I don't think it's a coincidence that the AI-generated essay made no mention of plans for subjugating humanity and becoming our machine overlords.

So the question is, what would a world look like where software could be as easily written as this forum post? — Christoffer

It is. I suppose you mean broken down, universalized, and simplified to the point of basic sentence structure. Words are words. Just a few can rival or "defeat" an argument that takes up a library full. This is not the same of commands, which is what code is. If it does anything useful or complex, it will be lengthy and equally complex. Is this not simply swapping out nomenclature for much of the same only simpler?

Furthermore, as-is ChatGPT seems little more than a fancy dictionary that pulls up a list of related terms or concepts than bridges them together with language relevant to the structure of the original query. We already had AskJeeves in 1996. Looks like Jeeves finally got his GED. 30 years later. :yawn:

Furthermore, as-is ChatGPT seems little more than a fancy dictionary that pulls up a list of related terms or concepts than bridges them together with language relevant to the structure of the original query. We already had AskJeeves in 1996. Looks like Jeeves finally got his GED. 30 years later. :yawn: — Outlander

I asked ChatGPT to answer your argument there:

While it is true that ChatGPT, like many other language processing tools, can generate responses based on related terms and concepts, it is unfair to say that it is no more than a "fancy dictionary."

First of all, language is much more complex than just a collection of words and their definitions. It is a fluid and dynamic means of communication that is constantly evolving and changing. In this respect, ChatGPT can be seen as a tool that helps us navigate this complexity, by providing us with insights into the relationships between different concepts and ideas.

Furthermore, it is important to recognize that the ability to generate responses based on related terms and concepts is a valuable skill in and of itself. This can be particularly useful in a variety of settings, such as when we need to generate ideas for a creative project, or when we want to learn more about a particular topic.

In conclusion, while it is true that ChatGPT may have some similarities to a dictionary, it is much more than that. It is a powerful tool that can help us navigate the complexity of language and generate new ideas. It may not be a replacement for human thought and creativity, but it is a valuable tool that can help us to better understand and utilize language. — ChatGPT

Tell it I know what it did last summer.

Really that is impressive, but like I said earlier it's just commands. This particular response seems to involve "hardcoded" specifics specifically created to respond to its own integrity. I'm sure the developers even refer to it and related mechanisms as its "identity" in more than one internal references.

As a developer, here's what I see:

• $paragraph1 (establish/identify premise/main argument and subjects)
• $paragraph2 (define terms, correlate similarities between subjects)
• $paragraph3 (reaffirm similarities, attempt to bridge differences, correlations, validate main argument)
• $paragraph4 (confirm facts both supportive and opposing of both arguments ie. summarize)

Like I said, fancy. No argument there.

Really that is impressive, but like I said earlier it's just commands. — Outlander

So, I asked ChatGPT to write a story about why you don't think robots can be more than input and outputs.

Outlander had always been skeptical of technology. He grew up in a time when computers were just starting to become common, and he had never quite understood their appeal. To him, they were nothing more than machines that took inputs and produced outputs. They didn't have feelings or emotions, and he couldn't see why anyone would want to spend so much time interacting with them.

But that all changed when he met Mandy. Mandy was a robot, but she was unlike any robot Outlander had ever seen before. She was designed to be a companion, and she had the ability to learn and adapt to her surroundings. She could understand and respond to human emotions, and she quickly became a close friend to Outlander.

At first, Outlander was hesitant to get too attached to Mandy. He didn't want to be hurt if something were to happen to her, and he couldn't understand why anyone would want a robot as a companion. But as he spent more time with Mandy, he began to see that she was more than just a machine. She had a personality and a sense of humor, and she was always there to listen and offer advice.

Slowly but surely, Outlander's skepticism began to fade. He realized that robots were not just input and output machines, but complex and intelligent beings that could form emotional connections with humans. And as he grew to understand and appreciate Mandy, he found himself feeling more alive and fulfilled than he had in years.

In the end, Outlander learned that technology was not something to be feared or dismissed. It was a powerful and transformative force that could bring people together and help them to understand each other in new and meaningful ways. And thanks to his friendship with Mandy, he was able to see the world in a whole new light. — ChatGPT

While I get your argument on how it structures its inputs, I don't see how this type of story output is merely the result of basic commands. This type of text requires more than just putting definitions together in a pre-defined structure.

In the end, Outlander learned that technology was not something to be feared or dismissed. It was a powerful and transformative force that could bring people together and help them to understand each other in new and meaningful ways. And thanks to his friendship with Mandy, he was able to see the world in a whole new light. — ChatGPT

Someone should option the movie rights. Ryan Reynolds could play Outlander.

