304 North Cardinal St.
Dorchester Center, MA 02124
The ChatGPT-Chat AI chatbot delivers amazing performance, it generates entire essays with just a few keywords. You want to “test” ChatGPT? Tricksare explains how it works. The OpenAI chatbot is called ChatGPT (short for Chatbot Generative Pre-trained Transformer) and is currently omnipresent. It is indeed highly amazing: you enter a question or any topics and the machine promptly answers – in many cases at a rather high level. For “ChatGPT”, it doesn’t even need a separate version, you can just start typing. Make yourself a picture and open the ChatGPT chat. Tricksare answers the most important questions about ChatGPT in the following paragraphs, briefly clarifies the terms AI and AI, warns against a supposed ChatGPT app and provides background information.
ChatGPT uses artificial intelligence (short: AI, artificial intelligence) to “understand” human language and generate answers that are almost indistinguishable from human expressions. ChatGPT is the prototype of a dialog-based chatbot developed by OpenAI and released in November 2022 as a beta version. OpenAI is a US-based company. Involved in the organization are entrepreneur Elon Musk (Tesla) and Microsoft. According to Wikipedia, the company wants to develop and market artificial intelligence on an open-source basis so that it does not harm society. The organization wants to carry out “free cooperation” with institutions and scientists and makes patents and research results available to the public. OpenAI’s GPT-3 language model is the basis for ChatGPT, it is based on deep learning technology, a subtype of machine learning. This technique, in turn, is based on the training of algorithms from several networks with massive amounts of data.
Yes, the chatbot can be used for free – at least in this preview phase. You only need an account with the developer. However, this is not likely to always be the case. By the way: With a script, ChatGPT can also be used as a voice assistant on the iPhone. But beware, there is no official application for smartphones and tablets yet and dubious developers are currently trying to make money with fake apps.
Politics, authorities and schools are at least alarmed. As various media report, there are bans. The academic conference International Conference on Machine Learning (ICML) has decided that authors cannot produce scientific articles completely with ChatGPT. In 2024, ICML will reevaluate the ban on AI-generated texts. The New York City educational authority is even taking it a step further and, according to Chalkbeat, has blocked access to ChatGPT in its networks because they fear students could use the AI tool for their work.
Stack Overflow, an internet platform for software developers, banned the submission of answers created with ChatGPT as early as 2022. And as reported by golem.de, the doubts have also arrived in Germany. The Brandenburg SPD used ChatGPT to submit a small inquiry to the state parliament. It was about how to prevent fraud in university studies and how fair grading can still be maintained in view of the new technological possibilities.
Since Microsoft announced that it would incorporate the chatbot into its search engine Bing, Google is at least alarmed. Although Bing is the second most used search engine in the world, it has been lagging behind the internet giant by a huge margin for years. ChatGPT could bring Microsoft’s service to a new level and make search queries “more interactive”. This not only gives the AI the dubious reputation of a potential Google-killer: some analysts even fear that the future of classic search services is at risk.
A big problem: how do you recognize texts written by ChatGPT? The concerns are varied. Users could cheat on school, study or exam papers, opinion makers could deliberately spread misinformation. Now, the ChatGPT makers from OpenAI have introduced a software to detect their own AI. The recognition is still only mediocre, as the developers admit themselves. In test runs, the software correctly identified computer-written texts in 26% of cases. At the same time, however, the system falsely assigned 9% of texts formulated by humans to a machine. So the new software is not yet reliable, but it’s a step in the right direction.”