What is a Generative Pre-trained Transformer?
By now, everyone has heard about #ChatGPT4 and all the ways this tool will disrupt the world, eliminate white collared jobs, and breed a new wave of “know-it-alls”. For some, this may be a bleak outlook of our future without knowing how AI will infiltrate our everyday lives, (very similar situation when the internet first appeared), but for others tethered to computers and the internet, this is the search engine we’ve all been looking for…
Despite some of the possible “negative” consequences, this new tool opens up an opportunity for accelerated productivity and research. We are drowning in a sea of overloaded information, and our traditional methods are strained. This may be the new #Google as our search for information is simplified and more direct.
Even though this tool will lead you to an answer quickly, users must understand its capabilities and limitations in order to realize its benefits. Just like during the early days of search engines, we curated keywords and phrases for engines to interpret your needs, only to sift through a just few hundred million results…
Chat GPT can easily sift through the noise with very little guidance to find you the “right” answer quickly, which is great for quick questions and basic facts. However, once you start asking hard questions, it becomes obvious there’s a mysteriousness of ths tool that only data scientists understand. For the consumers of this information, we should still exercise caution about what “it” tells you to be “truth”.
This new generative pre-trained transformer tool is great to answer all your silly questions, but if you don’t care about so much about “truth” and “facts” (and have plenty of remedial/redundant tasks to take care of), you can use this AI capacity to generate code specific to your needs. That in itself will avoid hours of sifting through #stackoverflow forums (only to realize you missed a closing curly bracket) and allow you to focus on higher lever challenges.
Bashing out Scripts
Finding facts, explanations, and information about the world may be great for our personal growth and awareness, but the real benefit from this tool is from generating low-code snippets without stumbling through syntax errors.
In the world of system administration, #bash scripting is a powerful tool and method of programming for system administrators, developers, and anyone else who works with command-line interfaces. Bash is ubiquitous in the computing world (comes native with Linux) and is heavily used to automate repetitive tasks, perform system maintenance, manage data, and creating workflows. All #Apple computers have Bash installed, and #Microsoft recently joined that bandwagon with its Linux Sub-System capability.
Like with any programming language, Bash scripts take time to write, test, and deploy. With Chat GPT, you can generate templates of codes and modify it for your needs.
For example, maybe you want to monitor the health of your site with a simple script that will ping your IP address every 5 minutes. Unless you write scripts regularly, you might need to consult your MAN page on your ping command, and stumble through all the syntax rules for building a conditional “While True” statement…
Or you can ask Chat GPT.
Those who code all the time, you may not be impressed with this example, but imagine if you had this tool during the early stages of your programming journey and avoiding all those hours dealing with syntax errors.
Don’t get me wrong. This tool by no means will make anyone a programmer overnight, but there’s a lot of potential for this tool to accelerate the learning process tremendously, and debug with ease. But of course, not all programmers can tap into this evolving technology just yet, and if you in graphic design, you might have to wait awhile…
There are many people who are worried about AI technology obtaining human like super intelligence (for good reason) that leads to a situation where it could dominate or even enslave humanity… But, fortunately, we are still very far from that possibility as it requires a degree of imagination and cognitive abilities unique to humans.
We can see this from the many examples where these AI tools fail to respond to more advanced elements of logic, reasoning, and imagination, some of which are hilarious. All of which are delivered with absolute confidence, no fatter how obvious it is factually incorrect…
One can spend countless hours trolling this bot to highlight all the language blunders only humans can understand and laugh about, but I was more interested its ability to generate scalar vector graphics (normally stored as an “svg” file extension).
Of course, I wanted to test the boundaries, so I asked Chat GPT if it could generate digital renderings of an impossible 3D form. Unsurprisingly, the bot failed, but was it I who failed to teach this machine?
We’ll probably never know, but for fun, I chose the Penrose Triangle for its simplicity and geometrically scalar form, which I felt was reasonable to ask a computer to generate.
But expectations were still a little too high…
On the left side of the following image is the “Penrose Triangle” Chat GPT created, and the right side is an actual Penrose triangle I created.
Just to make sure that this wasn’t just a misinterpretation of the request, I asked if it knew what Penrose triangle looked like. The written answer was spot on, but the text rendering was not even close.
Even after all this, I had to admit I was still very impressed, mostly due to the fact that this tool generated responses within in seconds while I spent hours fumbling through InkScape.
Thoughts Overflow Error
The science behind AI is not new and the term has been used since the 60’s, but the computational powers/capabilities at the time were not conducive, and information/data was sparse. Now that we have overcome the technical hurdles, the time is now to harness the design and application of this new and “free” tool. We’re now in an era where AI machines will exponentially multiply the sea of information on the internet, some of which may be incorrect, biased, and possibly harmful to humanity. We should never lose sight of the over consequence of this new tool, and for those who may not benefit from productivity gains, I encourge you to familiarize yourself with this technology to understand the magnified societal impact.
To reallly understand the impact, you might want to ask yourself:
- How do I know the responses are correct, unbiased, scientific, and does not harm others?
- Who should control the AI algorithms and source code? Private or public
- How is this tool regulated to ensure information is democratized and accessible?
- Are we living the prequel of Minority report?
One good documentary to catch up on some of the negative consequences of AI/ML is titled “Coded Bias” on Netflix. This movie provides a bleak view of how AI technology has gone wrong today…
Summary Outlook, not email
Honestly, I’m not too familiar with the AI roadmap, and I have no idea what to expect in the next few years. However, I am very confident that it’s time for everyone to start experimenting this tool as it starts to evolve and improve. Just like in the early days of the internet, there’s a lot of potential, and those who understand how it works, will be the ones to ride the wave…
I guess the moral of this story is that we’re still in the “80’s” of AI technology, but at this rate, Chat GPT5 will take use to Y2K or beyond…
But, on the plus side, this bot is programed to be very polite.
Disclaimer: Chat GPT did NOT write this whole article for me.