Master AI (Artificial Intelligence) Discussion/News Thread

[rdenney]

This is an interesting point and one I've thought a lot about.
My years (decades) of working in financial markets has taught me that when people say of the latest fad, "this time it's different", it usually isn't.
Yet, to me at least, AI does feel different and it worries me greatly.
I guess another thing I've learnt (or I'm trying to learn) is that it's ok to not know the answer. Living with uncertainty is ok

I wrote a long post about the proposition that innovation always leads to net gains in employment is no guaranteed, and about how all such innovations lead to economic displacements that can ruin lives even if they do ultimately result in a net gain in general wealth. I even wrote that trying to avoid those real consequences by stifling innovation usually backfires and sustains rather than avoids negative outcomes. I gave some examples. But then I thought some doofus will feed my post into ChatGPT and throw that back at me, and it made me think that all human discourse is a depressing waste of time.

Rick "wondering if people will like AI-generated music when it starts copying itself in repeated generations rather than mimicking primary human artistic creation" Denney

 

[respice finem]

...Rick "wondering if people will like AI-generated music when it starts copying itself in repeated generations rather than mimicking primary human artistic creation" Denney

I could think of a different scenario (not tomorrow, but relatively soon): AI generated artists, widely acknowledged before the "joke" is discovered.
And maybe even after...

 

[xanalog]

I like to pretend to be privacy-minded but mostly when it is convenient.
I asked some chatBot about the competition that is more conducive for those who are (or pretend to be) privacy-minded.
Herewith some options [verbatim] for like-minded members:
------------------------------------------------
The current AI landscape definitely favors data collection over user privacy and accessibility.
There are some AI tools that work without registration (like some open-source models you can run locally), but they often require more technical setup.
The trade-off between convenience and privacy is real, and it's understandable that privacy-conscious users get shut out of easily accessible AI experimentation.
Here are some popular AI models you can run locally without any sign-ups:
Text/Chat Models:

• Ollama - Easy installer that runs models like Llama, Mistral, CodeLlama locally
• LM Studio - User-friendly GUI for running various open models
• GPT4All - Simple desktop app with multiple model options
• Kobold.cpp - Text generation focused on creative writing

Image Generation:

• Stable Diffusion (via ComfyUI, AUTOMATIC1111, or Fooocus)
• DALL-E Mini/Craiyon - Though this one actually runs in browser
• InvokeAI - Local Stable Diffusion with nice interface

Code Assistance:

• Code Llama (through Ollama or LM Studio)
• StarCoder models
• WizardCoder

All-in-one Solutions:

• Jan.ai - Desktop app with chat and multiple models
• Msty - Another desktop client for local models

Most of these work best with decent hardware (especially for image generation), but many text models can run on regular laptops. Ollama and LM Studio are probably the easiest starting points - they handle the technical setup and let you download models with simple commands or clicks.
The setup is usually: install the software, download a model (few GB), then start chatting or generating locally with complete privacy.
----------------------------------------------------
Cheers!

 

[somebodyelse]

Rick "wondering if people will like AI-generated music when it starts copying itself in repeated generations rather than mimicking primary human artistic creation" Denney

I'm not sure if you're thinking about listening to the result of model collapse, or that they might find a way around it.

 

[whataboutmusic]

I like to pretend to be privacy-minded but mostly when it is convenient.
I asked some chatBot about the competition that is more conducive for those who are (or pretend to be) privacy-minded.
Herewith some options [verbatim] for like-minded members:
------------------------------------------------
The current AI landscape definitely favors data collection over user privacy and accessibility.
There are some AI tools that work without registration (like some open-source models you can run locally), but they often require more technical setup.
The trade-off between convenience and privacy is real, and it's understandable that privacy-conscious users get shut out of easily accessible AI experimentation.
Here are some popular AI models you can run locally without any sign-ups:
Text/Chat Models:

• Ollama - Easy installer that runs models like Llama, Mistral, CodeLlama locally
• LM Studio - User-friendly GUI for running various open models
• GPT4All - Simple desktop app with multiple model options
• Kobold.cpp - Text generation focused on creative writing

Image Generation:

• Stable Diffusion (via ComfyUI, AUTOMATIC1111, or Fooocus)
• DALL-E Mini/Craiyon - Though this one actually runs in browser
• InvokeAI - Local Stable Diffusion with nice interface

Code Assistance:

• Code Llama (through Ollama or LM Studio)
• StarCoder models
• WizardCoder

All-in-one Solutions:

• Jan.ai - Desktop app with chat and multiple models
• Msty - Another desktop client for local models

Most of these work best with decent hardware (especially for image generation), but many text models can run on regular laptops. Ollama and LM Studio are probably the easiest starting points - they handle the technical setup and let you download models with simple commands or clicks.
The setup is usually: install the software, download a model (few GB), then start chatting or generating locally with complete privacy.
----------------------------------------------------
Cheers!

I am positive that AI models are not in line with Europe's laws about privacy and that is not a good sigh for now.

Even if an app runs locally it could send data to the company and if its not open source no one can verify what is being sent to the mothership.

