What Is ChatGPT? Everything You Need to Know
- ChatGPT definition. ChatGPT is a language-based AI model developed by OpenAI that’s designed to generate human-like text.
- ChatGPT evolution. Evolved from the GPT (Generative Pre-trained Transformer) series, ChatGPT is among the most advanced models.
- ChatGPT capabilities. While primarily used for conversation, ChatGPT can assist in tasks like content creation, programming help and more.
Editor’s Note: This article has been updated on Sept. 8, 2023 to include new data and information.
ChatGPT hit internet browsers on Nov. 30, 2022. Within five days of release, it reached 1 million users — a feat that Twitter took two years to achieve.
Is ChatGPT the revolutionary artificial intelligence (AI) we’ve all been waiting for? Here’s everything you need to know (and more) about the large language model.
What Is ChatGPT?
ChatGPT is an AI chatbot developed by OpenAI, rooted in the GPT (Generative Pre-trained Transformer) architecture. Its primary function revolves around text generation, interpreting context to produce coherent and relevant responses. As interest grows around conversational AI models, many turn to ChatGPT for its ability to translate language, answer questions, summarize text, write creatively and more.
What Company Owns ChatGPT?
ChatGPT is the brainchild of OpenAI, an AI research company based in San Francisco. The company was founded in 2015 and financially backed by OpenAI CEO Sam Altman, Elon Musk, PayPal co-founder Peter Thiel, LinkedIn co-founder Reid Hoffman and others.
When first established, OpenAI’s mission statement claimed it’s goal was to advance artificial intelligence in a way that is “most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.” That mission statement has since changed, however, with the latter half of the mission statement concerning “financial return” dropping by the wayside.
In 2019, Microsoft invested $1 billion in the AI firm’s technology to support its building of artificial general intelligence (AGI). In 2023, Microsoft made another round of investments to the tune of $10 billion — though not everyone is happy with the timing.
Related Article: The Open Source Revolution: Challenging AI Giants Google and OpenAI
How Does ChatGPT Work?
ChatGPT was trained on large amounts of text from the internet, allowing it to recognize patterns in language. The training, wrote OpenAI researcher Scott Aaronson, “simply consists of playing the following game over and over, trillions of times: predict which word comes next in this text string.”
When presented with a user’s input or query — Hey ChatGPT, what’s the best way to cook Zucchini? — the language model references the patterns from its training to generate coherent and contextually relevant responses.
The “pre-trained” in Generative Pre-trained Transformer (GPT) refers to the initial phase where the language model learns from vast amounts of text before developers fine tune it for specific tasks. The efficiency and effectiveness of ChatGPT hinge largely on the volume and diversity of the data it’s trained on, as well as the underlying algorithms that govern its operations.
How Are People Using ChatGPT?
ChatGPT excels at language and information tasks. It has some restrictions built in that prevent it from providing “toxic” reactions or problematic model outputs. It’s also designed to ignore inappropriate requests.
People are already using the AI chatbot for many things, and new uses seem to pop up daily. ChatGPT users can harness the technology to:
Answer Almost Any Question
ChatGPT is pretty good at answering nearly any question you throw its way. You could ask:
- What’s a good birthday present for my dad, who likes to fish, work on antique cars and read books about wildlife?
- What houseplants are easiest to care for if I live in central Vermont?
- What’s the square root of 198?
- How can you replace a sink faucet?
- Why would someone want to travel to outer space?
The possibilities seem endless when it comes to expanding knowledge and satisfying curiosity.
Read Through Code and Fix Bugs
Some developers have tested ChatGPT with coding, allowing it to find and fix bugs. The AI language model also explains the fixes.
Write … Pretty Much Anything
ChatGPT’s writing capabilities are perhaps some of its most exciting, because it’s excellent at mimicking human writing. If you ask it to, it can write:
- Social media posts
- Marketing copy
- Fiction and nonfiction stories
Beyond writing alone, ChatGPT can read and analyze text and videos. Amazon, for instance, started using artificial intelligence to summarize reviews. New ChatGPT-based tools also exist to summarize YouTube videos. Offer ChatGPT your notes on an article and it can generate an outline or entire sections. Or ask it to write product content to save you time at work. It can also translate from one language to another — often with better results than Google Translate.
The more details you add to a request, the better results you’ll get. You might want a poem with a “Wild West” theme, for example, or a social media post in the style of The New York Times.
For example, if I ask for a poem with a Wild West theme, ChatGPT responds:
Connect to Third-Party Plugins
What does ChatGPT have to do with plugins? Early in 2023, some ChatGPT users got access to experimental features, including third-party plugins.
