• 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 disagree. AI does not need to be trained on everything possible to evaluate technique.

Maybe artistry.

You have experience with training AI models? Define "Technique" outside of existing top routines. AI *NEVER* grows beyond its training parameters - it fails miserably and "hallucinates". It's a complete and total myth that current AI models learn independently.
 
I have been using Anthropic Claude 4.5 to help write some code for a project. Not perfect and you need to keep it on track when problems come up but I have accomplished working code in about 10% of the time it would have taken me by myself and it is fully documented and backed up on GitHub. I have had terrible results with AI trying to analyse contracts (It just makes stuff up based on my questions) but for coding it is very powerful and will only get better.

The photoshop/ zoom in examples above got me thinking. Is it possible with AI to process audio signals to act like filters for crossovers, sub integration, and room correction but without any time domain changes or any ringing?
 
I have been using Anthropic Claude 4.5 to help write some code for a project. Not perfect and you need to keep it on track when problems come up but I have accomplished working code in about 10% of the time it would have taken me by myself and it is fully documented and backed up on GitHub. I have had terrible results with AI trying to analyse contracts (It just makes stuff up based on my questions) but for coding it is very powerful and will only get better.

The photoshop/ zoom in examples above got me thinking. Is it possible with AI to process audio signals to act like filters for crossovers, sub integration, and room correction but without any time domain changes or any ringing?

Coding is aperfect use case or AI, because it is based on intent.

Declarative languages have existed for a while (check out UML), and the obvious next abstraction is human language with logic behind it... which is what prompt programming is perfect for. No surprises there. If I tell ChatGPT "write some code I can integrate into my website to provide directions to my business under Vist Us" it is a total no brainer. I have also had a totally contrarian experience, providing a template and requirements, and got complete garbage back. Don't rely on it yet.

I would not extrapolate that into areas that abandon logic and start drawing from that realm we call "intution", which humans can excel at... and AI completel fails at.
 
Google has introduced an 'agent' that can call local stores to ask if a product is in stock.
LoL!
I am not certain if this is TMI for even google or an agent.

At the August 2010 Techonomy Conference, Eric Schmidt (of google) said:
"If I look at enough of your messaging and your location, and use artificial intelligence we can predict where you are going to go."
Similarly -the same month- during a WSJ interview, Schmidt stated:
"I actually think most people don't want Google to answer their questions. They want Google to tell them what they should be doing next."
Fifteen years on, google's foundational ethos and invasive approach to user data and prediction must have necessarily manifested within Gemini LLM's underlying code.

Do you really want google to record that you just ordered 3 items from SexToys-R-Us.com and shipped them to your mistress' address?:eek:
 
LoL!
I am not certain if this is TMI for even google or an agent.

At the August 2010 Techonomy Conference, Eric Schmidt (of google) said:

Similarly -the same month- during a WSJ interview, Schmidt stated:

Fifteen years on, google's foundational ethos and invasive approach to user data and prediction must have necessarily manifested within Gemini LLM's underlying code.

Do you really want google to record that you just ordered 3 items from SexToys-R-Us.com and shipped them to your mistress' address?:eek:

Perhaps I am wrong to think this way.....

I take modest comfort from the notion that the interest of Google and the like in my spending habits will expire when I expire.
 
The unfurled URL, arguably, begs the question: Does Olympic-level figure skating need to be revolutionized?
I mean... I'd say... no...
and/or I'd wonder why it needs to be revolutionized?
EDIT: Seems to be there are better things to revolutionize... maybe I am too naive.
The skating itself? No--innovations there should be artistic.

But the judging? Judging in figure skating has been various levels of scandalous for the sport's entire history. I put the use of what people now seem to call "AI" in the same category there as the use of video monitors by (American) football referees, which many who would oppose AI officiating have agreed results in better calls. But if it's really AI, then I think @pablolie makes a good point--AI rewards the status quo, not the innovator.