I believe that there are huge privacy concerns with AI....but privacy has many layers every small step matters.

 

[xanalog]

I believe that there are huge privacy concerns with AI....but privacy has many layers every small step matters.

You can't put the cat back in that box... that your shiny new phone came in!

Europe's laws about privacy...

LoL, post-mortem:
Someone should remind EU elite that the cat was let out of the box, way before (=2G days?) EU privacy concerns.

 

[Chr1]

No doubt the US elite would love their pharma companies to be able to bypass those pesky EU regulations re their synthetic opioids etc too. Lol.

 

[whataboutmusic]

You can't put the cat back in that box... that your shiny new phone came in!

LoL, post-mortem:
Someone should remind EU elite that the cat was let out of the box, way before (=2G days?) EU privacy concerns.

Yes true privacy does not apply in an online world but as you said you have to choose between privacy and comfort.

But you can take measures and act on what you can control....for example you can install GrapheneOS and not let google or apple track all of your data.

AI scares me -for other reasons too- but also because people willingly giving up private information and data unregulated...to a model that uses the data to build on them and those companies collect the data..and not just searches and posts but way more personal data and information.

At least EU took a step...before that most people didn't know what cookies are.

 

[Chr1]

Wonder if there's room for a thread titled : "The Risks and Ethics of Unregulated AI."

And another called : "So What. That Genie's Long Gone Too:...Discuss."

 

[Salt]

Wonder if there's room for a thread titled : "The Risks and Ethics of Unregulated AI."

And another called : "So What. That Genie's Long Gone Too:...Discuss."

Or "Just wait until the bubble implodes " , as some say ...

 

[antcollinet]

Young people are using this tool as a virtual friend or therapist.

So are old people.


It can be quite useful to talk through problems with. Perhaps dangerous though for someone with an actual mental health issue.

 

[antcollinet]

Not perfect but stunning AI generated duet by "Miley Cyrus" and "Adele:"

It is better than a lot of their real music!

Has that been done with the permission of the imitated artists - because if not, it is not OK.

 

[Chr1]

So, my two favourite concerns re AI are: The Use Of Deep Fakes for Nefarious Political/Societal Change. "

And the aforementioned potential multitude of scary new toxins.

On the flipside there's new antibiotics.
Plus, love the way it enables me on my phone re Lens, Circle To Search, photography etc tho.

 

[pablolie]

I wrote a long post about the proposition that innovation always leads to net gains in employment is no guaranteed, and about how all such innovations lead to economic displacements that can ruin lives even if they do ultimately result in a net gain in general wealth. I even wrote that trying to avoid those real consequences by stifling innovation usually backfires and sustains rather than avoids negative outcomes. I gave some examples. But then I thought some doofus will feed my post into ChatGPT and throw that back at me, and it made me think that all human discourse is a depressing waste of time.

Rick "wondering if people will like AI-generated music when it starts copying itself in repeated generations rather than mimicking primary human artistic creation" Denney

Efficiencies have always been the driver. Telephone switch board or elevator operators. The friendly on-site IT guy. Etc.

And even though everybody says AI is unique in impacting the service industry, that's utterly untrue. Salesforce replaced more sales operations people than anything before it in corporate. Jira and Confluence replaced PMPs galore. We have gone through these efficiency exercises a hundred times before.

My main worry is that AI does many things that we typically ask junior people entering the corporate world to do.

AI is a copy artist and plagiarist. It is great at searching and consolidating. But not at creating anything really new.

 

[xanalog]

I blew one the chatBot's fuse, with a simple task.

I asked it to create an image of a 'reverse Oreo cookie'.
2 outer layers of white cream filling with 1 inner layer of brown cookie.
Seven attempts (and 7 garbled images) later, it apologized and just gave up.

I worded my request so many which ways that.... ...but I did refrain from calling it names.

 

[sam_adams]

The friendly on-site IT guy.

Who?

See, also.

 

[Ralph_Cramden]

Google’s latest smartphone AI is leaving Apple in the dust​

https://finance.yahoo.com/news/googles-latest-smartphone-ai-is-leaving-apple-in-the-dust-160056479.html

 

[sam_adams]

Google "AI" results return fake phone numbers allowing users to be scammed:

Scam numbers crop up in Google's AI Overviews, per user reports

User reports say Google's AI returned false phone numbers that turned out to be scams over the real deal.

www.androidcentral.com

 

[somebodyelse]

Wonder if there's room for a thread titled : "The Risks and Ethics of Unregulated AI."

And another called : "So What. That Genie's Long Gone Too:...Discuss."

It's bringing a whole new class of security and compliance problems as those aspects get ignored in the rush. Like M365 Copilot access to files not appearing in the audit log - fixed only after it was reported a second time, and not announced to users, so companies won't know that their audit logs may be incomplete.
https://www.theregister.com/2025/08/20/microsoft_mum_about_m365_copilot/

 

[AdrianusG]

I guess it falls to trustworthy lawyers to determine liability then.

"trustworthy lawyers"

Yeah, but where to find them