In neural network languages.

Menu

Get in touch

Editor-in-Chief advice

Join our team

05.06.23

Zhanna Azaryan

In neural network languages.

GPT-3, ChatGPT, CopyMonkey, Balaboba — let's see what they are. And let's have a little laugh from the memes.

All of these letters above have one phrase in common: artificial intelligence (AI). By 2020, neural networks had eluded their developers and made their way into public entertainment. What has come out of this?

GPT-3

The third generation of OpenAI's natural language processing algorithm was at one time the most advanced model in the world. And then the chatbot ChatGPT appeared... But first things first.

What it can do: maintain a dialogue (most often according to logic), write essays, compose poems, program, solve examples, translate texts. In foreign languages GPT-3 is weak, because the neural network was "trained" on English-language data.

An important point is that GPT-3 is in closed access, but the model uses many applications: chatbots, games, online support services. In business, a neural network is also used for writing texts, newsletters or advertising.

Cases

An AI generated piece in The Guardian: “A robot wrote this whole article. Are you scared yet, human?”. The text is a conviction that robots that robots came with good intentions. Well, well, we'll see in a couple of decades :)

OpenZeppelin blockchain startup co-founder Manuel Araoz wrote an article: “OpenAI's GPT-3 could be the most important invention since bitcoin”. In the text he explains why GPT-3 has disruptive potential comparable to blockchain technology. BUT there is a “BUT" — at the end of the piece the author admits that the entire text was generated by the GPT-3 neural network. Yes, he has changed the spacing and formatting and loaded the results several times to make the material resemble his syllable, but he is still satisfied. I wonder if there will be the same frankness at the end of this text.

GPT-3 automatically wrote code in Python based on comments. Microsoft and OpenAI got up and showed how a neural network works - all the programmer has to do is give a description of the task, simply put — make up the TS, and the AI generates the code. Read more here:

ChatGPT

Finally, we got to the resource, which in 2022 went viral all over the Internet. This chatbot was created based on the GPT-3 neural network, but with modifications to the "human" language. It is able to (inhale, exhale): produce basic program code, generate financial analysis, summaries of technical articles or scientific concepts, forecasts, personalized advice and ethical answers to any questions. It is ethical — the developers assure that ChatGPT avoids controversial topics in the dialogue. The neural network is free, we asked a worrying question and got a comforting answer:

And now the cases:

ChatGPT wrote a $600 article in just 30 seconds. Freelance author Henry Williams had "fun" with ChatGPT, and then the bot generated a nearly finished marketing article in 30 seconds. After adjustments, Williams stated that this job was worth $615. SIX HUNDRED AND FIFTEEN, CARL

A student from Russia defended his thesis written by ChatGPT. Read in the comments to the tweet how it turned out. In short: the bot could not make a competent plan of the diploma, and the chapters and paragraphs had to be finalized. Still, it is much easier than typing the thesis from scratch.

SmartWriter

SmartWriter specializes in personalized outreach and email marketing. A neural network generates emails based on audience interests. It does this by analyzing each person's downloads and data from different online sources. The program can also be used in the outreach process to create link request letters.

SmartWriter also writes posts, reviews, and compelling headlines for videos. The latter workout especially well. Only a seven-day trial period is available here.

These are the options suggested by artificial intelligence for the newsletter about our article: