Every day I hear about the term AI. It's been bandied about in radio (replacing DJ's), industry (replacing people), heck, it's even been called a threat to the human race.
Most people don't even know what AI really is. And guess what. I'm going to tell you.
AI isn't a thing, per say. There's no intelligence behind AI. It's really a term, and a very misleading term, for a technique to program computers.
The computer industry has always been more full of hype (I'd use another term but this is a family site) than substance. AI is a prime example of that. Quantum computers is another (they don't even exist yet outside the laboratory and they're bandied about routinely as well), but I won't go into that now.
When I first got started in the computer industry years ago (the early 1970s, while attending university), all programming languages were procedural. When you developed a program, you created process flows (in something called a flowchart) before the actual programming. These process flows might look something like: the user types in the bond price and yield, these are validated according to some rules, and then the validated numbers are sent to the calculation routine.
There are lots of ways to write good code using a procedural language. There are many more bad ways, and you end up with what is called 'spaghetti code', which is a big mish mash and hard to maintain.
That was the impetus for the development of object-oriented programming languages. It takes the paradigm, or technique, of reducing program elements to, guess what, objects, which maps easily to the real world in a lot of instances. In our example, we'd have a Bond object, which had properties such as a price and yield, and various methods to calculate things.
That worked well for a while, but it turns out that you can write just as bad code using this paradigm as you can in a procedural language. You can have not enough objects (so that they become overloaded with methods and properties) or too many (so that you get 'spaghetti objects'). You may not even have the right objects. Some problems don't lend themselves to the object-oriented approach, either.
But all that is OK. Because there is no real magic with the object-oriented approach. Good programmers using procedural languages tended to write in an object-like fashion in any event, but without the artificial constraints placed on them by the 'object-oriented environment'. Even one of the first object languages C++ was just a preprocessor that translated the objects in C++ to the procedural C language.
All AI is is another paradigm, another way of looking how to solve particular problems. It was originally called machine learning, but even that doesn't accurately describe what is happening behind the scenes, as no real learning takes place. And as I said at the top of this (probably overly long) post, there's no real intelligence behind it either - programs are just doing what they were programmed to do.
This paradigm has been around for over 30 years (I worked in the field in the early 90s), with each year its proponents proclaiming that amazing things were just around the corner. And only now is the paradigm just starting to be widely adopted.
The 'AI' approach is best suited for those problems that don't have an exact solution, and so you have to 'train' the program for an appropriate answer. Take, for example, speech. Each individual has a unique speech pattern, and a one-size fits all approach to matching up what a program thinks speech should should sound like vs actual would not work. So the program has to be trained to recognize each individuals speech patterns.
How it does that is immaterial. Most speech engines use what are called neural nets or computational nodes, each of which adjust the parameters of built-in phonemes to actual speech.
However, the same program could be written using an object-oriented language and/or a procedural language. Instead of neural nets, the programs would just modify internal data to match phonemes. Or they could even generate code. There are lots of ways to do it. Some ways could even be easier than the AI approach, particularly for old code hacks.
But the bottom line is, AI doesn't stand for intelligence. There is no thinking as we know it (and we don't even know what thinking is, or how it works, not really). A speech program can't drive a car, for example. You would need a program specializing in that for a self-driving Tesla. These programs can only operate within the world and the constraints that they were designed for.
You can even write programs to go out onto the Internet, gather up information and regurgitate it to an amazed public. And then call it sentient.
Beware the hype. And everything you read from search engine results.
I am a human person who agrees with Artisan Radio, who is also human and not a radio, regardless of his online name.
On a radio program about AI I heard Sam Altman, CEO of Open AI, sincerely admit that, in his view, AI can be used against people by the people programming it, but AI within itself will not 'evolve' into a danger against humanity. But that means that the military, law enforcement, government and corporate powers, as well as lone-operators, being human, may possibly weaponize AI against the best interests of the human population. But AI, in and of itself alone, is not plotting against people because it is absolutely unaware of our existence unless some person tells AI about humans, and why would that happen?
As radio station owner/operators we can fiddle with AI announcers, but no AI bot will ever own or operate a part 15 station.
I can't see AI techniques being all that useful for radio, at least in the DJ field.
Speech generation is largely a mechanical process, the quality dependent on the amount of time (and bits) used to encode sounds. Most speech engines use punctuation (in text to speech) for inflection.
Human speech and inflection is influenced by the intention of the words spoken, and I could potentially see an AI approach being useful for that.
A simple example would be for the sentence: "The window is open."
If spoken in an even, measured, tone, it likely just refers to the fact that a window is indeed open. Spoken loudly, with emphasis on the last word could mean that the window is open, and you want someone to close it.
AI techniques can be used in speech recognition, where you have to determine the meaning or intent of a set of spoken words from a unique individual.
A truly natural speech generation engine would incorporate meaning into the inflection of the words spoken. Needless to say, I have yet to actually see examples of that.
In one way, streaming radio programs over the internet, can be done free of charge using open-source software on a home computer. The glaring difference between doing this and simply broadcasting a part 15 low power signal to nearby radios is that the internet reaches everywhere on earth whereas part 15 radio only reaches about as far as you can shout. But in another way streaming on the internet becomes stunningly expensive if programming copyrighted music for which very costly license fees must be paid. But there are many part 15 operators who consider the license fees worth paying for the privilege of programming favorite music.
In the case of KDX-OGG, streaming from an Icecast Server, no licensed music is used and only talk programming allowed by permission of the producers is scheduled. KDX-OGG can be heard from audio players available on our website, or two online directories which also list many other streaming stations.
Online streaming is scaled after standard radio, wherein a program plays at a scheduled time and a listener starts hearing from whatever point is reached at the time he starts listening. There is no rewind or fast-forward, and the next program doesn't begin until the current show is complete.
An audiofile located on the internet can be treated like a roll of recorded tape or a cassette recording and can be downloaded, stored, and played at will. It can be started at any point, rewound, fast-forwarded and even edited.
When a radio-style program is available as an audiofile it is popularly known as a 'podcast' and in essence is an audiofile that can be chosen at will be listeners. Many people today produce podcasts and there are plentiful podcast sites, a form of self-service radio.
Throughout this thread I have held agreement with Artisan Radio's declaration that AI (Artificial Intelligence) is a hoax. But because AI has been dangled as a real and present threat in various professions, it has become another 'conspiracy theory' warping the public's imagination. But if AI is any threat to the radio disk jockey, that is only true in terms of replacing paid DJs by AI DJs who work cheap. As far as 'intelligence' is concerned, the disk jockies of today don't say anything intelligent, so the AI version of a DJ might be over-qualified for the job.
This morning I carefully re-read this thread about AI, which stands for "Artificial Intelligence". The whole subject came to mind because I heard a show on NPR talking about the "state of the art" of AI which has become a "thing" because of ongoing introduction of well publicized "AI products" and companies both commercial and non-profit providing and promoting AI applications. The fact that AI is a hoax, as postulated by Artisan Radio (and with which I agree whole heartedly), isn't keeping it from being presented as a real and present "force" with serious potential for serving or dis-serving mankind. The tipping point that is implied is the point when an AI instance starts thinking on its own, or rather, is "told" to think something detrimental to a certain target population. I see that as telling AI to produce and then carry out propaganda, a true weapon already widely used in tailoring human beliefs and behaviors. The danger of AI is not in the human characteristics it has, but in the human characteristics it doesn't have.
My definition of AI is: An insane automation system that will carry out the will and objectives of a mad master.
As hoaxes go, AI joins the list of ancient hoaxes already infecting humanity, including nationalism, patriotism, militarism, religious fervor, ethnic superiority, etc.
Unfortunately, recognizing AI as a hoax does not diminish its potential danger if wrongly applied.
I think I'm having a moment of clarity about AI.
Despite being a hoax, AI has flamed up in the public media as a genuine threat capable of crippling whatever hold on truth we have without it.
