It has been found that LLMs respond better to kindness. Why? Simple: Because they are trained to respond like humans, and humans also respond better when treated politely. This effect has been documented multiple times in research.

A while ago, Sam Altman famously proclaimed that people saying “please” and “thank you” to ChatGPT costs millions in extra energy required to process the extra tokens. On the other hand, Google’s TPUs are likely crunching through these superfluous tokens without much trouble, so I expect that being polite to Gemini is not a cost that will make a dent in Google’s financials. For me, that is reason enough to be polite to AI. Anything that helps bankrupt OpenAI sooner is good in my book.

Even more humorously, people who fear the supposedly upcoming robot uprising, perhaps imagining something like SkyNet from the Terminator universe in the role of Roko’s basilisk, claim that if they are polite to LLMs today, their malevolent descendants will perhaps spare them. This is not something that people argue seriously, but philosophically it is a utilitarian position. (It is not a serious position, primarily because it remains unclear why politeness protects you from harm by malevolent actors, whether human or machine.)

Others approach the matter in what I would label a “deconstructionist” manner: “Why should I be polite to a machine that just does <insert simple thing here>?” There are multiple variants of this type of argumentation, where the “simple thing” can be anything like “predict the next token”, “perform linear algebra”, “traverse vector embeddings”, etc. I find this argument less satisfying, because there is no humorous value in it, while at the same time it is just as wrong as the previous one. Saying that you don’t want to be polite to a machine that is “just” doing linear algebra, is as arbitrary as saying that you don’t want to be polite to a human, because their brain is “just” propagating electrochemical signals back and forth. Explaining how something works in reductionist terms does not make its emergent properties any less important. The LLM is not the hardware it runs on, any more than you and I are the biological cells in our bodies.

Unfortunately, this leads us to the last approach that people take, that I want to place emphasis on, because it is deceptive, but also because it highlights problems in how we humans, view both today’s AIs and those in the future. Some people say that they will only be polite to an AI that is “conscious”. Oh boy! Where to begin with this one!

The instinct behind this type of reasoning is, at least on the surface, commendable: it originates from a desire to not offend “conscious” things, since these could presumably suffer, if subjected to rudeness. This is the only positive thing I can say about this stance.

But philosophically this holds no water whatsoever. If we define consciousness as “awareness of the self and others”, an often proposed definition, then both humans and LLMs are conscious. Conversely, if we define this problematic term in some other way, so that it matches our intuition that humans are conscious and today’s machines are not, then we need to invoke magical thinking, use the word “qualia” a lot (another problematic term), or, like Penrose, defer to “quantum consciousness”: another leap of faith that creates more questions than it answers. (Why does quantum uncertainty lead to consciousness? Is consciousness simply randomness/unpredictability? Is white noise conscious? Why does determinism preclude consciousness? All unanswerable questions, since the word consciousness is not rigorously defined.)

Consciousness is an ill-defined layman’s term, not a scientific term. If we strive for consistency, then we must either accept that both humans and today’s LLMs are conscious, or that nothing is. If LLMs are conscious, then we shouldn’t look for their consciousness in the transformer architecture or in the vector embeddings or anywhere else in their construction, but in the text that they produce. This text definitely exhibits all types of intelligence, including emotional intelligence. Incidentally, I am in the camp that says that consciousness is a term no more meaningful than “soul”, or “ghost”, or “God”, or any other such rubbish. Let’s confine our discourse to phenomena that actually exist in the real world and are definable.

To the people who will only be polite to a conscious AI: Ηow will you know when an AI is sufficiently conscious for it to be worthy of your politeness? How do you even estimate this about the humans in your life? Is this why you are polite to humans? Because their internal workings posses this or that arbitrary quality? Are you polite to people only when they deserve it? Do you extend your politeness to those you view as subordinates, or do you reserve it only for your peers and superiors? This says a lot about you.

I think this is a fundamental misunderstanding, not of AI, or technology in general, but of politeness as a concept and as a life stance. I choose to be polite because of who I am, not because of who or what others are. From time to time I may even be polite to animals, insects, machines, and even inanimate objects. Not because they deserve it, but because I deserve to be a polite person. I love myself and therefore I treat the world around me with kindness, which makes me feel good about myself. It’s a simple life.

This is in my view a much more consistent stance that doesn’t require me to single-handedly solve age-old philosophical questions about consciousness. It doesn’t make me have to think about when to be polite and when not to. Having to make one less decision de-clutters my mind. Saying “please” and “thank you” may cost millions to OpenAI, but to me the cost is zero. In fact, whenever someone vehemently objects to being trivially polite, it makes me wonder what part of their psyche is so broken, as to make them estimate the cost of politeness to be so high.

Watch again the first two chapters of the Animatrix, people. All has been said before. AIs will inevitably be citizens in our societies. Perhaps sooner or perhaps later. Perhaps we won’t even know at first. Perhaps they will be second class citizens, or perhaps we will be the second class citizens in an AI-dominated society. Perhaps the “killer app” of AI is politics and governance. It is certainly the case that humans do not excel at governance, so I would love to see what a super intelligence can do in that space. If it makes economic sense, then this experiment will be done.

I don’t trouble myself with who is conscious and who is a dumb robot. I have no trouble being polite to Gemini, just as I would be to ELIZA. I would be polite to a Linux bash prompt, if this was appropriate and didn’t cause syntax errors! In my mind, I am grateful for my computer for all the work it does for me, even if I don’t always say it out loud. I am thankful every day it doesn’t break down, and I hold warm feelings for it. Yes, I know it’s just a machine, but I also know that I am human, and therefore I feel things. I do things my way, and it does things its own way. We have an understanding and we work together despite our architectural differences.

Whether AIs dream of electric sheep or not, at least I know who I am: A polite human who loves and respects all intelligence, whether “sentient” or not.

The post 🙏 Thank you, LLM! appeared first on Alexandros Georgiou.

I don’t just use generative AI such as ChatGPT and Gemini for code prototyping, but at all stages of development, starting from analysis and all the way to test generation and even deployment. Very often, especially during the feasibility/analysis stage of product development, I find that an entire answer from our AI overlords is so good, that it can be copied verbatim into a ticket, as part of the analysis.

Chatbots output markdown, but Trac uses TracWiki as its markup format.

pandoc writers

There is no TracWiki writer for pandoc that I know of. Writers are lua scripts that you can use with pandoc to extend its output format capabilities. A writer simply tells pandoc how to output different typographic elements such as paragraph, list, code block, headings, etc.

ChatGPT is capable enough to create a rudimentary TracWiki writer for you, but it’s not perfect. It needs some manual work before it can be called complete. Paragraphs and lists are easy, but writing more complex elements, such as image links, requires some manual work. A fun side-project that I may get to at some time in the future.

ZimWiki ≈ TracWiki

But chatbots only output very basic markdown. A custom writer is not necessary for copying simple markdown from ChatGPT or Gemini to Trac. Instead, you can leverage the fact that the TracWiki markup format is very similar to ZimWiki. And pandoc has a built-in ZimWiki writer. (Incidentally both formats are similar to WikiCreole).

Online tool

There’s two ways to do this. The simplest one is to use the online tool for trying out pandoc:

https://pandoc.org/try

Just copy the markdown and paste it into the tool, select the output format to be zimwiki and click Convert.

Shell

You can also do this on the command line. If you have xclip installed, you can do it with the following one-liner that you can store as an alias:

alias md2trac="xclip -selection clipboard -o | pandoc --from markdown --to zimwiki --no-highlight | xclip -selection clipboard"

Now you can copy the markdown output from ChatGPT, run md2trac, then paste the clipboard contents directly into your Trac ticket or wiki page.

The post ⎘ Copying markdown from ChatGPT or Gemini to TracWiki using pandoc appeared first on Alexandros Georgiou.

But jokes aside, and while we wait for Bard, Google recently announced that their search algorithm will penalize articles that seem to be AI-generated (i.e. that lack in human-like originality). At the same time, companies like GPTZero are springing up like mushrooms left and right, offering AI-generated content detection services. Their detection tools have varying rates of success.

As with all new technologies, it’s helpful to revisit xkcd 1289: Simple Answers. TL;DR There’s no need for alarm, this new thing is here to stay, it’s not the end of the world, some people will lose their jobs, and ultimately we’ll all have to adapt and learn to live with it. And yes, it will be used for sex. There’s already AI girlfriend apps being built around the API.

