I find ChatGPT absolutely fascinating and at times it can be even very useful in your workflow.
The biggest issue I have with it, especially when it comes to science and engineering, is that it refuses to give any resources or references from were it learned from.
Which already is almost a direct no-no since it's impossible to fact check the answers.
The danger here is that many people WILL take that for granted, while it's absolutely essential.
With the risk of creating all sorts of false information. Even more so when the AI algorithm will get unintentionally biased by mistake.
Or will just simply make mistakes in general.
The biggest issue I have with it, especially when it comes to science and engineering, is that it refuses to give any resources or references from were it learned from.
Which already is almost a direct no-no since it's impossible to fact check the answers.
The danger here is that many people WILL take that for granted, while it's absolutely essential.
With the risk of creating all sorts of false information. Even more so when the AI algorithm will get unintentionally biased by mistake.
Or will just simply make mistakes in general.
I suspect Chat GPT was very well trained in one area and then they discovered it works in all areas. Am sure they are heavily crawling the web right now and their hardware is screaming for mercy as their engineers are sleeping less tweaking the design. Also it might matter whether your using the basic or the premium. It matters how it interprets the input as it needs converters for technical drawings, electrical drawings, architectural drawings and the related math. With the initial public release of Chat GPT the possibilities are endless despite those under estimating this technology, one shouldn't take it too lightly. Chat GPT is just one of the doorways to something major. Am pretty sure what were looking at is a zygote and not even an embryo.
How do humans learn? Are we truly capable of coming up with something quite novel. Learning is training the brain to process what's already there. We build upon each other. Chat GPT and other technologies are quite capable of hyper learning and evolve at a speed possibly not possible for humans and in the process probably upgrade us. We operate around the electromagnetic spectrum to analyze and interact with the world around us but we do it in a lazy way.
Last edited:
Microsoft is looking for ways to rein in Bing AI chatbot after troubling responses
https://www.cnn.com/2023/02/17/tech/microsoft-bing-ai-changes/index.html
ChatGPT is lovely. Fun for the novice amplifier designer. But: it cannot replace decades of amplifier permutations and following listening sessions. Why? Because what amplifier designers like Nelson Pass, John Curl and Bob Cordell have learned across 3-4 decades is in their heads. Their fingers. Their spines. And therefore never been fully expressed in books and blog posts that can be used as a training set for the neural network that feeds AI. Limited info in = limited deductions out. I will listen to design elaborations and insights from Nelson, John and Bob any day. Screw ChatGPT! I am talking craftmanship here. Not n-dimensional pattern recognition and correlation. Audio design is art - not pablum. 😎
What I mostly really like about ChatGPT, is that you can ask a question and it sticks to the subject as objectively as it can (read: trained for).
Which is very refreshing compared to human beings. Because in that case you get a whole pile of subjective BS, believes and politics on top of it. Including "opinions" you never asked for.
I find this particularly refreshing in the world of programming. Where I just want to know how and why things are being done. Instead of opening a whole war about people's subjective beliefs about a certain approach.
Or even worse, that people automatically assume you're a student, have all the time and interest in world and therefore being sent into the woods "because you have to learn"......
I just want my freaking thing to work...
Which is very refreshing compared to human beings. Because in that case you get a whole pile of subjective BS, believes and politics on top of it. Including "opinions" you never asked for.
I find this particularly refreshing in the world of programming. Where I just want to know how and why things are being done. Instead of opening a whole war about people's subjective beliefs about a certain approach.
Or even worse, that people automatically assume you're a student, have all the time and interest in world and therefore being sent into the woods "because you have to learn"......
I just want my freaking thing to work...
"Never send to ask for whom the bell tolls - it tolls for thee".
Thanks to "researchers" including but not limited to physics, biology, viruses, psychology who have no slightest idea where it would lead to
neither having any reflection and/or philosophy or religious background nor ever taken accountable for their actions.
There is a hope for humanity no more.
My thoughts align with this.
