Technology: Who’s Afraid of ChatGPT?

ChatGPT came roaring on the scene at the end of 2022, and it’s already become a familiar staple in most tech-savvy people’s lives. You may have used it yourself to create a grocery list or find out the pros and cons of various roofing materials. ChatGPT can automatically generate the text for anything you want to express in words in no time flat. 

As a freelance writer — one of the many careers that will be affected by generative AI —  I have to admit that chatbots feel like a bit of a nemesis, and I would not have delved into ChatGPT had it not been for the need to research this article. I am hardly the only one with qualms. Concern about generative AI programs in the European Union recently led to passage of the Artificial Intelligence Act, the first major acknowledgment that this technology needs careful regulation, or it could exploit human beings’ privacy and labor.

Meanwhile, tech companies worldwide are competing to create chatbots that are more human-like and can be programmed to perform a variety of human tasks. 

Indeed, there are experts who believe that ChatGPT is already as creative as the human mind. Jennifer Haase, author of “Artificial Muses: Generative artificial intelligence chatbots have risen to human-level creativity,” argues that insofar as “the ability to produce something new and useful” defines creativity, chatbots are competitive with humans and, in some cases, outperform them. It’s worth noting that ChatGPT 4, the more advanced version of Open AI’s famous chatbot, performed significantly better at creative tasks than earlier programs.

Better large language models are on the horizon. But for now, ChatGPT is a delightful way to pass some time — I prompted it to write a paragraph in the style of each of my favorite authors, in addition to other fun, time-wasting activities — and has a number of practical uses as well. 

What is Generative AI?

ChatGPT and other chatbots, like Microsoft’s Bing and Google’s Bard, utilize a type of artificial intelligence called a large language model (LLM). LLMs are new on the programming scene, but in just five years, their capacity has enabled chatbots to advance from rudimentary stock responses to ever more advanced capabilities. 

The magic behind an LLM is its progression from a recurrent neural network to a transformer model. Recurrent neural networks predict text in isolation, one word or character at a time, similar to how the predictive text function works on your phone. 

The transformer model, on the other hand, looks at longer pieces of information, sentences, and paragraphs. Both models receive an input, encode it, and then decode it to produce an output prediction. But only the transformer model is able to see words in context – where they belong in the grammar of a sentence, how frequently one word follows another, what the connotative value of certain words is – which results in a much stronger ability to create meaningful, more “human-like” responses.

The larger the LLM, the more data it can “train” on and the more sophisticated the responses will be.

The free version of ChatGPT reminded me, during one of our conversations, that its “knowledge cutoff” occurred in January 2022 – in other words, the company stopped training the chatbot with new data on that date. This is a significant limitation. Historical events have a tendency to shift perceptions and outcomes, leaving Open AI’s free chatbot with an outdated perspective on the world. 

The Roll out of ChatGPT

Open AI released its latest iteration of ChatGPT to the public in November 2022 with no advance notice. (In case you were wondering, GPT stands for generative pre-trained transformer). They did not regard their humble chatbot as anything more than a “research preview,” according to an employee of Open AI featured in an interview with Will Douglas Heaven for the MIT Technology Review. They hoped to use the chatbot to collect feedback that would help them make improvements; that was their best-case scenario.

Instead, ChatGPT became a runaway hit, exceeding the team’s wildest expectations. Open AI has recently signed a multi-billion dollar deal with Microsoft and created an alliance with Bain, which plans to use its AI programs in marketing campaigns for big clients like Coca-Cola.  

In March of 2023, a more advanced version of the chatbot, ChatGPT 4, was released as a paid feature. It will cost you $20 a month to access the “most capable” version of the ChatGPT. 

Practical Ways You Can Use ChatGPT Right Now

As a budding entrepreneur, you may already know some ways to leverage the power of predictive analytics in order to improve your business’s ROI. But what if you’re also a working parent trying to juggle many responsibilities at once? Does ChatGPT have any practical applications for your daily schedule?

In the spirit of the article, I asked what Open AI’s free version, ChatGPT 3.5, was useful for. What it told me is the bolded part: 

• Content Creation. You can use ChatGPT to craft web content, write emails to difficult relatives and create two-page screenplays where Putin and Dostoevsky discuss world peace. You name it; ChatGPT can write it. If you don’t like the outcome of the first response, change your initial prompt or question. You are likely to produce the best result by entering many similar questions or cues for the bot to respond to. 
• Drafting and Editing. A useful function for me has been summarizing notes I make while using voice dictation, but you can also use the chatbot to outline and draft presentations, grant applications, team memos, and term papers. 
• Learning Assistance. The chatbot can teach you and your children a variety of general facts, like the main industries in China or the number of times Nigel Richards has won Scrabble tournament matches. You can use it to help “reverse engineer” outlines out of existing drafts of essays and presentations – a useful trick to help get your main point across.
• Programming Help. ChatGPT has the capacity to code an entire website! It can also explain programming concepts and offer guidance on computer languages and frameworks.
• Language Translation. The chatbot will translate whatever it is prompted to translate into the language of your choice instantaneously (to check for mistakes, you should always run it through another translation tool, like Google Translate). 
• Conversation and Entertainment. It’s entertaining to set the chatbot to tasks, such as listing the native birds of Maine or writing a haiku about itself. It is entertaining to see how changing a single word in the original prompt can bring out “another side” to the bot.
• Virtual Assistant. ChatGPT creates a template message that you can input into your calendar to remind yourself and members of your team of an upcoming meeting. You can also use the app to make lists, organize your daily schedule, and brainstorm ideas about upcoming projects. 
• Simulation and Role-Playing. The bot can create detailed biographies for fantasy characters as well as help shape the fictional world in which the role-playing takes place. It can help with the creation of dialogue and plotting. 
• Idea Generation. By changing the input you give ChatGPT, you can create nuances even if the chatbot lacks that sophistication, something I find useful in ideation. In fact, ChatGPT is a great way to brainstorm ideas because it can provide a broad scope in such a short period of time.

Some Final Observations

According to the people who do this for a living, there’s a 5% chance that generative AI will cause the extinction of the human species. That’s a non-trivial risk of dying at the hands of some less benign descendent of ChatGPT. Definitely something to think about. 

Another sobering thought? Experts who responded to the same survey say that generative AI’s ability to “outperform humans on every task” by 2047 was 50%. That’s 13 years earlier than they predicted the robot takeover from last year’s survey.

This underscores an important point. Computer programming is evolving faster than the human brain. Those generative AI art photos your friends are posting on the internet, the ones with crazy hands and improbable color combinations, will soon be indistinguishable from real images and human-made art. The new and improved programs will also write novels, code their own platforms, invent mathematical equations, write case law, and maybe even figure out how to breach the space-time continuum  — all while humans sit around wondering if there are any leftovers in the fridge. 

Our failure to keep up with the technology we have created has important existential, legal, ethical, and social implications. But there are also practical considerations. If you are looking to purchase an existing business, you need to think about how chatbots and generative AI will affect the future of the industry you choose. Some industries will have the capacity to fully automate before long. Is that a positive or a negative for you?

In a rapidly changing world, it’s more important than ever to figure out where to land.