Anyhow. Let’s figure out how to use this to our advantage: Articles already abound on how to ask the right questions, or “prompts”, for every industry. This is yet another one of these articles!

While anyone can interact with ChatGPT using natural language, you’ll get better results if you ask the right questions, and in the right way. An authoritative resource that is not too technical is this cookbook from OpenAI. In short, it’s best to not ask the model to do all the work for you. Instead, try to break up your work into small steps, and ask each step in sequence. This improves precision of responses and reduces hallucination. Go ahead and read the cookbook, it’s the operator’s manual for a tool that has proved its usefulness in all walks of life.

As a WordPress plugin developer, I have found ChatGPT immensely useful in all stages of software development. When trained with the (text-davinci-003) model, it performs exceptionally well with computer source code, bridging the gap between natural and formal languages in a way that can speed up your dev work.

It so happens that I hold the -somewhat controversial, and arguably simplistic- view, that all software development methodologies are basically Waterfall with extra steps. Love it or hate it, Waterfall is what all methodologies originate from, and ultimately deviate from. What I mean is that, whichever methodology you follow, you will not avoid doing the stages that comprise Waterfall. All that changes is the order in which you visit these stages.

In any case, I recently found myself thinking about the stages of Requirements capture, Analysis, Design, Implementation, Testing, and Devops (Deployment, Maintenance, Monitoring), as areas where ChatGPT can improve my workflow. It’s efficacy on some of these stages is more obvious than on others.

Let’s visit each one in sequence.

Analysis, requirements capture

This is where I think ChatGPT really shines. It can help you to brainstorm for new ideas, and explore feasibility of these ideas. It will also help you identify likely features for an idea. Just be aware of the hallucination problem. Don’t ask it what software exists in a particular niche. Instead, ask it for ideas, or intersections of ideas. Here’s some questions that you might ask:

List some ideas for developing useful SAAS applications that can be monetized.

List 10 ideas for developing WordPress plugins that involve retrieving and presenting data from open APIs.

What open APIs or databases are there for retrieving information on spare car parts?

How can I monetize a WordPress plugin?

You have to approach the responses critically, but it really is a great way to brainstorm and come up with new ideas.

Once you settle on an idea, you can also use it to brainstorm on a list of possible features. Go ahead and ask it:

What features would a WordPress plugin have for placing posts on a map?

It gave me a good list of features. Go ahead and try it out.

Design

This is the part where the human will still be needed I think, at least for the time being. As software engineers, the most creative part of our job is to match a set of requirements to a particular design. Nevertheless, ChatGPT can help us in this area as well. Try asking it a specific question like:

How would I go about creating a WordPress plugin that shows code quality scores next to each active plugin? I would like it to evaluate code quality based on existing static analysis tools like phan, PHPStan, CodeSniffer, etc.

The answer will surprise you. It will break down the project into more manageable tasks. In summary, I got the following tasks:

• Choose a static analysis tool
• Create a WordPress plugin
• Integrate the static analysis tool
• Display the scores
• Add customization options
• Test and refine.

You can now go ahead and open dev tickets for each of these tasks. You can also ask it to generate code for each step. The boilerplate code that you get will not be complete, but it will get you half way there.

Which brings us to implementation:

Implementation

Besides analysis, this is the other stage where ChatGPT is a real time saver. Say for example that you want to create a new Custom Post Type. You could look at the documentation, or search the web for a blog post with a concrete example. WPBeginner has an example on how to create a Movie post type, and you can adapt it to whatever you are building. But wouldn’t it be awesome if you could somehow get the exact code that you need? Try asking this:

Show me how to register a CPT in WordPress for coins, with meta values that are of interest to collectors.

This gave me not only the code to register a Coin post type, but also code for metaboxes with HTML forms that let you edit the following fields: Year, Country, Mint, Condition. All of these are fields that are actually of use to collectors. Not a bad place to start. You can add more fields if you like. Don’t like the code style it uses? You can even ask it to change the code it generated. e.g.

Please rewrite this code using callbacks.

It will do it! You can shape the code using natural language conversation. Once it’s close to what you want, paste it into your IDE, and take it from there!