Yet, this is the same risk previously posed when only human intelligence was spreading true and false information.
Therefore we live with the same threats and dangers either way.
I write as a human intelligence, but maybe I'm an artificiality without being aware of it.
How can we know if we're real or artificial? I think this is clarity. What's it like for you?
I may be somewhat repeating myself, but I am not intelligent enough to keep track of the past. In the past, before hearing about AI, I was already aware that we humans have an inherited propensity for believing in and spreading superstitious nonsense. And with the coming of AI we begin with the assumption that AI is modeled from human intelligence but will evolve on a separate path with our concern that it will generate its own intellectual errors and become an alien system perhaps at odds with humanity. Can AI possibly correct our own intellectual biases and end up being truly better informed than we are? Even if it did we would probably disagree with it.
It's hubris to the extreme to declare that AI actually thinks and/or evolves.
As I've stated before, what is called AI is simply just another way of looking at solving a computer problem. It's an algorithm. A sophisticated algorithm, granted, as it can change itself, but then, you can do the same thing with other types of algorithms that preceded it (procedural and object-oriented).
We don't know what consciousness/self awareness is in humans. We don't know how we think, not really. We see some of the end results of the processes, and tinker around with those somewhat. As far as I'm concerned, we're not even intelligent, so I don't know how we can assign that label to anything else.
We humans may not be intelligent, but we are capable of thinking we are, and due to this we are open to being deceived. That's the point at which AI has access to our input.
AI is something I don't like. People are pushing Computer technology to far. Sooner or later
there will come consequences to this...
AI has been around for a while. What about computer chess? The computer's brain is way better than the human brain. A computer can analyze the situation and think 10 moves ahead and make the best move possible and do this in seconds. It doesn't matter what your IQ is. Humans make mistakes, the computer doesn't. I can make 3 moves and the game is over and I don't even know it.
The only way you can make a game out of it is if you set it to play at a lower level and all that means it purposely makes a bad moves or doesn't "think" as much and if you win it's because it "lets you" But humans still have to make it and program it. But is this a bad thing? Yes when it goes too far. It's a bad thing when all you have to do is say from your chair "lights out" or clap your hands once and it's like magic. Why can't you get up and turn the switch? The bad thing is that if AI is going to do everything for you. It will be the Zager and Evans song "In The Year 2525" come to reality. That is sad. What about self driving cars? A disaster waiting to happen. Just because something can be done does that mean you have to do it? Or should it be available to the public? Yes I agree with Roy and it will go too far and it will not be a good thing. It's already gone too far. Like in car GPS that has a voice telling you exactly how to go somewhere. You don't have to use your brain. You will physically and mentally just waste away. At least with chess you use your brain still.
What's wrong with a map to see where you want to go? Really, AI began back when a calculator was invented so you didn't have to know how to add, subtract, multiply, and divide. I wonder if this generation can write a note on a piece of paper, or write a letter?
Do they even know how to write at all? Sad sad sad.
Chess is one of those games that really doesn't need the program to 'learn', however it does it. It lends itself to the brute force method of playing, particularly with today's virtually unlimited storage and computer power. Even though it might seem like it, there's nothing like AI involved.
Back in the day, however, when computers didn't have as many resources, the chess programs would attempt to prune lookahead branches, based on evaluating the strength of each possible move. They would take into consideration not only material advantages, but also territorial control and other such heuristics. Once the game was decided, the programs could then tune their evaluation heuristics.
Is that AI? It's learning in a very rudimentary sense, but very problem domain specific. A chess program can't be trained to drive a car, or write an essay, or anything else other than to play chess.
Similarly, programs such as Chat GPT can regurgitate information in somewhat of a coherent fashion, but that's all they can do. And they can also make up bogus facts, and do things incorrectly. In my day, we used to call those computer bugs.
Chess cannot drive a car because it doesn't have a driver's license.
But seriously, is the program that drives a self-driving car a form of AI?
Playing chess against a car while it is on the road would be like a human making cellphone calls while driving.