[eltonioni]

Why is knife crime rising in so many places? How do we stop it? Can we balance punishment with prevention?

Because it is a symptom and not a cause.

No amount of legal intervention is going to make knife crime go away and fawning around storytellers hawking around a bad script isn't going to improve things.

[kanga]

https://www.abc.net.au/news/2025-02-03/ ... /104886722

Australian financial commentator 's analysis

"DeepSeek's emergence signals the beginning of the human-replacing phase of AI"

[OCB]

@kanga - knife crime/culture is indeed way off-topic wrt to this discussion.

It genuinely pains me to say - I’ve probably got more personal experience wrt knife crime than the majority here, but - it’s got nowt to do with AI.

We used to take such things up in the Politics section b4 it got bizarrely paywalled.

[kanga]

Why is knife crime rising in so many places? How do we stop it? Can we balance punishment with prevention?

...

Er, I didn't write that. I merely quoted the BBC gist of a documentary programme

[OCB]

@kanga soz - point taken

I don’t think folks here understand that the developments over the past few months are so astonishing, with regard to “expert level”

As I said - I’m a life-long technologist. 30+ years since I started to get paid to do this stuff on a monthly basis, & have been paid to do what I do almost every month since.

In terms of what I do for a living - no…I’m much closer to the wires & hardware than most.

[OCB]

Honestly - to prove the point how “smart” AI currently is, I’d challenge folks on here to think of a question.

It has to be a question that’s possible to be solved by reason - not external expert knowledge.

One step at a time.

I absolutely want an AI to explain to me why cylinder 7 might appear to be misfiring, without taking the head off.

[Mz Hedy]

I really must dig out and once again re-read my copy of "Machines with a Purpose" (New Scientist review here).
This is a more recent article from Scientific American which covers similar ground in much shorter format.
Basically, it all rests on human purpose being embedded in the machines designed by humans. Like several contributers on here, I learned to program computers back in the 1970s. None of us expected elements our code to be still in use 20 to 30 years later; especially the date calculation routines

Maybe a question to meet @OCB 's challenge is "how can we ensure that future AI engines answer the actual needs of the human user, not what the human erroneously said they needed?" A bit like the genie and the third wish.

[OCB]

@Mz Hedy - I hear ye, I grew up reading Asimov & Arthur C Clark.

We’ve gone from theory to reality in a few months.

Yes - skiffy warrants zero respect..

Skiffy debated this stuff decades ago.

[Flyingfemme]

"how can we ensure that future AI engines answer the actual needs of the human user, not what the human erroneously said they needed?"

The perennial problem. Users don't generally know what they want. They frequently confuse the problem with the solution - that is why we need systems analysts. AI seems designed to work for the end user without analysts. We are going to get a lot of systems that are useless/pointless and (hopefully) a stepping stone in the way to solving the actual problem. At least AI is cheap and fast, so an iterative process won't take forever. If users understand that the iteration will be required

[eltonioni]

Maybe a question to meet @OCB 's challenge is "how can we ensure that future AI engines answer the actual needs of the human user, not what the human erroneously said they needed?"

At the risk of bringing the tone down, "actual needs" seems more like a moral challenge. Otherwise AI will be reduced to ironing and sex robots in Currys window.

[johnm]

@Flyingfemme makes a very good point. I was never the best coder in the world and quickly moved on to systems analysis and project management. Finding out what customers needed rather than wanted was always a challenge and management were always part of the problem.

They would tell you what they wanted and often made it difficult to talk to the workers who could tell you what was needed if you asked the right questions This problem persists to this day and is why so many systems are a pain in the proverbial.

[kanga]

@Flyingfemme makes a very good point. I was never the best coder in the world and quickly moved on to systems analysis and project management. Finding out what customers needed rather than wanted was always a challenge and management were always part of the problem.

They would tell you what they wanted and often made it difficult to talk to the workers who could tell you what was needed if you asked the right questions This problem persists to this day and is why so many systems are a pain in the proverbial.

.. which was a particular aggravating feature in the public sector (civil and military, central and local government). Projects which could have been developed (and modified after introduction when needs or experiences evolved) in-house by close colleagues of the end-users were instead perforce (by ideological doctrine) outsourced or privatised. These were then developed to satisfy an early immutable specification, and delivered as an equally immutable turnkey project.

[eltonioni]

Are we close to understanding why things are done AT the little people rather than FOR the little people. Once "leaders" take the messy, imperfect human factors out of decision making they are just serving themselves. Which is nicely on topic...

Isn't that the big fear about Artificial Intelligence - that it does things AT humans rather than FOR humans? What's the difference?

[Mz Hedy]

@Mz Hedy - I hear ye, I grew up reading Asimov & Arthur C Clark...

You hear me, but did you listen?

I have read some Azimov and Clark, also some Gibson, Čapek, Grimm, Shakespeare, and Ovid*; all of whom have stories containing themes of 'be careful what you ask for'.

My question was a proposal in response for your request for one to be put to an AI to answer while showing its working.

I think @eltonioni has got closest to my meaning.

I think I'll stop at that - this is a flying forum after all and we're not using self-learning AI in our aeroplanes yet, are we?

* in English translation

[kanga]

https://www.abc.net.au/news/2025-02-04/ ... /104896770

Australian Federal government directive:

"Chinese AI chatbot DeepSeek banned from government devices over security fears"