No use trying to convince those who don't get it (they can't) or who are scared of it (they don't want to feel even more scared).
I found it quite interesting that it was Google researchers who wrote the first paper on the Transform architecture. The premise was that this architecture already fared better than RNN and CNN on some language tasks. However, Google let it lie low for a while.
In the meantime, OpenAI continued work on it, turned it into a service and product with the results that we know.
After this, MS does a partnership with them and even buys a chunky stake in them and integrates it in Bing.
Google scrambles to make what could be perceived as an equivalent announcement and a demo in Paris, but that event was an utter failure, a catastrophe that even had an effect on Google's stock.
So Google Research Tech -> now in MS Search.
ChatGPT-enhanced Bing has already:
- Made some threats to a human because the human insists it is wrong about dates. However, no need for AI here: MS makes threats to humans on its own
- Displayed an 'existential crisis' or 'panic' mode because it lied and then was told by a human that it can't remember past conversations. This was amazing to read (the convo itself).
That this is essentially purely a statistical modeling of high-probability completion of sentences and follow-up sentences is mind-boggling.
So now Google has to play catch up, on tech research they publicly shared themselves and let sleep until enhanced versions of it found their way into a competitor's service/product.
Right now, Bing > Bard.
Last edited:
When I tried ChatGPT it knew it couldn't learn from experience. Maybe enough for one session, but not longer. Its not to hard to bump up against its other limitations either.
Regarding people using it to avoid homework or avoid understanding something well in order to do it, the same sort of concerns have arisen with hand held calculators, PCs, Matlab, Google, etc. People aren't going to spend time to learn slide rules just in case the power goes out sometime. Instead people will learn Googlefoo, Matlab, etc.
Regarding people using it to avoid homework or avoid understanding something well in order to do it, the same sort of concerns have arisen with hand held calculators, PCs, Matlab, Google, etc. People aren't going to spend time to learn slide rules just in case the power goes out sometime. Instead people will learn Googlefoo, Matlab, etc.
Nothing wrong with new tools. We have had oscilloscopes, tone generators and distortion analyzer for a while... but we still cannot explain why a 1930's-style single ended triode 300B amp sounds better than a modern class D amp. Or why a Stradivarius sounds better than a modern violin... There is magic here. Something we don't understand - but we do hear and feel.
Tools are great at providing leverage to human activity. A shovel is useful when planting a tree when the alternative is digging a suitable hole by using one's bare hands. ChatGPT is certainly that - a tool.
Humans experience the magic. 🧙♂️
Tools are great at providing leverage to human activity. A shovel is useful when planting a tree when the alternative is digging a suitable hole by using one's bare hands. ChatGPT is certainly that - a tool.
Humans experience the magic. 🧙♂️
Not magic exactly. What we don't understand is how to measure close enough to like we hear, is all. Maybe interesting to run that idea by ChatGPT?
ChatGPT would probably end up being banned from diyAudio. At least I hope so.
Just had a little talk with ChatGPT, with the following excerpt:
Is psychoacoustics entirely consistent with auditory scene analysis or are there specific disagreements about human perception? Specifically I am interested in clear disagreements. Are you saying there are no clear, specific disagreements as to the validity of research findings?
"There are some areas where the findings of psychoacoustics and auditory scene analysis research may not be entirely consistent, and where there may be ongoing debates or disagreements. For example, there may be differences in the way that researchers define and measure certain perceptual phenomena, or in the emphasis they place on different aspects of auditory perception. Additionally, there may be differences in the specific experimental paradigms or stimuli used, which could influence the results obtained. However, it is important to note that both fields are highly interdisciplinary, and researchers in both areas often draw on a wide range of methods and concepts to better understand human perception of sound. Overall, while there may be some areas of disagreement or ongoing debate, the findings of psychoacoustics and auditory scene analysis research are broadly consistent in their characterization of human auditory perception."
So you sort of know in general way that there have been disagreements, but you don't know the specifics? Is that correct?
