You've probably heard about it, because the tech industry is rotating when it comes to artificial intelligence development.
In particular, the development of so-called chat bots is progressing rapidly.
One of the currently most popular representatives is ChatGPT.
But what is that anyway? What can it do? What is it good for?
What is ChatGPT?
ChatGPT has been open to the public for free since November 30th, 2022.
It is a prototype of a dialog-oriented chatbot with artificial intelligence.
It was developed by the US company OpenAI and is based on an improved version of GPT-3 (short for Generative Pretrained Transformer 3), also from OpenAI.
OpenAI was originally founded in 2015 as a non-commercial project with the aim of researching and developing artificial intelligence for the benefit of mankind.
Elon Musk and Amazon, among others, were significantly involved.
Microsoft has also been involved in the financing of OpenAI since 2019.
ChatGPT understands natural language input and questions and produces responses and text that are almost indistinguishable from real human responses.
In addition to English, he/she/it speaks other languages such as German, French and Spanish. ChatGPT is not limited thematically. The chatbot can be used in many ways and, for example, answers questions on a wide variety of topics, translates texts, summarizes them, creates essays on a given topic, writes poems or programs. During a dialogue, it remembers previous inputs and responds coherently to subsequent inputs.
OpenAI describes it the following:
We’ve trained a model called ChatGPT which interacts in a conversational way. The dialogue format makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. ChatGPT is a sibling model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.
How does it work?
ChatGPT was created based on an optimized version of the GPT-3 language model, called GPT-3.5. Huge amounts of text containing hundreds of billions of words were used to train the language model.
The texts come from various sources such as Wikipedia, online forums, websites, news sites, book databases, social networks and others. So basically from the entire internet.
The so-called reinforcement learning from human feedback is used for the training.
How can I use it?
Currently, anyone can use the OpenAI chatbot prototype for free.
It is sufficient to register with OpenAI. To register, the user only has to provide an e-mail address and a telephone number for a confirmation SMS of the registration.
If the user is successfully registered with OpenAI, he can start directly and conduct conversations with the artificial intelligence via a text input mask.
Could not have explained it better
The controversy
ChatGPT's impressive capabilities should not hide the fact that it is a prototype of artificial intelligence that still has numerous weaknesses. In addition to the often slow response times and the often limited availability due to overload, the AI chatbot shows some weaknesses:
- the obvious lack of creativity or understanding of irony
- tends to invent plausible-sounding answers or provide incorrect information in certain cases
- does not provide transparency as to which sources of information were used for the answers
- in longer answers, the information is often repeated
- responses may be biased, resulting from the bias contained in the training texts
- many copyright issues regarding the content created by the AI are still unresolved
Some examples
Always objective...
And surprisingly helpful...
Some responses, especially for more general questions tend to get pretty long. I wonder at what point people start to use AI for their master thesis.
You can try it yourself via https://chat.openai.com/
Let us know in the comments if you get some interesting answers or conversations going on...