For this week’s entry on our blog, we will explore the essential information you need to know about Chat-GPT. The increasingly popular and slightly controversial new AI program.
I’ll explain what it is, how it came to be, and what the idea behind it is. We would delve into the larger discussion of artificial intelligence and in the end and by the closing paragraphs, I would have helped you answer some of the most pressing questions you may have regarding this new phenomenon including the one that serves as the premise of this article. Let’s start with understanding what ChatGPT is.
Think NLP, then ChatGPT
“Generally speaking, AIs pre-train using two principle approaches: supervised and non-supervised. For most AI projects up until the current crop of generative AI systems like ChatGPT, the supervised approach was used. —-David Gewirtz.
When I was doing the research for this article, one of the things that became obvious to me was that people didn’t truly understand what goes on behind the scenes of the new ChatGPT technology that we’re all excited about for weeks on end now. The huge amount of time it would take to fully explain how it works and the tension around whether it should be allowed to exist because it could represent a threat to humans and their jobs didn’t help its case either.
On the basic level, ChatGPT represents an artificial intelligence chatbot developed by a company named OpenAI that provides meaningful responses to users based on the context and intent behind their questions.
One brief exploration I love about what ChatGPT is, how it works, and what its purpose is comes from David Gewirtz who is a computer scientist and distinguished lecturer at UC Berkeley.
David, in one of the articles he contributes to various websites mentioned how the technology’s unique power lies in its ability to parse questions and produce fully fleshed-out answers that are polled from publicly available information or data.
According to him the reason for the buzz surrounding the technology is due to the use of the non-supervised training approach that supposedly became suddenly feasible before its development and introduction. This fact, he believes you may not have known, reflects in its “ChatGPT” name, as in, Chat Generative Pre-trained Transformer.
To dive a little into the technical description, ChatGPT is what you can call a computer program that leveraged natural language processing methods and was introduced to allow computers to harness massive amounts of data to generate ‘learned intelligence’ that gives them the ability to respond to any text that a person feeds into them with a new piece of information that is relevant to the context of the conversation. The program is believed to be better than any other before it at producing sentences that sound like they were written by a human solely for this reason.
Since it’s the non-supervised training approach that was used to develop ChatGPT that makes it so different from other AI programs like itself, let’s look at what this approach is in the next paragraph for a fuller understanding.
It would be impossible to anticipate all the questions that would ever be asked, so there really is no way that ChatGPT could have been trained with a supervised model. Instead, ChatGPT uses non-supervised pre-training — and this is the game changer. —-David Gewirtz.
In layman’s terms, what ChatGPT adopting a non-supervised training approach meant is that the program is built in a way that lets it understand the syntax and semantics of human language and uncover patterns and underlying structure of data from the one it was given to memorize on its own prior to launch and which serves as the repertoire where words are taken from to form human-language sentences response to users.
It might help to read the previous paragraph again.
The reason for its sudden success and the apparent massive potential that came hand in hand with the threat to humans and their jobs that has led people to the kind of question that is the title of this article is due to how this approach confers ChatGPT, a computer program, with an almost limitless knowledge in a way that has never been possible before because its developer can just continue to train newer versions of the program with more and more information or data.
In contrast, where a supervised training approach is adopted, an AI program is trained to map certain outputs or answers to specific inputs or questions.
An explanation of this fact was given by David who noted that “For example, an AI could be trained on a dataset of customer service conversations, where the user’s questions and complaints are labeled with the appropriate responses from the customer service representative. To train the AI, questions like “How can I reset my password?” would be provided as user input, and answers like “You can reset your password by visiting the account settings page on our website and following the prompts.” would be provided as output.”
Where ChatGPT gets data from
According to OpenAI, the creators of ChatGPT, the program was trained on a dataset known as WebText2 which is a library of high-quality, well-annotated, and very very large text data from websites like Wikipedia and lots of books.
It’s reliably reported that ChatGPT trained on about 45 terabytes(1 terabyte = 1,000 gigabytes) of exclusively text data from these sources.
This massive amount of data according to David Gewirtz in a ZDNet article is what “allowed ChatGPT to learn patterns and relationships between words and phrases in natural language [Human language] at an unprecedented scale, which is one of the reasons why it is so effective at generating coherent and contextually relevant responses to user queries.”
So, will it kill your company?
When you think about it, the shrewd capability of ChatGPT and newer AI programs like it that are certain to be fed with all the books and other information that represent the entirety of the raw human knowledge that has ever existed in humanity is likely to remain unknown.
At the moment, the potential that it could kill your company is there. People are already seeing how several occupations like writing, web designing, law etcetera and companies that provide services in these industries could be heavily affected by the growing knowledge base and adoption of artificial intelligence programs like ChatGPT.
Although as a language model, ChatGPT in particular was mainly designed as a tool to provide information and assistance to individuals and organizations in terms of business insight discovery, and recommendations, as well as to provide solutions such as to help automate their customer service, improve their marketing strategies, optimize their operations, and so on, we cannot rule out its potential for doom.
In other words the possibility that AI programs can harm you or your business is there even if you believe that because ChatGPT in particular cannot take physical actions, it therefore cannot kill your company or any other company for that matter figuratively.
It is likely that your company, if it fails to evolve and adopt AI tools like ChatGPT, could face serious challenges and struggle to compete with those that offer AI-powered solutions in the future and as a result die a natural death thereof.
When I asked ChatGPT itself this same question, it suggested that instead of worrying about the program killing your company, you should rather focus on how to use AI and machine learning programs like it to your company’s advantage just as I did with this article and hope that you are able to keep enough pace with the evolution of the AI tools and technology in general so that your company can remain relevant for years to come.
Learn how to create prompts and use ChatGPT efficiently from the next article on our blog where you can also get insights from other pieces about business, technology, and how we engineer some of our most interesting projects.