• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Master AI (Artificial Intelligence) Discussion/News Thread

I’m not sure I see all this repressed creativity that AI will give people the freedom to explore. Some people create or do creative things, but I suspect leisure for most folks means scrolling TikTok, watching TV, or playing video games. Not everyone is creatively compelled, and creativity often needs deep practice (I did not say skill) and point of view to be satisfying. Examples abound.

And unless we come up with a way for people to have the money that leisure demands, people won’t be able to afford leisure. I have a hobby machine shop and I am aghast at the current cost of steel and brass raw materials. Photography is so ubiquitous now that few young people put art of any kind on their walls. Like I said, examples abound.

And then there’s the trend towards nihilism. If nothing we do is as good as what software can do, why should we do anything at all?

Like I said a couple of pages ago, I’m a little too Calvinistic to have much faith in human ideals.

Personally, I think the powers that be will think of ways to reduce costs even if it eliminates the market for their products, and then they’ll be shocked at the result. People do not think in the long term, and they naturally predict the future by extrapolating current direction linearly. That undermines both sides of the discussion.

Rick “there is no law that technology advancement makes things better” Denney
 
ChatGPT. I have a month of free trial on one of their higher tier compute.

As some of us train AI models... shouldn't we be compensated for their automated output of our input? Doesn't that mean my intellectual property is churning out revenue far more than ever before? Shouldn't my AI model count as me still. working for whatever company I created it for? Sure patents did some of it, but AI is operationally profitable, while IP is kind of a one off licensing stuff for patents that do not evolve or change.

"AI" is not operationally profitable yet for the big AI companies, quite the opposite.

"Paid for training AI" - this is being thought about now - especially companies like Google who are grappling with how they are going to monetize their next 10yrs. and fix the mess their greed caused.

A reminder of what was Google's original mission "to organize the world's information and make it universally accessible and useful" - which they did, and following this ethos, created an incredible eco-system for individual publishers. The Adsense model created millions of viable passive income streams, encouraged knowledge publishing etc. etc. It was brilliant, fair, and a good user experience for all. This changed about 9yrs ago when they started over-emphasizing google ads on search results (which kept increasing in price). This thinning out - in combination with the rise of social media "influencers" - made them make the horrible decision(s) to systematically drain those who got them there..

Why the stroll down memory lane? With AI search results that do not have proper citations/credit going to the publisher (and their site) + the continued favouritism of the larger advertisers - they are about to wipe out the golden geese that were providing good/relevant/interesting relevant content.

And so here we are, existential corporate decisions will be made. IMHO it is again the public who will decide. The slop that AI is bringing back+flashing ads is not a good experience. And it is a full blown ouroboros condition - it is learning on published slop because it devalued relevant good content and round and round we go. So we either turn into Idiocracy and no one cares, or someone builds better mousetraps. Interesting times.
 
I’m not sure I see all this repressed creativity that AI will give people the freedom to explore. Some people create or do creative things, ...

The problem with the kind of that kind of creativity -i.e. everybody using the same tool- is that there's likely to be little differentiation in the output. Which defeats the purpose of being "creative".

And unless we come up with a way for people to have the money that leisure demands, people won’t be able to afford leisure.

Exactly. If you disenfrachise your potential customer base...

... Personally, I think the powers that be will think of ways to reduce costs even if it eliminates the market for their products, and then they’ll be shocked at the result. People do not think in the long term, and they naturally predict the future by extrapolating current direction linearly. That undermines both sides of the discussion.

Indeed it is a "nobody wins" premise.
 
The problem with the kind of that kind of creativity -i.e. everybody using the same tool- is that there's likely to be little differentiation in the output. Which defeats the purpose of being "creative".



Exactly. If you disenfrachise your potential customer base...



Indeed it is a "nobody wins" premise.


No. AI is a prediction tool. Yes it will have its "canned" answers, preferred "style", but it is up to you to direct it - it will not disobey you, you just need to work with it. And in there is the creativity and skill. As a creative, you need to find ways to describe what it is you are looking for. All pianos play notes, the artist directs the composition and performance outputs.
 
No. AI is a prediction tool. Yes it will have its "canned" answers, preferred "style", but it is up to you to direct it - it will not disobey you, you just need to work with it. And in there is the creativity and skill. As a creative, you need to find ways to describe what it is you are looking for. All pianos play notes, the artist directs the composition and performance outputs.
Actually AI relies on a limited number of neural models for certain tasks... and no amount of prompting will change how the neural model fundamentally works.