Rick "upon further review..." Denney
 
I have been using Anthropic Claude 4.5 to help write some code for a project. Not perfect and you need to keep it on track when problems come up but I have accomplished working code in about 10% of the time it would have taken me by myself and it is fully documented and backed up on GitHub. I have had terrible results with AI trying to analyse contracts (It just makes stuff up based on my questions) but for coding it is very powerful and will only get better.

The photoshop/ zoom in examples above got me thinking. Is it possible with AI to process audio signals to act like filters for crossovers, sub integration, and room correction but without any time domain changes or any ringing?
The question I have when reading accounts like yours is this: You are qualified to evaluate Claude's imperfections and thus gain advantage from the efficiency it adds. But how did you become qualified? I submit that it is likely you are able to evaluate Claude precisely because you are asking Claude to do something you already know how to do yourself (through training and experience). Now: Put yourself in the shoes of brand new programmers just starting their first jobs. They may have a little training but very little experience in a production environment, and their only training may be in the use of the AI tool rather than in evaluating the AI product. How will they become qualified?

This is a huge issue in my view when considering the application of AI to many technical activities, especially those commonly used by amateurs and those that lack the syntactic rigor of, say, writing a computer program.

Rick "improvement is rarely continuous" Denney
 
The skating itself? No--innovations there should be artistic.

But the judging? Judging in figure skating has been various levels of scandalous for the sport's entire history. I put the use of what people now seem to call "AI" in the same category there as the use of video monitors by (American) football referees, which many who would oppose AI officiating have agreed results in better calls. But if it's really AI, then I think @pablolie makes a good point--AI rewards the status quo, not the innovator.

Rick "upon further review..." Denney
I never intended to dedicate years of my life to in-line skating. In-line skating just slowly escalated into an everyday and all day thing. The tiny nuances of edge control, body control/body English that changed my skating where more apparent after each of the previous nuances became part of my skating repertoire. I took up weight lifting, swimming and cycling all for the sole purpose of improving my in-line skating. I discovered by absolute fluke when studying a form of kung fu that has no kicking and required an hour of tai chi 3 times per week that tai chi was a critically important game changer for my in-line skating development. My entire life evolved to improve the skating experience. I had to learn about diet, the matters of hypertrophy and the difference in training that would bulk me up or tone me up, strength training versus fitness training and many other subjects to support the in-line skating development. I became so fit from my toes to my ears and so edgy about in-line skating that I looked like a cross between a professional boxer, a world class sprinter and a professional basketball player. It was known at the 11km park loop in Vancouver that the Vancouver Canucks professional hockey players chased me around the park. I was skating around the 11km park loop one day and a passenger coach bus was parked beside the 11km path, many young males being address by what appeared to be a leader and as I skated by I easily and clearly listened to them being pepped talked about me and my in-line skating, about my dedication and that I do it all without pay and they do it with pay etc. It was very life changing. Of course it was asphalt and concrete in-line skating and not figure skating but the investment, the fitness, the performance, the need, the commitment, the intention was all there and it was the vehicle in my life for everything going on for some years. There is no way that a computer could analyze that. There is no software or even sensors that can realize what is required in so many tiny microscopic details that are skating. Every single matter of skating is of critical importance at a high level of skill, athletic performance and the entertainment performance that is figure skating. Everything and I mean every single tiny microscopic micro-managing little thing that is required to skate at the high level is occurring from the tiny edges one stands upon all the way up through the bones and joints controlled by musculature that the overall majority of the entire population on Earth does not even know it has the ability to control with high accuracy and the fantastically absurd balance required to perform high level skating. After skating everyday and some days for up to 11 hours per day at the very high level of fitness that I was operating at I was walking down railroad tracks for miles at a time with no real effort required for balancing upon the railroad tracks. It was so absurdly easy, it required so little concentration and was effortless mile after mile. The only way these matters can be analyzed and judged is by fellow humans that understand everything about skating.
 