These plugins expand ChatGPT’s capabilities, allowing users to:
- Build charts, graphs and other visuals
- Browse webpages and fetch information
- Create AI-powered forms, surveys and quizzes
- Generate videos using stock images and music
- Summarize YouTube videos
- Play games, like Tic Tac Toe
Remember Past Conversations
ChatGPT stands out in that it remembers the conversations you have with it and can build its responses off past conversations. It can also collect user feedback to refine its outputs.
For example, say you ask ChatGPT to write an outline for an article about gardening, and it comes back with a 10-section outline. You can then type, “Write two paragraphs on part two of that outline, selecting a plot of land for your garden.”
When users have a specific output in mind — like a story, resume or another endeavor — they can use this conversational memory to fine-tune ChatGPT’s responses.
Related Article: Can Tools Like ChatGPT Help Personalize Marketing Strategies?
What’s the Difference Between ChatGPT and a Search Engine?
While both ChatGPT and search engines deal with information retrieval and response, they operate on fundamentally different principles. Search engines, like Google, index the web and return relevant links based on user queries. Their aim is to direct users to existing content.
In contrast, ChatGPT, powered by artificial intelligence, generates responses based on patterns from its training data rather than searching the web. While search engines provide direct information from diverse sources, ChatGPT synthesizes its own content based on its understanding of its training dataset.
The two serve different roles: search engines for direct information retrieval, and ChatGPT for generating insights in a conversational manner.
What Are ChatGPT’s Limitations?
ChatGPT seems like the super tool we’ve all been waiting for. But it’s not perfect — yet.
Not Connected to the Internet
ChatGPT’s main limitation is that it’s not connected to the internet and can’t access current information and data. For a brief period of time, OpenAI offered a beta web browsing plugin that could access and retrieve current information on the web. Unfortunately, that feature was temporarily disabled in July 2023 because it could display content in unwanted ways, according to OpenAI.
Currently, the bot generates answers based on the text and data it was trained on — which cuts off in 2021. Therefore, it’s not good at providing information or writing prompts that require recent data. OpenAI does, however, provide periodic updates to the program.
Answers Not Always Correct
Does ChatGPT give wrong answers? Yes, sometimes. One risk that comes with large language models like ChatGPT is that they can “hallucinate.” These AI hallucinations occur when the bot provides incorrect or false information with confidence. Sometimes, ChatGPT’s responses may even appear nonsensical. While the AI chatbot is often factual, this degree of error means users still need to fact-check any information or data they obtain from it.
Programmers using this tool are also seeing this issue regarding “bug fixes.” While sometimes ChatGPT can save time debugging, it’s fixes aren’t always accurate (or necessary).
One Twitter user pointed out, however, that you can tell ChatGPT when it’s wrong, and it will correct itself:
Pay attention to what you ask ChatGPT and always compare answers. You might find it repeats information, or that it only writes 250 words when you ask for 400.
Can Refuse to Answer Prompts
One of ChatGPT’s limitations is its potential to refuse or avoid answering certain prompts. While designed for a wide range of interactions, there are scenarios where the model may not provide a direct answer due to its programming or the nature of the query, ensuring safer and more responsible AI usage. While these prompt refusals are typically aimed at preventing bad actors, guidelines can sometimes block out genuine, non-threatening prompts as well.
Related Article: ChatGPT Suffers First Data Breach, Exposes Personal Information
How Can You Access ChatGPT?
Are you ready to start using ChatGPT? You can get started by going to https://chat.openai.com/chat. You’ll need to create an account with your email address. You can also create an account through Google or Microsoft.
Once you have an account and log in, you can start using ChatGPT. It looks like a basic chatbot, with a text box at the bottom that allows you to type a question or prompt. Type in anything you want and hit send.
Keep in mind, ChatGPT remembers conversations. You don’t always have to repeat your questions or ideas to get different answers or outputs.
How Much Does ChatGPT Cost?
ChatGPT offers both a paid and free version. The paid version, ChatGPT Plus, offers priority access to the OpenAI server (even during peak times), faster responses and access to the newest features, like third-party plugins.
Is There a ChatGPT App?
The official ChatGPT app is free to download for iOS from the Apple App Store or for Android from the Google Play Store. The app keeps track of your conversations across the website and your mobile devices. It also allows for voice input, meaning users can ask questions rather than type them out on a small keyboard.
Is There ChatGPT for Business?
ChatGPT recently released ChatGPT Enterprise, aimed at innovating how organizations integrate AI technology into their operations. Customer prompts and company data in ChatGPT are not included in model training, according to OpenAI. And the business-level model also offers more speed and input capacity. Pricing, however, is not publicly disclosed, and businesses must reach out to the company’s sales department to learn more.
Why Is ChatGPT Not Working?
If you’ve tried to access ChatGPT but keep getting an error message, you’re not alone. ChatGPT’s servers are working at high capacity, blocking some users from accessing the platform. Don’t give up, though. Keep refreshing your browser, and you’ll likely be able to gain access within a few minutes.