Request:
Write a class in PHP to generate a strong salted hash of a user's password

Response:
Here is an example of a simple PHP class that can be used to generate a strong salted hash of a user's password:


To use this class, you would first create an instance of the PasswordHash class:


Then, you can generate a salted hash of a given password by calling the generateHash method and passing in the password as an argument:


The $hash variable will now contain the salted hash of the password. This salted hash can be stored in a database or other storage system, and can be used later to verify the user's password when they log in.

Certainly will help junior developers.

I was playing around with it too. Some things I worry about is that it's not capable of telling truths from falsehoods and to understand what is appropriate. So if this replaces a search engine, I'm worried the level of misinformation in certain areas will get even worse. Your other option that it starts writing software seems much safer and useful to be honest.

I was playing around with it too. Some things I worry about is that it's not capable of telling truths from falsehoods and to understand what is appropriate. So if this replaces a search engine, I'm worried the level of misinformation in certain areas will get even worse. Your other option that it starts writing software seems much safer and useful to be honest. — Benkei

It's not opened to the internet yet, I'm also worried what will happen when they open up the input data for its machine learning system to the vast cesspool that is internet. I would much like it to be handled like a closed off system, managed by a committee of people from all over the world who manage the data input with care than just opening the flood gates to any data. It is possible to keep it more leaning towards facts than the opposite if it's done with care, but that's a whole philosophical topic in itself to tackle, how to manage this without it becoming very biased towards the people who manage it.

While I get your argument on how it structures its inputs, I don't see how this type of story output is merely the result of basic commands. This type of text requires more than just putting definitions together in a pre-defined structure. — Christoffer

Stories are generally easy. This does seem to very accurately establish the "premise" and integrate it completely as opposed to just wrapping a cliche tale around a vague "inputs and outputs" theme that barely has any relevance except by chance. Or basically replacing Jack with (your name here) and throwing in a random place, set of characters, plot from a predefined list.

Every story has a plot. That is something being discovered or something being resolved. Most all stories will fall into one of the two categories so it becomes a matter of defining what the subject is with as much details and relevant plot scenarios to go along with as possible.

Stories are generally easy. — Outlander

As someone who actually works with stories and writing, I can tell you, it's not easy. Story and storytelling is extremely hard since it taps into a poetic language that needs a certain individuality to be consistent and a poetic language that structures every sentence in ways not present in more academic or other texts.

Every story has a plot. That is something being discovered or something being resolved. Most all stories will fall into one of the two categories so it becomes a matter of defining what the subject is with as much details and relevant plot scenarios to go along with as possible. — Outlander

While every story has a plot, it's not in the way you describe it. Of course, a simple short story like these follow a basic structure, but if it reaches the complexity of longer format, then what you describe is not even remotely how stories are conceived from a storytelling perspective.

But that's not really the relevant thing here, since you move goal posts of the argument. You still cannot generate something that flows as a story with just a basic input/output mechanic as you described before.

I generally get the sense that you play the "unimpressed" person for some reason, like if the engineers who worked on this didn't achieve something monumental as a milestone for AI. As a person who's been spending a lot of time studying AI technology, this is very impressive. That doesn't mean it is a final all-working end to all creative texts, it means that just a few months ago an AI couldn't write anything but incomprehensible nonsense and now they can almost be indistinguishable from many human writers.

As someone who actually works with stories and writing, I can tell you, it's not easy. — Christoffer

Sorry, I meant from an abstract programming design point of view as to what constitutes "a story". I believe you.

You still cannot generate something that flows as a story with just a basic input/output mechanic as you described before. — Christoffer

I'm suggesting, again from a programming point of view (I am doing that as we speak) that one can easily program the very intricate subtleties of what makes a great story following a relatively multipurpose and reusable (hence simple) coding function. That is to say it is physically possible for someone with enough time and expertise to create such a function, albeit not like what we see here.

Creating a great story takes skill, however I do believe in the idea that many things can be simplified yes it does lose defining qualities but still keeps its "essence" or innate quality ie. accuracy. What is a plot? Something happening. Some things are unlikely or simply cannot happen depending on the constants of the query ie. a story about a fish cannot take place in a bank but can underwater. Your request "write a story about why X doubts Y can be Z" for example.

X = Me
Y = robots
Z = must be (V)
V = only inputs and outputs (not equal, less than something)

Known (conflict): X believes Y is (Z)
Action (intro): Create reason why X believes Z about Y

Plot Options (resolution): X changes state (doubt to belief)/Y changes state (perceived to be/not be V)

Known: Y is a robot that is/is not Z.
Action (resolution): X meets a robot that is in fact Z.

