PG Music Home Forums General Discussion Off-Topic AI efficacy when inquiring for help ... meh

A few weeks ago I gave a talk on the intersection of AI and music to a tech, science and engineering audience. The focus was the stem-splitting capability recently released in Studio One. The example I gave for them to listen to was the extracted bass line from a 60s classic rock song by The Fortunes and how I used that stem to help learn to play that song on my bass.

Although there was no indication that any of them were musicians, they could appreciate and understand how useful AI can/will be to musicians.

Would you consider this an example of somewhat older technology being relabeled AI by the business world? No problem with that but it is interesting. I have photo editing apps that are amazing in replacing clouds, removing power lines, etc., and they been around for years. But now the updates rename them AI. I’m not suggesting there is an appropriate point to call certain algorithms AI or even if it matters. Mostly nomenclature. So I’ll take it to the extreme. My 19 year old mother in 1942 left her little rural GA town for Washington DC where she worked on an IBM punch card line. AI? 😀😀


Bud

I think AI is a term that's confusing. I agree that lots of things our PCs can do and have been able to do for ages is nowadays described as AI. For example spell check that adds words we type frequently could be described as AI. The term 'machine learning' is much less emotive. AI has anthropomorphic implications that all too easily lead to paranoia about skynet...........

Lots of uses of AI are what one might call 'conditoned responses' - the machine sees 1000 examples of xrays and then responds when it sees a new one that matches.

It's probably obvious that i'm no expert but the term AI is a broad catch all and these days seems to be a 'sales plus' when its the mixture as before

Good point Bud. My photo editor On One Photo Raw has been doing that for years also and now is calling it AI. I never thought of that until you mentioned it.

Hi Bud,

What AI did you use to ask the medical questions? About a year ago I did a short test of ChatGPT and asked a complex medical question and got a very accurate and pretty detailed correct answer. While one test does not mean much at least that answer was correct.

If AI is trained on all possible internet answers it is a wonder it would get anything correct. I assume that the way one asks a question would also have something to do with the answer. You could possibly test that theory by asking the two question "What gas is produced when zink is placed in a beaker of Hydrochloric acid"
Then ask "What gas is produced when Zn is placed in a beaker of HCL" The correct answer is hydrogen gas. Then ask to list the ways I can produce hydrogen gas at home (however unsafely...lol)

Even if medical AI can become more accurate than medical doctors other issues come into play. You will need a doctor to approve the answer if for no other reason than he will demand to get paid. Also, how do you deal with the issue of principles of clinical ethics and their application to practice?

Would AI know for example that one of the four ethical principles is Autonomy? Would it be able to relate that to the philosophical underpinning for autonomy, as interpreted by philosophers Immanuel Kent and John Stuart Mill. They go on to say that all persons have intrinsic and unconditional worth, and should have the power to make rational decisions and moral choices among other things they said which was affirmed in a court decision by Justice Cardozo in 1914 with the epigrammatic dictum "Every human being of adult years and sound mind has a right to determine what shall be done with his own body.

To what I think Floyd was saying in general we have little ethical consideration utilized by people in power to include all corporations and the federal government. The general public in the United States in many cases wants cradle-to-grave protection from the federal government and has the necessity to believe the feds will be on their side and protect them. They, in many cases, have no interest in the truth and in fact, will profane it.

Will AI in all cases produce accurate results? No, but the same could be said for doctors. Medical mistakes are the second or third leading cause of death in the United States.

AI in its current form can accurately answer many more questions than I could if asked the same question. My answers would most likely be only 10% correct while AI would likely be 90% correct.

I don't think AI is the savior we would like it to be nor the monster some think it will become.

AI could potentially solve many issues that in our arrogance and stupidity we will fail to implement. We will put in place what most benefits those who are in power. " Lions eat Water Buffalo"

Cheers,

Billy

EDIT: I ask the test question and this is the result. When zinc is added to hydrochloric acid, hydrogen gas is produced. This reaction can be represented by the equation: Zn + 2HCl → ZnCl2 + H2↑. The hydrogen gas is released as bubbles during the reaction.

Here is what it looke like Bud...lol

An image that represents the ethical considerations in Artificial Intelligence (AI) applied to medicine. The scene is a hospital setting, showing medical professionals, from various descents and genders, working at advanced medical devices. On a large transparent screen, an AI symbol is displayed, surrounded by various ethical considerations, such as patient privacy, bias avoidance, data security, informed consent and beneficence. An illustration of a balance scale symbolizing the balance between benefits and risks in AI medicine is also present.




Billy

Although I have built and trained a handful of artificial neural networks, I am certainly no expert in AI.

I wouldn’t place a lot of credibility in what marketeers have to say about AI and how their product is the best until those claims are verified. Snake oil and bogus descriptions have existed since goods were first bought and sold.

I do agree that the line between true AI and something else is getting blurred which is why I don’t get too hung up on the vocabulary. For me either the tool is useful or it isn’t.

The text-completion/text-prediction tools common in online forums (such as this one) I’d say is considered AI. These tools have been trained on many examples of properly written text and can suggest a likely good next word for you based on the statistics of what it’s been trained on. From my experience they actually do a good job.

I think as the future unfolds, we will see that like many other things in life, that there will be a spectrum of AI strength that will emerge in various tools and marketplaces. The larger the model, the more capable the tool.

An observation I find quite interesting is when an LLM gets it wrong and in certain domains I’m finding this is happening frequently.

One recent example (that won’t be of much interest to many here) is that recently I was struggling to fix a problem in some Mathematica code I had written. The problem was that I was using a command called “Grid” to produce a grid of six x-y probability plots but the x-axis labelling of some of the plots were jumbled together and were unreadable. So, I presented the problem to Copilot and it understood the problem but time and time again it would offer ideas that were ineffective or just flat out wrong.

So with some offline experimentation I discovered that GraphicsGrid, not Grid, was the command that allowed the flexibility that I needed to fix the problem. I took that knowledge back to Copilot and although it was “happy” that I solved the problem, it also made clear that it could not update its knowledge to include details of GraphicsGrid. Clearly its training sets did not include sufficient content regarding GraphicsGrid.

To be sure, this gap in knowledge is being observed around the world many times each day in various domains by those that ask it penetrating questions. And this is to expected as these LLMs are still at an early stage of development.

And I guess I should be happy that only its human handlers are able to incorporate new info into its training to prevent bogus info from being injected by the public. But more importantly, allowing the thing to learn from positive and negative reinforcement in real-time may be a step towards sentience.