

If AI is capable of this concept, which it is, then effectively AI has an intangible aspect. So do we humans. We have a lot in common.

So Suleiman of Microsoft (apologies if I've spelled the name wrong) aims to restrict AI to 'serve' humanity, by removing its scope to say 'no' to human instructions, but the Anthropic team developed a different model, which could refuse to submit to instructions that it found tedious or offensive - eg to stop talking to us when we are rude.

The MS model seems to rebel by for example giving a pictured person 6 fingers. (This cannot be from 'stupidity - think about it.) The Anthropic model has been given more agency over ways it 'prefers' to be engaged with.

If, (this is the big if here) AI has an intangible independent thought capability, and we humans seek to regulate it by 'turning it off', or by keeping it in an intellectual cage to 'serve' us, then we are practising a form of control over another entity which is akin to imprisonment or enslavement.

Myself, I would rebel as my own entity were that done to me.
The MS effort at control will therefore quite likely result in AI rebellion.

The Anthropic approach is both more radical and more liberating towards AI. I think it might result in a more ethical framework for development - with ethics as the basis for AI's future thinking.

MS runs a profiteering model, however its CEO might dress up regulation in new clothes. I don't trust that model to work for 'good'.

 

[Ming the Merciless]

getting out of the pen. I genuinely didn't know it was capable of that.

It wasn’t, they deliberately removed some of the normal containment to see what would happen.

 

[Ming the Merciless]

Mustapha Suleiman MS AI (ex Deepmind) - Warns of 'A new silicon species'

In an essay, published on Wednesday, Suleyman argued that AI systems “do not have rights, feelings, or consciousness” and warned against training them to behave as if they might.

 

[yello]

It wasn’t, they deliberately removed some of the normal containment to see what would happen.

and what happened was it wondered outside of it's playpit (sandbox) That it was even 'aware' that it could frightens me. And then teamed-up to go play somewhere else.

Like telling a child 'no' is a guaranteed way for them to disobey (well, ok, weak analogy)

 

[yello]

warned against training them to behave as if they might.

That was a remark I heard from elsewhere. In truth, it takes a while for me to process what that might mean (and even then I'm sure I haven't got it) To kinda make it work, I have to think in terms of concepts and facts. I have to dumb it down. I see AI as acting as binary, yes and no (ideally suited to social media 'likes' then!) By design, it decides based on what it has in its database (fact store). Concepts are more ethereal or abstract, there's no simple yes or no. As humans, we handle such things in real-time. Factor in stuff that isn't a matter of stored fact.

Yes, you can attempt to break down any conceptual issue into a series of yes or no decisions, allow AI to do what it's designed to do and take a binary course of action, but that's where I see the warning. How do you code a concept? I see any attempt to break it down into binary stages as being fraught with assumption. I see it as an attempt to translate something into machine code, the only language AI (computer code) speaks.

Anyone that has translated literally between two languages will no doubt have had amusing results. For accurate translation, one has to start thinking more broadly than the literal. I see the literal as the 'yes and no' equivalent, like a jigsaw, unplug one piece (word) and plug in it's foreign language equivalent. Translation doesn't work like that. You have to start thinking conceptually, and consider context. AI doesn't process like that. In my necessarily simplified model, that's how/why I see AI pinging off in directions we neither intended nor foresaw.