Artificial Intelligence on steroids
image source
I've been playing with GPT-4 for the last few days.
"What is GPT?"
GPT stands for Generative Pre-trained Transformer, a natural language processing model developed by OpenAI. GPT models are designed to generate human-like text by predicting the most probable next word or sequence of words in a given sentence or text. They are pre-trained on a wide range of reference material, such as books, articles, and web pages, which allows them to learn the statistical patterns and relationships between words and phrases, which enables a response in natural language.
"What's the point?"
GPT-4 will have the ability to perform multitasking. It will be able to handle multiple tasks simultaneously, making it a versatile tool that can be used for a wide range of applications. For example, it could be used to develop chatbots that can perform multiple functions simultaneously, such as answering customer queries and processing transactions. One of the most exciting potential applications of GPT-4 is in education. With its ability to reason, understand context, and multitask, GPT-4 could be used to develop personalised learning platforms that cater to students' needs and learning styles. It could also be used to develop intelligent tutoring systems that provide students with personalised feedback and guidance.
Deep Thought
image source
GPT-4 is expected to be able to reason and understand the context better. It is likely to have a deeper understanding of the relationships between words and be able to understand more nuanced concepts and ideas. This means that it will be capable of generating more relevant and accurate responses to queries, making it a valuable tool for businesses and organisations that rely on natural language processing technology. I've already seen examples of sales proposals written by GPT, along with systems designed to offer customers alternative solutions (Read cheaper) if an offer is rejected.
image source
One of the most exciting potential applications of GPT-4 is in education. With its ability to reason and understand context, it could be used to develop personalised learning platforms that cater to student's individual needs and learning styles. It could also be used to develop intelligent tutoring systems that provide students with personalised feedback and guidance. However, the development of GPT-4 is not without its challenges. The sheer number of parameters it is expected to have will require significant computing power and resources to develop and train. Additionally, there are ethical concerns around the potential misuse of the technology, particularly in fake news and propaganda. I did a little test that I posted about on the Fediverse yesterday. I asked GPT to write me a blog post about Linux.
I then ran it through a couple of plagiarism checkers. The only thing they could find was six words used in the same string but on websites that had nothing to do with Linux, which leads me to believe that it is possible to post GPT-generated content and pass the work off as your own because there is no mechanism to disprove ownership! You need to understand that the technology is not taking content from anywhere but creating it based on what it has learned. Scary stuff!
image source
Image created by irisworld
My actual name is Pete. Here is why I have the username dickturpin.
You can also find me on Social Media
 |
https://mastodon.org.uk/@dick_turpin |
 |
https://castlecannon.house/profile/dick_turpin |
|
https://twitter.com/dick_turpin |
 |
https://www.facebook.com/peter.cannon3 |
This work is licensed under a Creative Commons Attribution 4.0 International License.
Create your license: Choose
Podcaster. FOSS, Linux & Creative Commons Advocate.
"Be who you are and say what you feel because those who mind don't matter and those who matter don't mind."
https://twitter.com/17152901/status/1636323415761068033
The rewards earned on this comment will go directly to the people( @dickturpin ) sharing the post on Twitter as long as they are registered with @poshtoken. Sign up at https://hiveposh.com.
We've had plenty of people re-writing info from elsewhere to make Hive posts and this just automates that. I think there were already plagiarism checkers as well as tools to try and fool them. It's an arms race.
Chat GPT may be different as I think it can pull from lots of sources. If you don't know what those are then you don't know how reliable it is. There are already tools trying to detect when it was used, but there may be costs involved.
I think some cases will stand out where the quality of posts from a user radically changes, but with new ones it may be harder. It definitely presents challenges and it will get worse. The genii is not going back in the bottle.
My Hive votes will continue to go to stuff I like. That can involve building a relationship with the author and machines may struggle with that, for now.
Welcome to the future grandad!
It doesn't "pull sources"; it literally creates its response based on what it has learned. The AI creates the textural response GPT gives you, so there is no plagiarism. It's kind of like asking your friend to write something for you. Unless that friend reveals that they, in fact, wrote it, there is literally no way of knowing other than if your friend uses words or phrases that everyone knows you would never use, but here's the rub. Openminds is working on that too and looking to personalise responses, so for example, as you know, I use a lot of South East London phrases, dialects and nuances. Fairly soon, GPT will learn that and respond accordingly, i.e. in the same language I or anyone else would normally use in their day-to-day interactions.
Well it is using existing text sources to 'learn' from. Some may be unreliable. There may be possible analysis of the style of writing, but that will need some other form of 'AI'. I do see people being accused of using it by HW.
There's also some discussion over on Mastodon with some people very against it. The technology is here and it will get used. I've only played with it, but it may have real uses.
If you get caught using methods that some on Hive do not approve of then you are likely to get slammed. Some may disagree on what should be allowed, but being honest helps. If I generate an image then I will say so.
I suspect the vast majority will prefer to create their own content. I could see myself using it for work purposes, but it would really feel like cheating if people used it for the Fediverse, Twitter or even Hive.
The clincher will be when people start linking it to their accounts with scheduled Tweets or posts whereby the AI knows to produce a set amount of characters for each particular platform based on a list of topics submitted to it.
"OK, Google (Alexa), today's GPT topics are..........." 😐
Not everyone has your morals, so we need to be vigilant. If people are posting like a bot then they may be one.
Chat GPT can only color within the lines; that is, it can only draw from approved sources, and if you read enough of its responses you find out that its outputs are somewhat formulaic.
So there's an opportunity, here, to use it as a tool to create a creative template that you then go back and tweak. Goodbye writer's block!