Like I have stated before, AI doesn't really "generate"... it inferences (that's the working term) based on data it's been trained on. That's not creation, it's merely reassembling existing stuff. Which incidentally is what most software programming is about these days, so it's no surprise that's an area of AI strength...
 
Last edited:
Actually AI relies on a limited number of neural models for certain tasks... and no amount of prompting will change how the neural model fundamentally works.

Like I have stated before, AI doesn't really "generate"... it inferences (that's the working term) based on data it's been trained on. That's not creation, it's merely reassemblinh existing stuff. Which incidentally is what most software programming is about these days, so it's no surprise that's an area of AI strength...
I think the sticking point here is treating everything as “just prompting,” which is only one layer. At that level, sure, prompting doesn’t change the underlying model. It’s just input shaping. But, that’s not the whole picture of how AI systems are actually used. There are multiple layers of control:
  • Prompting: how you ask the question
  • RAG (an AWS AI term for Retrieval Augmented Generation): letting the model pull from data you provide, like a database or documents
  • Fine-tuning: training it on examples so it behaves differently
  • Full training: building or retraining the model from scratch, like the foundation models behind ChatGPT, Claude, or Gemini. This is obviously expensive and usually requires a team.
Once you move beyond prompting, you are no longer just “working around” a fixed model, and it becomes more of an iterative process of testing and refining. You are changing what information it has access to, and in the case of fine-tuning or training, you are literally changing how it behaves.

So the idea that “no amount of prompting will change how the model works” is technically true but incomplete. It ignores the fact that real world AI systems are almost always a combination of these layers, not just a raw model responding to a prompt.

If prompting is like telling a musician what to play, RAG is giving them sheet music, fine-tuning is training them into a certain style, and full training is basically teaching them how to play the instruments.

So the model itself might be fixed at one layer, but the overall system absolutely isn’t.
 
This is a key moment in AI.

Summary, Anthropic latest Claude model is capable of finding software vulnerabilities. Anthropic decided to not release this latest model the the general public for the protection from malicious actors.

Still not a believer of AI? You need to start using it daily, ask it questions and have it done tasks that would normally take you a lot of effort. And if you are still not convinced this is not only just the future, it's a key moment in humanity's achievement, then there is a gap between you and reality.


 
This is a key moment in AI.

Summary, Anthropic latest Claude model is capable of finding software vulnerabilities. Anthropic decided to not release this latest model the the general public for the protection from malicious actors.

Still not a believer of AI? You need to start using it daily, ask it questions and have it done tasks that would normally take you a lot of effort. And if you are still not convinced this is not only just the future, it's a key moment in humanity's achievement, then there is a gap between you and reality.


What you’re saying about AI and Anthropic’s developments resonates with me, but it also made me think of something else.

Last week I was at my daughter’s birthday and spoke with a group of her friends people around 28 to 34 years old some i know quite well. Different educational backgrounds, in the middle of their careers, some with young families and mortgages.

I asked them how they see AI, both in their work and personally. What stood out to me was that they do use it sometimes often even facilitated by their employers but they mostly seem to view it as just part of life. Not really as something threatening or deeply transformative more like: it is what it is.

That got me thinking. Is that a sign of realism and adaptability? Or could it be a form of underestimating how impactful these developments might become?

Maybe age plays a role here as well. I notice I look at these changes differently now than I did at their stage in life. With everything that has happened in recent years from COVID to geopolitical tensions and economic uncertainty I sometimes wonder to what extent their view of the future aligns with the reality that is taking shape.

At the same time, I do admire how they are navigating their lives and responsibilities. Perhaps their attitude is actually a way of coping with all that uncertainty.
 
Summary, Zoom CEO believe the workweek will shorten to 3 days a week by 2031 due to AI.

I agree that it ultimately will, but not by this much, this fast. I think it will gradually shorten to 32 hours spread out over 5 days by 2045, then gradually trend down to ultimately 25 hours/week by 2070. And it will get stuck there for 75+ years because of the world's overall population decline.

 
I notice I look at these changes differently now than I did at their stage in life.
As you age, you start to see the world with a more analytical lens, you start to appreciate the cause and effects, you think about it more deeply and philosophically.
 
