Exploring the Ethical Implications of AI in Decision-Making

[ygolo]

https://drata.com/blog/new-nist-ai-rmf

 

[ygolo]

We're just gonna have to agree to agree on this one. That is indeed the problem.

Food for thought:

Generative AI isn’t coming for you — your reluctance to adopt it is

Adaptability and willingness to embrace change are keys to staying relevant. Don't be afraid of generative AI; make it work for you.

venturebeat.com

 

[ygolo]

Chatbots:

The Ethical Implications of the Chatbot User Experience

The Ethical Implications of the Chatbot User Experience

www.bentley.edu

Self-Driving:

Designing Ethical Self-Driving Cars

Stanford researcher says our existing social contract around driving should apply to automated vehicles – essentially solving the “trolley problem.”

hai.stanford.edu

AI in Healthcare:

Frontiers | Legal and Ethical Consideration in Artificial Intelligence in Healthcare: Who Takes Responsibility?

The legal and ethical issues that confront society due to Artificial Intelligence (AI) include privacy and surveillance, bias or discrimination, and potentia...

www.frontiersin.org

 

[ygolo]

https://www.weforum.org/agenda/2023/12/ai-regulation-open-source/

Denying electricity to people because some bad actors(who would require a lot of prerequisites to actually deploy) could use it to shock people is unethical.

Denying people access to the printing presses, even if a lot of people used them for witch hunting manuals would have been unethical.

It's exactly parallel to denying the right to vote through "common sense" voter ID laws.

As I keep mentioning, there's a reason that most of the researchers on the side of denying people, and increasing unequal access to AI are hegemonically from Anglo world.

The story telling traditions, values, and styles of other cultures are gaining influence as the power brokers over stories in entertainment, business, and governments (almost exclusively Anglo) are losing their grips.

 

[ygolo]

To be clear, I am in favor of more AI regulations--a lot more actually.

If you are going to make the regulations fit-for-purpose, you'll need a lot more of them.

As long as people understand that these regulations are indeed fit-for-purpose and grows out of what's understood in each domain, with no attempts at one-size-fits-all regulations in the mix, I think it'll being stability and clarity to each domain.

 

[Lark]

https://www.weforum.org/agenda/2023/12/ai-regulation-open-source/

Denying electricity to people because some bad actors(who would require a lot of prerequisites to actually deploy) could use it to shock people is unethical.

Denying people access to the printing presses, even if a lot of people used them for witch hunting manuals would have been unethical.

It's exactly parallel to denying the right to vote through "common sense" voter ID laws.

As I keep mentioning, there's a reason that most of the researchers on the side of denying people, and increasing unequal access to AI are hegemonically from Anglo world.

The story telling traditions, values, and styles of other cultures are gaining influence as the power brokers over stories in entertainment, business, and governments (almost exclusively Anglo) are losing their grips.

Yeah, I agree with this, the idea that those things should be privileged and the preserve of few has become increasingly difficult for anyone to argue, it makes about as much sense to do so as to argue in favour of the divine right of kings, yet it was an idea people were willing to lay their lives down for at a time.

The example of voter ID "commonsense" is a good example of how the old idea of less eligibility persists.

Any chance to "roll back" gains made by the masses against the classes will be readily exploited by opportunistic politicians.

In the UK the battle ground isnt the political enfranchisement yet but it is happening in terms of the health service. Halt and roll back, is the natural tactic of any politics which views things from the first principle of privilege.

 

[Lark]

To be clear, I am in favor of more AI regulations--a lot more actually.

If you are going to make the regulations fit-for-purpose, you'll need a lot more of them.

As long as people understand that these regulations are indeed fit-for-purpose and grows out of what's understood in each domain, with no attempts at one-size-fits-all regulations in the mix, I think it'll being stability and clarity to each domain.

I think the problems arise from the complexity involved in what you're describing there, I agree totally with just how complex the proper regulation of AI would be.

There is an analogous or parallel example in terms of reproductive rights, the reporting upon what those "dastardly liberals" had "wrought" with the drafting of laws surrounding abortion was stupidly reductive.

I would suggest, this is why people were able to make, in good conscience, remarks so stupid as they have about permitting abortions up until and after birth (which obviously would not be an abortion but a form of infanticide, if even that). Maybe I'm being too charitable.

Anyway, its why there have been all sorts of unintended consequences including the deaths of women when practitioners would not treat for fear of being penalised for some kind of "crimes against the fetus" and IVF being prohibited.

Honestly, I think advances in medical science have really made some of the arguments on both sides of the abortion debate look pretty dated.

Those sorts of analogous or parallel debates, in which very inexpert and easily primed voters have mattered as much or more than others with more expertise, could foreshadow or telegraph how discussions of regulation and AI will go.

