ChatGPT: A Guide To What It Can and Cannot Do

ChatGPT has taken the world by storm and continues to be the talk of the town. Many are thrilled by this development, while others fear they, ultimately, are going to lose their jobs over AI. If you want to know what the buzz is about, then this article is for you. 

The world has vastly and rapidly evolved over the years, and we are now living in the future of technology. Multiple companies and individuals are trying to create and improve technologies that may address the burdens of humankind, and these technological advancements include artificial intelligence. The release of ChatGPT triggered a heightened and more intense global race in the development of artificial intelligence.

In March this year, the company’s latest AI model—GPT-4—is now being used to update the performance and capabilities of ChatGPT. As an aftermath, the stakes are getting much higher. So, read along and discover what makes ChatGPT the current buzzword and what makes it special.

What Is ChatGPT?

ChatGPT is an artificial intelligence chatbot that makes use of natural language to create human-like responses to conversations. You can have conversations with the ChatGPT chatbot on whatever topic you like. This AI language model can generate responses to questions and compose different kinds of written content, such as essays, poems, social media posts, web articles, emails, letters, and codes.

ChatGPT is one of the numerous generative AI tools where you can enter prompts to generate human-like texts, images, and videos created by AI. Other examples of generative AI include Dall-E, a text-to-image model by OpenAI, and Lensa, an AI-powered app that can convert selfies into illustrious self-portraits.

OpenAI ChatGPT is trained based on a compilation of websites, articles, images, and social media posts scraped off the internet and real-time English conversations with OpenAI’s hired human contractors. The chatbot “learned” the proper writing grammar and structure and today’s most frequently used phrases.

The GPT in the chatbot’s name stands for Generative Pre-Trained Transformer, which refers to its method of processing requests and formulating chat responses. OpenAI trains ChatGPT on reinforcement by human feedback on top of the reward models that rank the best replies. Human feedback helps augment the AI chatbot with machine learning to help improve its responses in the future.

You can liken this popular chatbot to an automated customer-service chatbot on a specific website or social media account where you can converse with it to raise questions and concerns. However, the former is several inches smarter.

Who Created ChatGPT?

The genius behind this famous chatbot is OpenAI. OpenAI is a research company based in San Francisco, California, and was founded by Elon Musk, Sam Altman, and other notables in 2015. Just three years succeeding the company’s founding, Elon Musk withdrew ties with the company for personal reasons.

OpenAI’s founders started it as a non-profit company with the goal of producing artificial intelligence that can contribute to the development of humankind. However, in 2019, the company’s leadership announced that OpenAI is already a for-profit organization with the hopes that more people will invest.

OpenAI launched ChatGPT in November of 2022.

How to Use ChatGPT?

If you are curious as to how ChatGPT works, then you can try using it for free. Here’s how:

1. Create An OpenAI Account

To do this, go to chat.OpenAI.com and create an account using your email address or by directly connecting your Microsoft or Google account. You must create an account on OpenAI’s website to gain access to ChatGPT.

2. Accept The Terms and Conditions

Once you have created and logged into your OpenAI account, you will proceed to read the fine print. Once you agree, click on “Next” and “Done.”

3. Start Chatting

After having a full-fledged account on OpenAI, you can get started. You are free to ask ChatGPT whatever questions you have.

How Does ChatGPT Work?

ChatGPT uses OpenAI’s GPT-3 language model—a neural network machine-learning model and the company’s third generation of Generative Pre-Trained Transformers. The chatbot uses deep learning, or a subset of machine learning, to come up with human-like texts via transformer neural networks.

The chatbot works via its Generative Pre-Trained Transformer. Transformers are referred to as specialized algorithms to discover long-range patterns in a sequence of data. That is exactly how ChatGPT knows what words, sentences, and paragraphs to say when you converse with it. Its pre-trained transformer allows it to stay on the topic even on long stretches of text.

Since a transformer needs massive data, its training comprises two stages—pre-training on generic data and fine-tuning on tailored data for the specific tasks it has to perform. In the case of ChatGPT, the chatbot was trained on a vast reservoir of texts found online to learn language structures and rules and fine-tuned dialogue transcripts to figure out how humans converse and the characteristics of human conversations.

Human trainers converse with the AI chatbot and rank its responses. The responses, otherwise known as reward models, help the trainers determine the best answer. To keep training ChatGPT, you can upvote or downvote a response by clicking the thumbs-down or thumbs-up icons near its response. Additionally, you may compose written feedback to improve and fine-tune how ChatGPT provides responses.

Can You Use ChatGPT on Mobile?

There is currently no mobile ChatGPT app. However, you can access this AI-based tool via your smartphone’s mobile browser. All browsers support the website so you don’t have to worry about issues concerning this.