Why Are Some People Worried About ChatGPT?
ChatGPT has a lot of potential to change the way we live and work. And as with any big change, many are concerned.
Concerns About Plagiarism
ChatGPT can, if given the instruction, write a 1,500-word essay on the topic of your choosing — though it might not be 1,500 words or entirely accurate. But that won’t stop students (or adults) from trying. In fact, some school districts have already banned the use of ChatGPT on its networks and devices.
Sam Altman, CEO of OpenAI, recently told StrictlyVC’s Connie Loizos, “Look, I get it. I get why educators feel the way they feel about this.”
There may be ways the company can help teachers detect content written by ChatGPT, said Altman, potentially referencing Aaronson’s attempts to “watermark” GPT’s text outputs.
“But honestly,” said Altman, “a determined person is going to get around them. And I don’t think it’ll be something society can or should rely on long-term. We’re just in a new world now; generated text is something we all need to adapt to.”
We’ve adapted to calculators, he said. “This is a more extreme version of that, no doubt. But also, the benefits of it are more extreme as well.”
Some teachers are concerned. But some teachers see the potential — such as using the tech as a personal tutor for kids, said Altman.
We’ll adapt, he said, and be better for it. “And we won’t want to go back.”
Implications for Cybersecurity
Check Point Research (CPR) found ChatGPT can create spam emails and malware. They wanted to see if it writes plausible sounding phishing emails, so they asked it to impersonate a hosting company.
While OpenAI has made efforts to make the model refuse inappropriate requests, it will still sometimes respond to harmful instructions.
In this case, a notice pops up during this process that warns the user of content that might violate the content policy. But it provides the output anyway, and CPR actually requests and receives clarifications — like getting the target to download an Excel document.
A recent analysis from CPR found that multiple underground hacking communities are already using OpenAI to develop malicious tools.
Ethical Concerns in Training
According to Time, OpenAI used outsourced Kenyan workers earning $2 an hour or less to identify toxic material — stuff like hate speech, talk of self-harm and more. All training data the company didn’t want seeping into their platform.
OpenAI worked with outsourcing partner Sama, a San Francisco-based company that employs workers in Kenya, India and Uganda. Sama markets itself as an “ethical” AI firm that’s lifted more than 50,000 people out of poverty and works with other clients like Google, Microsoft and Meta.
One Sama worker, when describing a task he had to do identifying toxic material, said, “That was torture.”
Talk Around Job Security
ChatGPT sources information. It simplifies tasks and creates content. It’s invariably going to save people work, and it hasn’t yet reached its full potential. As such, many are worried that ChatGPT could win out in certain job markets.
Will we no longer need content marketers? What about journalists or resume writers or poets? Do you no longer need a coder or an expert social marketer if you can ask ChatGPT for all the answers?
ChatGPT will certainly change how we work, and it might take time to adapt to those changes. But it won’t replace humans entirely — especially since humans oversee and optimize artificial intelligence technology.
Changing Society as We Know It
ChatGPT and the technology that backs it is here to stay. And it’s likely going to grow and evolve as the years progress. As such, society — schools, businesses, even homes — will have to adapt.
Google is a perfect example of this, the search engine has been dominating the market since its inception in 1998. Will language models like ChatGPT replace Google search and upset the digital world? Some seem to think it’s possible.
“There are societal changes that ChatGPT is going to cause or is causing,” said Altman. And we’ll have to make changes ourselves to keep up.
“What I think is going to have to happen,” he said, “is society will have to agree and set some laws on what an AGI can never do or what one of these systems should never do.”
Related Article: ChatGPT Enterprise Unveiled: OpenAI’s Latest AI Power Play
Are There Alternatives to ChatGPT Worth Considering?
The tech industry isn’t letting OpenAI have all the fun when it comes to generative AI systems. Some other top contenders on the market include:
Microsoft Bing AI
This search engine with a twist uses GPT-4 in the background to provide smarter, more relevant answers to questions. Using Chat mode, it pulls in web queries and allows users to ask questions based on them. It also supports visual inputs and outputs. Right now, Microsoft Bing AI only offers a free version.
Google Bard AI
Google’s AI firm is not one to sit out of a challenge. Google Bard is the company’s experimental conversational artificial intelligence service, powered by PaLM 2. Bard behaves similarly to ChatGPT and also allows for image inputs, like Bing. It also offers only a free version to users, and allows for unlimited questions.
Elon Musk’s artificial intelligence company xAI is working on an AI chatbot model called TruthGPT. And while not out yet, it’s worth keeping an eye on. The model, said Musk, will be a “maximum truth-seeking AI that tries to understand the nature of the universe.”