I never intended to dedicate years of my life to in-line skating. In-line skating just slowly escalated into an everyday and all day thing. The tiny nuances of edge control, body control/body English that changed my skating where more apparent after each of the previous nuances became part of my skating repertoire. I took up weight lifting, swimming and cycling all for the sole purpose of improving my in-line skating. I discovered by absolute fluke when studying a form of kung fu that has no kicking and required an hour of tai chi 3 times per week that tai chi was a critically important game changer for my in-line skating development. My entire life evolved to improve the skating experience. I had to learn about diet, the matters of hypertrophy and the difference in training that would bulk me up or tone me up, strength training versus fitness training and many other subjects to support the in-line skating development. I became so fit from my toes to my ears and so edgy about in-line skating that I looked like a cross between a professional boxer, a world class sprinter and a professional basketball player. It was known at the 11km park loop in Vancouver that the Vancouver Canucks professional hockey players chased me around the park. I was skating around the 11km park loop one day and a passenger coach bus was parked beside the 11km path, many young males being address by what appeared to be a leader and as I skated by I easily and clearly listened to them being pepped talked about me and my in-line skating, about my dedication and that I do it all without pay and they do it with pay etc. It was very life changing. Of course it was asphalt and concrete in-line skating and not figure skating but the investment, the fitness, the performance, the need, the commitment, the intention was all there and it was the vehicle in my life for everything going on for some years. There is no way that a computer could analyze that. There is no software or even sensors that can realize what is required in so many tiny microscopic details that are skating. Every single matter of skating is of critical importance at a high level of skill, athletic performance and the entertainment performance that is figure skating. Everything and I mean every single tiny microscopic micro-managing little thing that is required to skate at the high level is occurring from the tiny edges one stands upon all the way up through the bones and joints controlled by musculature that the overall majority of the entire population on Earth does not even know it has the ability to control with high accuracy and the fantastically absurd balance required to perform high level skating. After skating everyday and some days for up to 11 hours per day at the very high level of fitness that I was operating at I was walking down railroad tracks for miles at a time with no real effort required for balancing upon the railroad tracks. It was so absurdly easy, it required so little concentration and was effortless mile after mile. The only way these matters can be analyzed and judged is by fellow humans that understand everything about skating.
I get all that--but figure skating is an artistic sport that is analyzed by how the body moves externally, not by how the body feels internally. Can that be analyzed? I suspect so. Maybe not now--I don't know--but eventually. I doubt AI could judge the art because it is art, but judging the moves for accuracy and correctness is just a matter of technology. I hear the commentators making specific mechanical observations all the time--he didn't quite rotate enough, she put weight on the off skate--stuff I don't understand but that they observe only with their eyes.

I'm not saying I like it, but the appeal of it seems clear to me.

I am saying that figure skating as a sport has been judged in very inconsistent and scandalous ways at various times over many years. They have completely altered the judging methodologies several times trying to address those issues, but still it remains as an issue that pops up from time to time. I can see why using an AI-driven technical evaluation would be so seductive.

But that's the issue with so many AI use cases--they are seductive because of the problems they solve, making it tempting to ignore or minimize the problems they create.

Rick "admiring anyone who can remain upright gracefully when good balance is required" Denney
 
Some companies like Amazon and Airbnb restrict online bots being used on their websites for reinforcement learning.

The NY Times reports start-up companies are creating digital clones of popular websites to use for reinforcement learning.

 
I often enjoy coming across new words that somehow creep into American-English.
Latest one is AI "alignment" and its research thereof.;)
 
For better or for worse:

IMG_1891.jpeg

Today I asked AI to identify some snow footprints from my patio.

Human tracks for scale. Can you identify the animal tracks?

Those are classic red fox tracks in snow.Here’s why I’m 100% confident it’s a fox (and not a coyote, dog, or cat perfectly straight, almost single-file line – foxes are famous for placing their hind feet almost exactly in the holes made by the front feet when trotting (very energy-efficient in snow).
Print size is about right for an adult red fox: roughly 1.8–2.3 inches long (your boot prints give great scale).
Oval/elongated shape with clear claw marks at the tips.
Four toes, nicely symmetrical, with the two middle toes close together and the outer toes a little farther back – textbook canid pattern, but much neater and smaller than a coyote or dog.
The stride (distance between each set) and the narrow trail width also match a trotting fox perfectly.