The same steps, as enumerated in this article’s third main heading, must be followed to use ChatGPT on your phone. The chatbot will work the same way it would when you access it from a computer.

What Kind of Questions Can You Ask ChatGPT?

You can ask ChatGPT any question you like, be it simple or complex, and expect the chatbot to give an answer. You can ask close-ended questions, such as “When did Abraham Lincoln issue the Emancipation Proclamation?” and open-ended questions, such as “How do you define life?” Best of all, it can debug and write codes because it is well-versed in STEM disciplines.

There are no limits as far as the questions to ask are concerned. However, ChatGPT uses data up to 2021 only, and for anything past that year, ChatGPT doesn’t know.

Regardless of its limitations, ChatGPT is versatile and can be used for more than just human conversations. You can use ChatGPT to accomplish the following:

• Draft emails
• Compose music
• Code computer programs
• Create an article title
• Write social media posts scripts
• Solve math problems and equations
• Discover keywords for SEO
• Create write-ups for websites
• Play games
• Compose product descriptions
• Assist with job searches, such as writing a resume and a cover letter
• Describe complex topics in a simple manner
• Ask trivia questions

And a lot more! Unlike a typical chatbot, ChatGPT is able to store questions in its memory bank to carry on a conversation in a fluid and human-like manner.

Can ChatGPT Provide Wrong Answers?

Like all other technologies, this AI tool is not without limitations and may give faulty answers and incorrect information. Thus, it is critical for you to take its responses with a grain of salt and double-check whether the given information is really accurate. Since it’s in the process of learning data based on the text its users provide, misinformation may happen.

With that, it is crucial for you to rate the answer and give feedback to help develop and improve the model.

Since ChatGPT has limited data only, it cannot determine the current date. So, giving it a prompt to give what day Easter is going to be this year won’t be able to generate an answer. And this is how ChatGPT differs from actual human conversations and search engines.

What Are the Other Limitations of ChatGPT?

As we’ve mentioned, ChatGPT has limitations despite looking grand and impressive. Aside from giving limited and possibly incorrect information, here are ChatGPT’s other limitations: 

Unable to Comprehend Human Language Complexity

Since the chatbot is trained based on input, it may be unable to comprehend the human language complexity to its full extent. Its response may often seem shallow and may lack emotions and insight. 

Responses May Sound Unnatural

It may overuse some words like the or since it can only predict its responses. As a result, you will have to review and revise any content you will produce through AI tools to make it sound more natural and human-like. A good piece of writing is relatable and may reflect the persona of its writer, and creating write-ups through AI-powered tools will lack in this aspect. 

Does Not Cite Sources

The chatbot summarizes, but it does not cite sources. It doesn’t even provide insights or analyses into any statistics or data. It may provide statistics, but no explanation of what the statistical data means. 

Can’t Understand Sarcasm

Because the chatbot is AI and without human emotions, it does not have the ability to understand irony and sarcasm. It is based on a dataset of texts and provides answers based on that data alone. So, it can’t replace humans (yet). 

Can’t Cover Multiple Questions in A Single Response

The chatbot is not divergent and cannot design its answers to address multiple questions in a single response. So, if you ask it whether an elephant would make a good pet based on its size and give a follow-up question that would say, “what about a dog?” The chatbot may only focus on the size variations of the question and not on whether or not an elephant would make a good pet based on what it is. 

Are There Ethical Concerns Surrounding ChatGPT?

While the chatbot is helpful and useful in accomplishing several tasks, some ethical concerns depend on how it is being used. Two of the several ethical concerns include:

Deceitful Use

Users may use the chatbot unethically by utilizing it to cheat, scam, impersonate, or spread fake news. Since the output is AI-generated, it can be difficult for authorities to spot the person behind the unlawful content. Many educators are concerned by the fact that students can use the tool to write papers, plagiarize, and cheat.

Fortunately, for these educators, there are tools available to check papers for plagiarism and cheating.

Cyberthreat

Since the AI tool is able to write codes, it can also be used as a tool for cyber threats. Threat actors may find ways around OpenAI’s safety protocol and use ChatGPT to create malware.

Is ChatGPT Worth Using?

We understand where the hype encompassing this AI tool is coming from, and we can’t deny the fact that, sooner or later, AI can be used for anything. So, if you want an easy route of accomplishing something, such as drafting an email or getting ideas for an article, then ChatGPT is worth using. However, always remember that while it is impressive, it has limitations and faults, so don’t rely on it too much. 

If you are looking for other AI-powered tools and chatbots for your marketing stints, check out our recommended ones in this Postoplan review and MobileMonkey review.