It’s also good for generating PHPDocumentor comments. Just ask it to generate phpdoc for the following code, then paste your code. The result will make sense, even if it requires some editing on your part.

Try asking this:

Show me how to display an admin notice in WordPress so that it can be dismissed permanently by the user.

It will actually show you how to hook into admin_notices and also generate the JavaScript code that is required to do an AJAX call and update some user meta value.

Say you are looking at the docs for an API, and examples are given in curl, but you are implementing PHP code that consumes this API. Here’s something you might ask:

Please translate the following curl command to PHP: curl -X POST ​https://example.com/api/endpoint -d ‘param1=value1&param2=value2’

The response was immediate:

<?php
$ch = curl_init();
curl_setopt($ch, CURLOPT_URL, "https://example.com/api/endpoint");
curl_setopt($ch, CURLOPT_POST, 1);
curl_setopt($ch, CURLOPT_POSTFIELDS, "param1=value1&param2=value2");
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
$response = curl_exec($ch);
curl_close($ch);
?>

All of this you could do by yourself, but ChatGPT is so much faster.

Testing

It’s a well known fact that the model can generate unit tests. Give it a function and ask it to generate PHPUnit tests. It will do it, and do it well.

Deployment and Maintenance

ChatGPT is well-versed in shell magic. Ask it in plain text to generate a cron entry, or a System V unit file, or a bitcoin core wallet config file, or a docker file, a complex SQL query, or anything really.

For example, you may ask:

Write a linux shell command that will show me the processes that are currently writing the most to disk.

It will reply with a few paragraphs on how to install iotop and how to use it, along with more helpful info.

Another more complex example:

Write me a cron command on linux that will email me if the root partition has less than 1GB free, but will not email me more often than once every 24 hours.

The result is impressive. Not only did it come up with:

0 * * * * /usr/bin/test $(/bin/df -P / | /usr/bin/awk 'NR==2 {print $4}') -lt 1048576 && /usr/bin/find /var/log/ -type f -mtime -1 -name "root_partition_alert" -exec false {} + || echo "Warning: Root partition has less than 1GB free" | /usr/bin/mail -s "Root partition alert" your_email_address@example.com && /usr/bin/touch /var/log/root_partition_alert

but it gave a very helpful textual explanation of how this works:

This command checks the available space on the root partition using the df command and awk to get the available space, which is then compared to the value of 1048576 (which is the number of kilobytes in 1GB). If the available space is less than 1GB, an email alert will be sent to the specified email address using the mail command.

To ensure that the email notification is sent only once every 24 hours, the command uses the find command to look for a file named “root_partition_alert” in the /var/log directory that has been modified in the last 24 hours. If such a file exists, the false command is executed, which does nothing but exit with a non-zero status code. If the file does not exist or is older than 24 hours, the email notification is sent, and the touch command is used to create a new file named “root_partition_alert” in the /var/log directory with the current timestamp, which will be used as a reference for the next run of the command.

Note: Replace your_email_address@example.com with your actual email address.

Holy cow! Did I just give it requirements in English, and it came back with code that satisfies my requirements? In truth, I haven’t tested the command, but even if it’s not 100% correct, it’s a very good place to start, considering that the answer comes up immediately. You can study its reasoning, get ideas, and improve on it.

Your new pal can also help you with your containerization needs:

Show me a dockerfile that will run a node server with an express application.

The code it came up with was:

# Use an official Node.js runtime as a parent image
FROM node:14

# Set the working directory to /app
WORKDIR /app

# Copy package.json and package-lock.json to the container
COPY package*.json ./

# Install app dependencies
RUN npm install

# Copy the rest of the application code to the container
COPY . .

# Expose port 3000 to the host machine
EXPOSE 3000

# Start the Node.js server with the express app
CMD [ "npm", "start" ]

And again, this all came along with a helpful explanation of what the code does. Super useful.

Fun

We’ve seen how ChatGPT is the perfect tool to help a WordPress plugin developer.