"Yes, that is correct. While I am aware that there have been debates and differences in the findings and methods of psychoacoustics and ASA, I do not have access to all the specific details and the nature of those disagreements. The research in these fields is complex and evolving, and it is not uncommon for different groups to have different interpretations or conclusions based on their data and methodology."
...IOW, there is a lot it doesn't know if you get down to a certain level of detail 🙂
Also, it tends the ramble and repeat itself when it doesn't have a clear answer.
Is psychoacoustics entirely consistent with auditory scene analysis or are there specific disagreements about human perception? Specifically I am interested in clear disagreements. Are you saying there are no clear, specific disagreements as to the validity of research findings?
"There are some areas where the findings of psychoacoustics and auditory scene analysis research may not be entirely consistent, and where there may be ongoing debates or disagreements. For example, there may be differences in the way that researchers define and measure certain perceptual phenomena, or in the emphasis they place on different aspects of auditory perception. Additionally, there may be differences in the specific experimental paradigms or stimuli used, which could influence the results obtained. However, it is important to note that both fields are highly interdisciplinary, and researchers in both areas often draw on a wide range of methods and concepts to better understand human perception of sound. Overall, while there may be some areas of disagreement or ongoing debate, the findings of psychoacoustics and auditory scene analysis research are broadly consistent in their characterization of human auditory perception."
So you sort of know in general way that there have been disagreements, but you don't know the specifics? Is that correct?
"Yes, that is correct. While I am aware that there have been debates and differences in the findings and methods of psychoacoustics and ASA, I do not have access to all the specific details and the nature of those disagreements. The research in these fields is complex and evolving, and it is not uncommon for different groups to have different interpretations or conclusions based on their data and methodology."
...IOW, there is a lot it doesn't know if you get down to a certain level of detail 🙂
Also, it tends the ramble and repeat itself when it doesn't have a clear answer.
In the field of artificial intelligence, some seek to create machines that can perform human-like tasks, while others aim to develop machines that can perform tasks that humans cannot do yet. The latter group is focused on a bold vision of the future that involves quantum computing, GPU clusters, and a dramatic leap forward in human evolution.
The goal is not merely to make AI as good as the experts, but to make it as good as the experts 1000 years from now using quantum computing or GPU clusters, to catalyze a leap in evolution. This could enable us to have access to technologies like teleportation among other advancements that are yet to be developed. Ultimately, this future may be slightly different from what it would have been without the AI catalyst.
The implications of this vision are vast. If we can develop machines that are as intelligent as humans 1000 years from now, what else might we be able to achieve? Could we eradicate diseases, eliminate poverty, or even venture into outer space?
However, some skeptics doubt that this vision is attainable. They argue that quantum computing and GPU clusters are still in their early stages of development and that it could take several years, if not decades, to unlock their full potential. Additionally, some people worry that creating machines that are far more intelligent than humans could be hazardous, resulting in unintended consequences and even existential threats.
Despite these concerns, those who support this vision remain undeterred. They believe that the possible benefits are too significant to disregard.
"AI is not just a tool that makes our lives a little more manageable," they say. "It is the key to unlocking a future that goes beyond our wildest dreams."
Only time will tell whether this vision will materialize, but one thing is certain: the future of AI is one of the most exciting and transformative areas of research in the world today.
I don't know what universe you're living in, but the one I am living in, all of those differences are perfectly understood.
For quite some decades actually.
If you were mentioning black holes and tunneling effects instead, I would totally get the word magic.
Because from a physics point of view these things are kinda bizarre.
Not according to Dr. Earl Geddes.
A quote:
"The bottom line here is that we know so little about how humans perceive the sound quality of an audio system, and in particular the loudspeaker, that one should question almost everything that we think we know about measuring it. From what we have found most of what is being done in this regard is naive."
http://www.gedlee.com/Papers/Comments on howard.pdf
http://www.gedlee.com/Papers/papers.aspx