So, definitely a red fox moving at a steady trot across the snow. Very cool find!
 
For better or for worse: Today I asked AI to identify some snow footprints from my patio. ...

"Those are classic red fox tracks in snow.Here’s why I’m 100% confident it’s a fox (and not a coyote, dog, or cat perfectly straight, almost single-file line – foxes are famous for placing their hind feet almost exactly in the holes made by the front feet when trotting (very energy-efficient in snow).
Print size is about right for an adult red fox: roughly 1.8–2.3 inches long (your boot prints give great scale).
Oval/elongated shape with clear claw marks at the tips.
Four toes, nicely symmetrical, with the two middle toes close together and the outer toes a little farther back – textbook canid pattern, but much neater and smaller than a coyote or dog.
The stride (distance between each set) and the narrow trail width also match a trotting fox perfectly.

So, definitely a red fox moving at a steady trot across the snow. Very cool find!

Yes, I don't know if the presentation errors in the text are the original output or copy-paste-post things: missing space after initial period, incorrectly formed m-dash, returns after every sentence, then an extra line break to separate the conclusion. Nobody wants to see that sort of thing presented systematically.

But all that pales before the gratuitous language flourishes attempting to convey casual human conversation. I'd need anti-nausea pills. This quote from Dave Winer's Scripting News is about that stuff on an unfortunate Alexa product, but it could equally be deployed as a cardinal rule:

The new Amazon Alexa with AI has the same basic problem of all AI bots, it acts as if it's human, with a level of intimacy that you really don't want to think about, because Alexa is in your house, with you, listening, all the time. Calling attention to an idea that there's a psuedo-human spying on you is bad. Alexa depends on the opposite impression, that it's just a computer. I think AI's should give up the pretense that they're human, and this one should be first.
 
I don't know if the presentation errors in the text are the original output or copy-paste-post things: missing space after initial period, incorrectly formed m-dash, returns after every sentence, then an extra line break to separate the conclusion. Nobody wants to see that sort of thing presented systematically.

Most AI chat interfaces output markdown by default. Since there's really no official standard for markdown (or rather several competing 'flavours') it's pretty common to see double linebreaks converted to none or the reverse, depending on where from or where to the text has been copied. Not really the bot's fault. The 'incorrectly formed m-dash' is a correctly formed n-dash which is pretty much universal outside the US (and much better looking IMO) and often recommended in modern style guides (e.g Bringhurst).

I agree about the 'chattiness'. It gets stale very quickly but many people seem to like it. It can be mitigated somewhat with a suitable system prompt.
 
The 'incorrectly formed m-dash' is a correctly formed n-dash which is pretty much universal outside the US (and much better looking IMO) and often recommended in modern style guides (e.g Bringhurst).

Amen to that.
 
Most AI chat interfaces output markdown by default. Since there's really no official standard for markdown (or rather several competing 'flavours') it's pretty common to see double linebreaks converted to none or the reverse, depending on where from or where to the text has been copied. Not really the bot's fault. The 'incorrectly formed m-dash' is a correctly formed n-dash which is pretty much universal outside the US (and much better looking IMO) and often recommended in modern style guides (e.g Bringhurst).

Good points. I've re-used exactly zero output from the chatbots, and entered exactly one prompt into a DDG search box, so I'm not fluent—cut+paste from PDF in ancient times delivered similar horrors, so I get the idea. But unless @Petrushka was aiming for a warts-and-all effect, surely you'd clean up the worst of the slop before presenting it to others?

As for dashes em and en, my hybrid education (high school in both US and Oz) sometimes leaves me betwixt and between. I'm almost horrified that I prefer the US habit.

I agree about the 'chattiness'. It gets stale very quickly but many people seem to like it. It can be mitigated somewhat with a suitable system prompt.

Yeah I really do prefer my computers to be computer-y. The uncanny valley is a disturbing place. Full verisimilitude will be worse of course.
 
Last edited:
Back
Top Bottom