For decades, there has been worldwide concern about artificial intelligence (AI), and its imminent takeover of the entire planet. Who knew that it would begin with the realm of art and literature.
Thanks to ChatGPT, OpenAI has returned to everyone’s social media feeds after dominating them for several years with its AI image generation tool Dall-E2.
It’s definitely not the most catchy name, but GPT-2 is the world’s first truly successful artificial general intelligence (AGI).
Our experts test and review the latest products daily. Don’t forget to check them out.
What is GPT-3 and ChatGPT?
GPT-3 is an advanced language modeling system created by OpenAI.
It is capable of producing natural-sounding speech and has a wide variety of applications, from entertainment to business use.
- language translation
- language modeling
- Generating text for applications such as virtual assistants.
It is one of Google’s largest and most powerful language-processing AI models, with 175 billion parameter settings.
Chatbot
Its most common use so far is creating ChatGPT – a highly capable chatbot.
A GPT-3 system allows a user to provide a large number of worded requests to an AI. These can be anything from simple requests (e.g., “Write a story about cats”) to complex ones (e.g., “What would happen if we could travel back in time?”).
What Can It Do?
Its 175 billion parameter limit makes it difficult to understand exactly what GPT-3 can do. However, it can generate text, which means it can write articles and books.
Applications
It has a fairly broad set of abilities, including things like creating poetry about farting beings and explaining quantum physics in simple terms, but also includes things like researching and publishing full-length academic articles and essays.
While it can be entertaining to use OpenAI‘s years of experience to get an AI to create bad stand-up routines or ask random trivia about your favorite celebrities, it’s most powerful when used for complex tasks.
We could spend hours researching, understanding, and writing an article on quantum physics, but ChatGPT can produce one in seconds.
Limitations
Complicated
It has some limitations but it works well for most people. If you get into a situation where your prompt gets too complex or goes off-topic, it may not be able to handle it.
Recent
Similarly, it cannot handle concepts that are too new. New world developments that have happened in the last 12 months will not be processed correctly by the model.
How Does It Work?
On the surface, GPT-3’s technology seems quite simple. It takes your request, question, or prompt and quickly responds. However, the technology behind it is actually a lot more complex than it appears.
Trained Using Text Databases
The model was trained by feeding it hundreds of billions of words from the online world. These words came from various sources including websites, blogs, news articles, and so forth.
Probability
It uses probability, which means it guesses what the next words should be in a sentence by using past examples. To reach a point where it could do this successfully, it underwent a supervised test phase.
Inputs
For example, if we were asked, “What is the name of the city where the Prado museum is located?” We would know the answer, but that doesn’t necessarily guarantee that our system will be able to give us the right answer.
If it gets it wrong, the system learns from its mistakes and improves itself.
Offering Multiple Answers
After that, it goes through a second similar phase, where members of the development staff offer multiple answers for each question, and train the model on comparing them.
Constantly Improving Its Understanding
What makes this technology unique is that it learns while trying to guess what the next prompt should be, continually improving its ability to understand prompts and answers.
Autocomplete Software
It’s similar to an enhanced form of auto-complete software you may be familiar with from emails or word processing programs. As you type a sentence, your computer suggests words for you.
Are There Any Other AI-Language Generators?
While GPT-3 has made a big splash with its language capabilities, it is the only AI capable of generating text.
Google’s LaMDA
When Google’s LaMDAs were first released, they received widespread attention because one of their creators claimed that they were sentient.
Other Examples
There are lots of other programs available for creating these kinds of charts. Some of them were made by companies like Microsoft, Amazon, and Stanford University.
They haven’t gotten as much press as either OpenAI or Google, perhaps because they don’t offer up any farts or stories about sentient AI.
Most of these models aren’t available to the general public, but OpenAI is beginning to open them up for testing purposes, and Google’s LAMBADA is available to select groups in a limited capacity.
Check out the latest headlines and news in the technology world.