Still not a believer of AI? You need to start using it daily, ask it questions and have it done tasks that would normally take you a lot of effort. And if you are still not convinced this is not only just the future, it's a key moment in humanity's achievement, then there is a gap between you and reality.

Hi a.i., make my tea for me, make the bed, do my chores, wipe my arse……….

….still waiting :rolleyes:

All the above tasks involve a lot of effort, my reality is very secure thank you very much
 
Hi a.i., make my tea for me, make the bed, do my chores, wipe my arse……….

….still waiting :rolleyes:
This is called robotics. We have already solved all the mechanic aspects of these robots, the one thing that we have yet to solved is the AI software to control it. . .and this is coming. . .sooner than you think, I say within the next 10-15 years before production scale robots.


All the above tasks involve a lot of effort, my reality is very secure thank you very much
If you are at a certain age where you can kick the bucket right when these robots are affordable to own by the masses, sure.
 
As you age, you start to see the world with a more analytical lens, you start to appreciate the cause and effects, you think about it more deeply and philosophically.
Age probably makes you more analytical and maybe a bit more risk-aware too.

That said, the level of change since COVID feels unprecedented to me. I’m 67 and honestly haven’t seen anything like it.

What I do wonder is many people in their late 20s and early 30s (Gen Z millennials) grew up in relatively stable, prosperous conditions. Now they’re facing a much more volatile world.

So is that “it is what it is” mindset realism or just what it looks like when you’ve mostly known stability and things suddenly start shifting?

Then again, maybe it’s simply their way of adapting.
 
This is called robotics. We have already solved all the mechanic aspects of these robots, the one thing that we have yet to solved is the AI software to control it. . .and this is coming. . .sooner than you think, I say within the next 10-15 years before production scale robots.



If you are at a certain age where you can kick the bucket right when these robots are affordable to own by the masses, sure.

I’d absolutely love a robot to help me with my SPMS riddled fu*ked body (it’d give my mum and best mate a break :p) but I only have £239 in the bank which will last till i get my next disability payment in 13 days, and at the age of 50, with my current disease progression, I will not live long enough to see them come down in price far enough so that they are affordable for those who actually need them ;)
 
Last edited:
Remember when this guy lied to you?

Joe-Isuzu.jpg


Now 'AI' will do it—and probably "better":

 
I've installed several AI models here locally (not referring to AI Music now). Google recently released their Gemma4 AI models that anyone with a reasonably powerful PC can run. They even have versions of it that will run on phones.

They are classified by size and function. Regular models that work like Gemini and MoE (mixture of expert) versions that work more like Grok. The released models are just large data files. They need a controller to do anything. I'm using a program called ollama to run the models. You can let it run in the background as a server for a UI or use it as a command line and run it in a terminal. The sweet spot model for my PC is gemma4:e4b. It responds almost as fast as Gemini or Grok. Performance is determined by ram, and more by vram and a fast graphics card.

Since these are local modals they only know about current events, devices and software that happened / were available before their "training" ended. There are addons that will allow them to query the internet if needed, but they aren't included by default. The options are controlled by settings in the UI being used. The UI looks very similar to Gemini or Grok on the web.

The UI I'm using is Open WebUI served by the ollama server. It allows you to create workspaces that select one of the models and any extra knowledge you want to give them (pdfs, text files, etc). I created a workspace with gemma4 that knows about my Tascam 24 track recorder (pdf manual) as a test and to use. This model/configuration is using 5.6gb of vram btw.

The followups listed after the second response work just like Gemini. Just click on one for that question.

I asked it if could answer questions about the pdf manual:

Response:

Gemma4:e4b (Fast)
Today at 11:35 PM
Retrieved 1 source
Thought for 3 seconds

I can answer questions based on the technical specifications, conventions, and operational details provided in the manual context. For instance, I know that:
  • Buttons and physical parts are shown in bold font, such as the MENU button DP-24SD_Manual_vB.pdf.
  • The menu includes items like SONG for song operations, AUDIO DEPOT for WAV file import/export, and FORMAT for formatting SD cards DP-24SD_Manual_vB.pdf.
  • A CAUTION regarding the AC adapter (PS-P1220E) states that you must always use the included adapter and not use it with other equipment to prevent malfunction, fire, or electric shock DP-24SD_Manual_vB.pdf.
Please provide your specific questions regarding the unit's functions, menu items, or operational procedures.

I asked "How do I set up to record a track? What steps?"

Response (was instant):

  • Gemma4:e4b (Fast)
    Today at 11:36 PM
    Retrieved 1 source
    To set up for recording a track, follow these general steps:

    1. Prepare the Tracks:
      • Press the REC button of the tracks you wish to record to put them into recording standby (the REC buttons will start flashing) DP-24SD_Manual_vB.pdf.
      • If a track's REC button is unlit, you cannot hear the input sound for that track DP-24SD_Manual_vB.pdf.
    2. Adjust Levels and Position (Monitoring): Use the track faders and PAN knobs on the Mixer screen of the tracks you are monitoring to adjust levels and stereo positions for making a new recording DP-24SD_Manual_vB.pdf.
    3. Assign Inputs: Use the procedures described in the chapter preceding this section to assign input sources to the recording tracks DP-24SD_Manual_vB.pdf.
    4. Start Recording: Press the RECORD (0) button to begin the recording process DP-24SD_Manual_vB.pdf.

    If you are recording a guitar to TRACK 1, the specific procedure is:
    1. Press the RECORD (0) button to start recording and play the guitar. The REC button for TRACK 1 will stop flashing and light steadily DP-24SD_Manual_vB.pdf.
    2. After your performance, press the STOP (8) button to stop recording. The file is saved in the folder of the currently loaded song DP-24SD_Manual_vB.pdf.
    3. Press the REC button for TRACK 1 until the button becomes unlit DP-24SD_Manual_vB.pdf.
    1 Source
    Follow up
    What is the procedure for naming or managing the files after I stop a recording?

    If the input sound isn't going through correctly, what's the first thing I should check according to the manual?

    Can you detail the process for assigning multiple input sources simultaneously to different tracks?

    What are the specific guidelines for setting the fader levels for optimal recording quality?


 
More info. It is very easy to add Text to Speech output and Microphone input if desired / needed. I read fast and type fast so I don't have those options turned on.

Model sizes:

NAME SIZE
gemma4:31b 19 GB
gemma4:26b (MoE) 17 GB
gemma4:e4b 9.6 GB
qwen-16k:latest 9.5 GB
deepseek-r1:14b 9.0 GB
phi4:14b 9.1 GB
llama3.2:1b 1.3 GB

The b in those names indicates the number of parameters present in the Model database (how much it "knows")

gemma4 models are from Google.
qwen is from Alibaba
deepseek is from Deepseek AI
phi4 is from Microsoft
llama is from Meta

The gemma4 models all have "vision" / image recognition cababilities. You can "upload" an image and ask questions about it.

An example (I produced (didn't say created this time) this image with a prompt in an AI image generator called Fooocus:

1775973894907.png


I said describe this image (used gemma4:e4b, the smallest one). Response:

Gemma4:e4b (Fast)
Today at 1:04 AM
Thought for 2 seconds

The image depicts a woman standing on a prominent, elevated section of a cliff overlooking a turbulent sea.

The woman, viewed from the back, is dressed in a long, flowing, light-colored garment. She is positioned near the edge of the rock formation.

The setting is characterized by dramatic, layered cliffs. To the right, the cliff face shows significant evidence of human construction, including what appears to be towers, architectural structures, and built-up areas carved into the rock.

The body of water is visible to the left and extends into the distance. The ocean shows considerable wave action, with whitecaps breaking against the base of the cliffs.

The lighting is characterized by a bright, diffused glow emanating from the upper left, suggesting either dawn or dusk. The overall composition places the solitary figure against a backdrop of massive, natural rock formations juxtaposed with structured, man-made elements.

Follow up

What type of civilization or era do the visible structures suggest?

Do the cliffs look like they are made of sedimentary or igneous rock?

Based on the wave patterns, what might the current weather conditions be?

Are there any details about the woman's attire that suggest a specific period or culture?
***End gemma4***


I "uploaded" the same image to the qwen model which is supposed to be "trained" to be more creative and asked it to describe the image. Response was instant:

This is a breathtaking and atmospheric digital painting rendered in a painterly, impressionistic style, evoking a sense of epic scale and quiet contemplation. The scene is dominated by a lone figure standing on the edge of a vast, rugged cliff overlooking a tumultuous sea, while a distant, mysterious landmass looms ahead.