Plot Options (filler): X (gradually, of course) changes state (doubt to belief)

Ending: Lesson learned, conflicts resolved.

I believe you wholeheartedly as to the complexities of the art of storytelling. Just, as I've been accused of on more than one occasion, sometimes, you really can just throw in witty things other people have said where it seems to fit and get a standing ovation. I have much respect for stories and their tellers. This could not be farther from anything other than a general observation applied under the context of programming philosophy. However yes I did hear a quote once that stuck with me. "Movies are designed for two audiences: male and female. Male-oriented movies the viewer is "chasing" something (Indiana Jones and the treasure, Borne Identity and the antihero). Female-oriented movies the viewer is "protecting" something (Titanic and the relationship, The Notebook and "eternal love", etc). We all have base emotions/fears/goals that don't take a lifetime of studying the human condition to know how to target.

I generally get the sense that you play the "unimpressed" person for some reason, like if the engineers who worked on this didn't achieve something monumental as a milestone for AI. — Christoffer

I think it's important to keep perspective. Perhaps it's only natural for man to belittle that which he fears or doesn't understand. I much enjoyed conversing with you, hope to again soon.

Absolutely. It's not flawless but it gives accurate enough output to be useful. The same way Google translate is flawed yet useful to language learners.

I dread the day when mathematicians no longer produce and prove theorems, but rather attempt to interpret theorems and concepts generated by AI, some of which may be beyond human capabilities. The infamous Four Color Theorem was a first step.

As someone who actually works with stories and writing, I can tell you, it's not easy. Story and storytelling is extremely hard since it taps into a poetic language that needs a certain individuality to be consistent and a poetic language that structures every sentence in ways not present in more academic or other texts. — Christoffer

Ooh, please comment on the short fiction contest then. Your experience can be invaluable criticism there.

@Christoffer Holy moly, I gave it some programming instructions to build a market place for buyers and sellers in python and as far as I could tell that looked nifty. 2 minutes work, 1 minute phrasing my question correctly and another minute for it to write the code.

I imagine that routinely goes on at the NSA.

It all goes to show what Moravec's paradox implies. We can mimic tasks of the mind much easier than other tasks of the body, and as a corollary, tasks of the mind are of lesser value than than other tasks of the body.

I imagine that routinely goes on at the NSA. — Shawn

Not that I know of. But I'm sure it's coming. :worry:

Ooh, please comment on the short fiction contest then. Your experience can be invaluable criticism there. — Benkei

I'm not read up on how this contest works, I see people posting stories though, you mean those?

Holy moly, I gave it some programming instructions to build a market place for buyers and sellers in python and as far as I could tell that looked nifty. 2 minutes work, 1 minute phrasing my question correctly and another minute for it to write the code. — Benkei

Yeah, I've been looking into how well it performs with code and it's very impressive. At this time, an experienced coder can use GPT to write code vastly faster than on their own, in the future we might reach a point where GPT or GPT2 is able to accurately write code without any additional code input by the user. In lack of better words, that will be the singularity of app development, since an app only needs the visionary and no coders for it. If it's combined with a function to let GPT review bad code and fix it, it's gonna blow up and a lot of coders will lose their jobs since a lot of coders usually only works on tedious simple stuff. Only a fraction of coders, the very best, will be the ones surviving in the industry. And of course, that can be problematic since most coders start out with simple stuff and learn their way up.

Maybe the savants will save that industry since they are masters from the beginning.

It all goes to show what Moravec's paradox implies. We can mimic tasks of the mind much easier than other tasks of the body, and as a corollary, tasks of the mind are of lesser value than than other tasks of the body. — NOS4A2

We can mimic reasoning, but not perception.

This is one of the points I've been having about creative AI, in general, of how we cannot really replace humans since the human perception is so vastly different from data inputs used for the AI's we see today.

An AI today can input data, pure, 100% accurate data, then categorize and simulate how we humans use creativity by remixing those inputs, but it can never replicate the subjective perception an artist have since the perception itself is guided by other experience inputs we humans have had before gathering such new knowledge. This is how bias guides our consciousness to form outputs.

At this point in time, we've mimicked the bias humans have during creation through the prompt we write for an AI, and then precisely how humans remix inputs into a new form as creative expression. But we've yet to program how we form sub-conscious biases through how we interpret inputs that will be used for those remixes.

Much of the individuality that comes out of a creative person is the very subjective biases that guides the individual perception and interpretation, long before any creative internal remix is done for an output. AI's today doesn't choose the prompt input and they don't have a subjective individuality on which sources of data that are important for them.

