Loopy Pro: Create music, your way.
What is Loopy Pro? — Loopy Pro is a powerful, flexible, and intuitive live looper, sampler, clip launcher and DAW for iPhone and iPad. At its core, it allows you to record and layer sounds in real-time to create complex musical arrangements. But it doesn’t stop there—Loopy Pro offers advanced tools to customize your workflow, build dynamic performance setups, and create a seamless connection between instruments, effects, and external gear.
Use it for live looping, sequencing, arranging, mixing, and much more. Whether you're a live performer, a producer, or just experimenting with sound, Loopy Pro helps you take control of your creative process.
Download on the App StoreLoopy Pro is your all-in-one musical toolkit. Try it for free today.
Comments
It is different, yes. It simply remembers your entire conversation history. You can ask it anything about what you previously discussed and it should be able to recall it, analyse it etc. It also now has a feature where it will kind of create its own custom instructions based on things you say. For example, if you casually mention that you're a software dev, you might see a message popping up after that which says something like 'memorising'. It is marking that info as significant, to improve its future answers' relevance. But... It still tends to waffle on, answers are often still prone to hallucination. Far from perfect, still useful in skilled hands tho!
Don’t give them too many ideas for hybridisation.
I’d say we have to sort out ‘human rights’ before anything else, certainly before the rights of a mechanical mannequin.
Right on.
I think you covered it. Assumptions are unavoidable. But as you say we should all strive to be kinder by considering what we say and exercizing some self control when typing away. I try to be better but it's small steps.
BTW, I think it was established in the 1986 documentary "Christine" where Will Darnell explained so well:
I understand what you said, but it did create a story to fill in the void when it cannot find the answer.
I asked it the following question: "Provide a detailed analysis of the third movement of Bartok's String Quartet No.4." I did the analysis myself (but never published it.) I searched the web for a good analysis of this movement, but couldn't find one that's good enough—basically, no real "detailed" analysis exists, thus the question. FYI the movement is tempo marked as "Non troppo lento" (slow, but not too much), and begins with a slow E Pentatonic scale descending passage, alternately played by the 2 violins and viola. All notes were held to form a sustained, static chord accompanying a melody played by the cello.
And here's part of the answer:
"...The first section, marked "Allegro" (fast and lively), begins with a lively and energetic melody in the first violin, which is accompanied by rapid arpeggios in the other three instruments..."
No human errors, no matter how amateur he/she is, would rival this. Again, I agree with you completely that it's up to us to check whether ChatGPT gets its answer from reliable sources, etc., but in this case, I don't believe any human sources for the analysis could be this wrong. It looks more like it cannot find the analysis of this specific movement of the quartet, and so pulled other analyses of other quartets to create the answer instead.
It’s almost like chatGPT makes some assumptions.
…but you can roll it in glitter
Would you put on a good show for the reflection in the mirror, or rather the origination of that image?
There is no need for a forum war. I will extend an invitation to any particular individual(s) to discuss my assumptions in a PM
My dad died last year and I had a screen shot of his medical prescription. I wanted to find out what meds where associated with a heart condition so I posted the screen shot and asked chatgpt what meds were heart meds.
It gave me a detailed synopsis of each one and gave a summary of only the ones that were associated with a heart condition. It did it in seconds. Would have taken me a fair bit of time to pull that information myself. It's really quite amazing technology.
Implicit in a lot of this thread is a notion that LLMs (of which ChatGPT is one) are designed for fact/truth discrimination. They are not; it is not what they are designed to do. They aren’t “intelligent “ in in the sense of being designed to analyze information for truth. They are designed to generate language consistent with the corpus they were trained on.
They are essentially predictive text engines trained on an ENORMOUSLY (really really really enormous) LARGE amount of source material. If the corpus they are trained on has any bad information in it, that information will make its way into what it returns.
LLMs are very good at generating text that SOUNDS accurate—and for the average person, the quality of the sentences will be better than what they might write themselves in terms of style. But they often spit out convincing sentences that are factually wrong.
I have a few friends that find it useful for programming because the corpus seems to have sufficient material that it supplies reasonably relevant code —because these friends are expert coders, they quickly recognize when it gives them bad code. A couple of friends, also expert coders, work in areas for which the corpus must not have much relevant code, because they have found it not very useful except for code they don’t need help with.
Little discussed is how much benefit these systems are compared to the energy they consume (lots) or the ethics of companies making profit that is 100% reliant on other people’s work (the corpus on which these systems train).
It seems that there's also, implicit in your making this point, a suggestion that humans have some advantage over AI because humans have some innate superiority at identifying "truth". I would suggest that this is not an advantage humans have over LLM-AI. Or, if humans do have some advantage, far more is required to establish that than simply to say LLM-AIs "are not designed for fact/truth discrimination." Humans have evolved to adopt beliefs that maximize fitness, not truth. See, e.g, https://www.scientificamerican.com/article/did-humans-evolve-to-see-things-as-they-really-are/
LLM's simply are not designed to do this -- any attempt to add 'truth discrimination' is essentially a hack. If you read the writing about LLMs from people that are both hugely knowledgeable AND have no vested interested (i.e. no profit motive) they have a lot of enlightening things to say about what this technology can and can't do -- even with refinement. Jaron Lanier has written some really good pieces going through this -- there are a lot of technologists that have a vested interest in selling LLMs as delivering more than they do/can -- because they have a huge profit motive.
Humans with expertise in a field are by no means infallible -- but they are able to identify errors in a way that LLM's cannot. There are certainly areas where various types of AIs are less fallible than individual humans.
When one switches freely between discussing AI and LLMs (a very particular, if impressive, application of machine learning/AI), it can be confusing. LLM's are a particular category of tool with particular limitations. Other AI tools have other applications and limitations. I think it is important not to treat LLMs as AI writ large. LLMs are amazing at what they were designed to do -- but they are not designed for the kind of analysis that experts in a field do. They just aren't. I don't mean "they aren't there yet", I mean that isn't what that tool does.
The next big push relative to the advance in A.I is reasoning and eliminating hallucinations which I believe OpenAI are currently working on. The goal is to establish A.I as being reliable and trustworthy.
AI has just started. Not sure what you are referring to relative to a burst.
What an exemplary post with all reference articles, love that part of it and wish more people here (and elsewhere) would link to their claims. Not sure I agree with your take on it but that's another story.
The Goldman Sachs article is over a year old and well out of date.
LLMs have come a long way since then and advances show no sign of slowing up. Quite the contrary.
I've been following advances very closely. Huge amounts of money and effort are getting plowed into this. A.I is going to be deeply imbedded into every aspect of our lives if we want it or not.
Thanks for the article. In conclusion it does say that A.I will pay off but at the moment it's constrained by GPU availability. There's a bit in there that states (conservatively) that in 10 years 25% of human jobs will be replaced. That's quite a decent return of investment. I think it will be quicker than that.
Don't get me wrong. A.I is going to be the biggest disrupter and impact to humans than any other technology. I'm not hugely optimistic that we will handle the transition well.
But no. The bubble isn't about to burst. There is no bubble. It's only unrelenting progress. Quite frightening really.