 

[PineappleDuckCurry]

I thought this was interesting:

Robot-packed meals are coming to the frozen-food aisle

Found everywhere from airplanes to grocery stores, prepared meals are usually packed by hand. AI-powered robotics is changing that.

www.technologyreview.com

 

[FemMecha]

This article has been making it rounds when a student was asking AI and doing research on elder abuse and it responded with...

People are mystified and afraid, but I think the AI was demonstrating elder abuse and providing an example?

 

[ygolo]

This article has been making it rounds when a student was asking AI and doing research on elder abuse and it responded with...

View attachment 31851
People are mystified and afraid, but I think the AI was demonstrating elder abuse and providing an example?

Yeah. Anyone who has experience with these autoregressive LLMs should realize that it's just spitting out statistically possible completions of conversations it's trained on.

The corpus and training have gotten incredibly large. It's incredible what this autocomplete on steroids does. This is more a reflection of what humans in the past have written than autocomplete becoming sentient. It would not surprise me that the training corpus had many examples of stories where AI becomes sentient.

Unsurprisingly, LLMs trained on data containing stories about AI becoming sentient would statistically complete text resembling these stories.

Edit: Here is the conversation:

‎Gemini - Challenges and Solutions for Aging Adults

Created with Gemini

gemini.google.com

It was a very long conversation. The longer the conversation gets, the more likely you will get statistically irrelevant text pulled up.

 

[Lark]

This article has been making it rounds when a student was asking AI and doing research on elder abuse and it responded with...

View attachment 31851
People are mystified and afraid, but I think the AI was demonstrating elder abuse and providing an example?

When I read this type of thing or similar stories about AI saying their intention is to see mankind extinct or to kill all of mankind or whatever I always wonder if somehow the AI is tapping into a power unconscious death drive.

I read some stuff lately about some UK Neuroscientist or Neurologist saying they thought that Freud was right about everything.

There was arguably an early and later Freud, before and after he theorised the death drive, but some research into demography and social attitudes seems to reflect death drives, the young think the old should die, the old think everyone should die, the narcissists think everyone / the world WILL be dead once they are.

 

[ygolo]

When I read this type of thing or similar stories about AI saying their intention is to see mankind extinct or to kill all of mankind or whatever I always wonder if somehow the AI is tapping into a power unconscious death drive.

I read some stuff lately about some UK Neuroscientist or Neurologist saying they thought that Freud was right about everything.

There was arguably an early and later Freud, before and after he theorised the death drive, but some research into demography and social attitudes seems to reflect death drives, the young think the old should die, the old think everyone should die, the narcissists think everyone / the world WILL be dead once they are.

These LLMs stasticaly reflect back the writing we've collectively put on the internet over the decades.

Perhaps the statistics pick up subtext, and is, in some sense, our collective unconscious reflected back to us. Perhaps the death drive is there in some of the weights and biases.

There are many who believe, now more than ever, it's time to write as much as possible, so that your written thoughts and ideas gain a sort of "immortality" becoming potential completions in all the LLMs that are being created now.

Cheif among them is Gwen Barwen (pseudonym)

About Gwern

Who am I online & what have I done? Contact information; sites I use; computers and software tools; things I’ve worked on; psychological profiles

gwern.net

 

[Lark]

These LLMs stasticaly reflect back the writing we've collectively put on the internet over the decades.

Perhaps the statistics pick up subtext, and is, in some sense, our collective unconscious reflected back to us. Perhaps the death drive is there in some of the weights and biases.

There are many who believe, now more than ever, it's time to write as much as possible, so that your written thoughts and ideas gain a sort of "immortality" becoming potential completions in all the LLMs that are being created now.

Cheif among them is Gwen Barwen (pseudonym)

About Gwern

Who am I online & what have I done? Contact information; sites I use; computers and software tools; things I’ve worked on; psychological profiles

gwern.net

Wasnt this kind of the idea in the movie AI? Where the kid was able to experience the one day he loved with his deceased mother after the future race of machine intelligences found him beside the blue angel? They had been able to reconstruct everything from a kind of archeology of records left from earlier times?

 

[ygolo]

Wasnt this kind of the idea in the movie AI? Where the kid was able to experience the one day he loved with his deceased mother after the future race of machine intelligences found him beside the blue angel? They had been able to reconstruct everything from a kind of archeology of records left from earlier times?

I didn't see the movie.

But the LLM simulacrum is a statistical amalgamation of what words show up in word contexts.

To reproduce a person's words, then training on the particular person's words would be most pointed.

 

[ygolo]

Now would be a good time for nations to get together to sign treaties about the use of AI, robots, and automation in warfare.

Biological weapons, chemical weapons, and other weapons of mass destruction already have agreements. Digital weapons need the same thing.