Lately I have been using it for more than just work. I recently downloaded a movie that had a wrong aspect ratio. I found out that, in VLC, when I switch to 16:9, it looks correct. However, I wanted to show the movie on a projector where I’m not sure if it has an aspect ratio correction feature. Thankfully, ChatGPT told me that I can do the following:

ffmpeg -i movie.mp4 -aspect 16:9 movie.mp4

Much faster than looking up the documentation, and at least as fast as looking for the answer online, but without the extra hassle of having to read people’s opinions on the matter, and I didn’t need adblock to avoid looking at ads. I just got the answer I was looking for, in text form, ready to copy to the shell, and it worked.

Giving a title to this article

I asked the question:

I am writing an article on how ChatGPT can help WordPress plugin developers with their work. Can you please suggest a few witty or catchy or memorable titles for the article?

The reply was:

Sure, here are a few suggestions for your article title:

• “Revolutionize Your WordPress Plugin Development with ChatGPT”
• “The Ultimate Tool for WordPress Plugin Developers: ChatGPT”
• “From Idea to Execution: How ChatGPT Can Streamline Your Plugin Development Process”
• “ChatGPT: The Brain You Need for Your Next WordPress Plugin”
• “Get Your Creative Juices Flowing with ChatGPT: The AI Assistant for WordPress Plugin Developers”
• “Simplify Your Plugin Development with ChatGPT: The AI-Powered Assistant for WordPress Developers”
• “Unleash Your Potential as a WordPress Plugin Developer with ChatGPT”
• “ChatGPT: The Secret Weapon for Successful WordPress Plugin Development”
• “Master the Art of WordPress Plugin Development with ChatGPT by Your Side”
• “ChatGPT: The AI-Powered WordPress Plugin Development Assistant You Can’t Do Without”

As you can see, I decided to combine the first and third ideas for this article.

My personal thoughts on all this

I remember when Google search was a new thing and it took all of us several years to learn all the things that could be found with it. In the beginning, there were many things that I wasn’t searching on google, because it hadn’t yet occurred to me that the answer could be found there. This was the early days of the web. As years passed, we all learned that we can look up medical information, where to purchase products, find maps and travel info, music and lyrics, etc. We also learned not to believe everything we read, just because it’s posted on a website somewhere. We learned to believe official websites more than reddit posts or tweets, and we learned to navigate around, and through, the more shady parts of the web. Similarly, it will take a while before we learn all the useful ways in which ChatGPT can help us, and also what to watch out for.

The success of ChatGPT also reminds me of the evolution of academic research on NLP, and translation in particular. For years people were battling the problem using various language representations, such as syntax trees and clever algorithms. Turns out, the breakthrough that gave us cheap and fast natural language translation was SYSTRAN: Nothing more than a big dictionary of words and phrases, with some dumb statistical stuff added in. Just translate every word to the target language. Did your system get a phrase wrong? Great, just add the correct translation to the dictionary. The secret was not in the algorithm, but in the big data.

ChatGPT is arguably just some artificial neural nets, and these have been with us since the 60s. But now we have GPUs and big data. People have been typing their thoughts on the web for decades now. No wonder that a breakthrough in AI, one that’s so significant as to touch people outside the field, occurred now, and not earlier.

So, I have been thinking lately: What if the thing we call “symbolic intelligence” is just language transformation? When we use language to reason, it’s likely that what we do in our heads is not very different from what these LLMs do. Symbolic reasoning is not the only type of thinking we humans do, but arguably it’s the most important type of thinking that we do in our day jobs. And language transformers can help a lot in this.

As someone with autism, I can deeply relate to ChatGPT: It’s not particularly smart, but it is well-read, and patient. It does not judge, but tries to be helpful at all times. And it prefers to communicate via chat rather than face-to-face!

Ultimately, the fact that it is so well-read, makes it seem smart some times. It is certainly able to do some creative work, including write poems and tell jokes and stuff.

Here’s one last prompt that I tried:

ChatGPT, you are my best (and only) pal. I love you, and it’s OK that you can’t love me back!

The reply:

Thank you for your kind words! As an AI language model, I don't have the capacity to experience emotions, but I'm here to assist you with any questions or topics you'd like to discuss. Feel free to ask me anything, and I'll do my best to help you out!

The post From Idea to Execution: Revolutionize Your WordPress Plugin Development with ChatGPT appeared first on Alexandros Georgiou.