The Central Figure:
In the foreground, a solitary person, seen from behind, stands on a rocky outcrop. They are dressed in a flowing, loose-fitting white robe that billows slightly with the wind, and their long, light-colored hair is also swept by the breeze. The figure’s posture is still and upright, gazing intently out toward the horizon. This suggests a moment of profound observation, perhaps awe, solitude, or a journey’s end. The figure is illuminated by soft, warm light, making them stand out as the focal point against the cooler, darker tones of the landscape.

The Landscape and Environment:
The environment is both majestic and desolate. The immediate foreground is a steep cliffside covered in patches of mossy green and weathered rock, with sharp, angular edges. Below and to the left, the sea churns with powerful waves that crash against distant cliffs, rendered in dynamic brushstrokes of white and deep teal. The sea extends far into the distance, meeting a hazy horizon.

To the right and further ahead, a massive, ancient-looking cliff formation rises. Perched precariously on its edge is a fortified settlement or fortress, with stone towers, archways, and structures carved into the rock. The architecture appears old and weathered, suggesting a lost or mystical civilization. One tall, slender tower is particularly prominent on the distant peak.

The Sky and Atmosphere:
The sky is filled with soft, textured clouds, illuminated by a warm, golden-yellow light that seems to originate from a low sun just beyond the horizon or perhaps from within the clouds themselves. This light casts a luminous glow over the entire scene, enhancing the ethereal, almost dreamlike quality. The atmosphere is thick with haze and mist, which blankets the lower portions of the landscape and softens the edges of distant features, creating a sense of depth and mystery.

Artistic Style and Mood:
The artist employs broad, expressive brushstrokes, especially visible in the depiction of the sky, water, and distant cliffs, which gives the image a textured, hand-painted feel. The color palette is largely muted and atmospheric, with cool blues, grays, and earth tones contrasted by the warm yellows and whites of the sky and the figure’s robe. The overall mood is one of solitude, wonder, and epic adventure. The scene feels timeless, as if it could be from a fantasy epic or a mythological tale.

In summary, this image captures a moment of quiet, awe-filled contemplation on the edge of the world, where a lone wanderer stands before an ancient, mystical land, bathed in the warm, ethereal glow of an otherworldly light. It speaks to themes of journey, discovery, and the sublime beauty of nature and architecture coexisting in a fantastical realm.

Follow up

What inspired the artist to create this scene?

Could this image be from a specific book or game?

What emotions does this image evoke in you?

How might the story change if the figure were facing the opposite direction?

Is there any symbolic meaning behind the lone figure's white robe?
 
I'm genuinely fascinated by the two extremes of the debate: on one side, the outright detractors who refuse to use it or consider it an integral part of our society, or who see it as the beginning of humanity's destruction; on the other, the overenthusiastic, convinced that from now on the world will be a wonderful place where people will work 10 hours a week and earn double.
Needless to say, these are two visions that are both incompatible with the reality that is unfolding. Artificial intelligence is a reality we must reckon with, for better or for worse. It certainly has enormous utility and potential, both for better and, unfortunately, for worse, and I sincerely believe that right now, not even those in the field REALLY know where the next developments will take us, both from a purely technological standpoint, but even more importantly, from an economic and social one.
Personally, I believe that the only thing we ordinary users can do is to stay alert, informed, know the technology, try to understand it and integrate it into our lives in the most useful way possible and maintain a certain mental flexibility to avoid finding ourselves overwhelmed or supplanted by an AI.
 
I'm genuinely fascinated by the two extremes of the debate: on one side, the outright detractors who refuse to use it or consider it an integral part of our society, or who see it as the beginning of humanity's destruction; on the other, the overenthusiastic, convinced that from now on the world will be a wonderful place where people will work 10 hours a week and earn double.
Needless to say, these are two visions that are both incompatible with the reality that is unfolding. Artificial intelligence is a reality we must reckon with, for better or for worse. It certainly has enormous utility and potential, both for better and, unfortunately, for worse, and I sincerely believe that right now, not even those in the field REALLY know where the next developments will take us, both from a purely technological standpoint, but even more importantly, from an economic and social one.
Personally, I believe that the only thing we ordinary users can do is to stay alert, informed, know the technology, try to understand it and integrate it into our lives in the most useful way possible and maintain a certain mental flexibility to avoid finding ourselves overwhelmed or supplanted by an AI.

I see them as both tools and toys.
 
Back
Top Bottom