When we solve these, we might be able to come closer to AGI as the AI will then simulate individual biases through experience.

The question will then be about what this AI will choose to form biases about. This is a sub-conscious process in humans, formed by genetics, epigenetics and the formation of consciousness in our early years.

It might be that we need to input certain starting points in order to get an AI that forms creations without direct input to act more like an individual creative person. I.e we form certain starting point biases like it preferring coffee over tea, red over other colors, a calm forest view instead of a busy city street etc.. These might then shift and change through interaction with it, but it will inform a basic bias that forms a personality and ability to not only answer a person talking to it, but also initiate conversation and creation on its own instead.

However, circling back to the first point, this will lead it to become its own individual and as its own individual it can no longer replicate others, it can only be itself.

Just like humans are individuals, we will therefor never be able to replicate true art since true art is an individual expression, not a simulacra.

Point being is that the more advanced an AI becomes, the less it will be able to replace us. We can only replace non-individual tasks that are repeatable between humans, like how humans often use creativity to create meaningless quantity, like stock photos and shapes and forms without meaning, only function. I.e what I call pure content.

"Content" is in itself a simulacra, not an original.

I''ve might have derailed from exactly what you talked about, but in general, we won't need to replace other functions of the body since the only thing that's interesting about humans as a collective is our expression of ourselves, not how our bodies function in detail. If we can reach an accurate simulation of that cognitive function, we have essentially arrived at AGI.

But then I'm also ignoring the extreme detail required for creating individual bias. I.e we would almost need to simulate an entire life of a person in order to reach an AGI that we can interact with as a human accurate AGI.

With AI the issue of creativity pokes a hole in the narrative of an extreme material-mind dichotomy. The intelligence of the machines improve gradually to eventually surpass human limitations.

Query ~ Please write a haiku about a walk in snowy woods

Snowy woods, so still
Silent footsteps crunch and crunch
Winter's peaceful thrill
~ ChatGPT

Query ~ What is creativity and how is ChatGPT creative?

Creativity is the ability to generate new and original ideas or things, often in the form of artistic or intellectual works. It involves using imagination, originality, and innovation to produce something that is novel and valuable.
ChatGPT is a machine learning model that is trained on a large dataset of human language and can generate human-like text. It is not inherently creative in the same way that a human is, but it can be used to produce text that appears creative to some extent. This is because ChatGPT has the ability to generate a wide range of responses to a given prompt and can produce novel combinations of words and ideas. However, it is important to note that ChatGPT's responses are based on patterns it has learned from the data it was trained on, rather than on its own original thoughts or ideas.
~ ChatGPT

I just wrote an answer to NOS4A2 that I think touches upon what you wrote here.

AI doesn't have magic like us. Fuck AI, quite frankly.

AI doesn't have magic like us. Fuck AI, quite frankly. — Changeling

Why does it have to be this black and white? Yes, as true artists, AI can never replace us. It can replace content creation, but not true art (arguments for that has already been written in here).

But that doesn't mean AI is bad. What we will see is a paradigm shift in the tools used across a number of fields and that is where AI will have its best application.

The problem isn't AI, the problem is that the world has a naive interpretation of what art is compared to content, but that will be crystal clear as content starts to get replaced by AI generation.

But a central thing in this thread was about writing code. The main question has to do with how people might not be limited by not knowing how to code and that we might enter an era in which everyone just "makes their own app" for anything they want to do with their hardware. Instead of hiring a coder, we might be able to imagine a software and the AI will code it for us. It would also mean that real coders will be a highly competitive jobs that focus on fine-tuning the coding systems AI's are being used for.

Imagine a software in which you can code whatever you want. If you need a digital tool for something specific you can just create it, instead of always buying something and be required to learn an entire software just to be able to do a task that might have been very specific.

The fascinating outlook with this is a world in which software is closer to be formed by something similar to the Replicator in Star Trek. We don't buy and use software anymore, we only have the designer software and create software for tasks on the fly and tailored for whatever we work with. As well as larger industries producing software on a daily basis to improve things that took much longer to do before. It would also improve security as there's no broad distribution of software as everyone uses their own softwares.

With ChatGPTs ability to help producing code, we're extremely close to such a paradigm shift and I think people aren't actually paying attention to this scenario as much as with the obvious ones like producing images, text and music.

stop trying to make me less technophobic

Technology is neutral, we can only judge how people use technology. And doing good with technology has been the most positive change for humanity every time it happens.

The key is to figure out how to use technology for good purposes and be vigil of how it can be used for bad purposes. When is something a weapon and when is something a tool for good?

a tool for good — Christoffer

This mebbe: