How ChatGPT is changing the way we use artificial intelligence
Assistant Professor of Computer Science Ryan Mattfeld offers insight into how newly developed technology might fit into our lives.
The public was quickly fascinated by the power of a new artificial intelligence technology, ChatGPT, a chatbot developed by research and implementation company OpenAI and released late last year. It has already demonstrated its ability to provide detailed answers to complex questions while using the information it processes and user feedback to improve its responsiveness.
"Machine Learning and Artificial Intelligence" by mikemac marketing is licensed under a CC BY 2.0 license.
ChatGPT has proven to be versatile, with users using the technology to compose music, debug computer code, write restaurant reviews, generate advertisements, and answer quiz questions. He is able to give his answers in a conversational way and has generated enthusiasm for its potential, as well as concerns about how it could be used.
But what exactly is ChatGPT and what does it say about the current and future state of AI? Today at Elon's, we caught up with assistant professor of computer science Ryan Mattfeld, who uses ChatGPT as a topic for class discussion and has an idea of how it could transform our use of AI.
When did you first hear about ChatGPT and what was his first reaction?
Ryan Mattfeld, Assistant Professor of Computer Science
ChatGPT was launched on November 30, 2022 and I first found out about it in early December. The first example I saw was software written by ChatGPT, which included an explanation of how the code works. I was immediately skeptical of the abilities and at first assumed the example had been chosen with care. I kept learning more about ChatGPT, so at the end of the semester, I created an account and started playing with it.
I was immediately impressed. I gave him an assignment used in one of my 2000 level courses the previous semester, which included a specific programming task embedded in a story/scenario about building robots. This was several paragraphs long, contained unnecessary information, and referenced parts of the class that ChatGPT was not aware of. I thought it would block ChatGPT.
I made a mistake. He immediately produced an accurate solution, including an explanation of the code he wrote and a description of how he solved the provided problem. I quickly sent an email to everyone in my department alerting them that ChatGPT was real and asking for their attention.
People are used to search engines that can direct them to answers to their questions or provide resources for more information. Why ChatGPT is taking a big step forward?
ChatGPT and search engines have two different purposes. The main purpose of a search engine is to try to direct you to specific resources. The main goal of ChatGPT is to generate reasonable responses to inputs using natural language. The most critical difference is that the main focus of ChatGPT does not include precision. It's definitely a secondary goal, but it's not a guarantee.
That being said, there is definitely an overlap. Part of ChatGPT's main goal is to condense a wide range of data sources into one useful response. So when ChatGPT provides an accurate answer, it simplifies searches, enables dialogue, and provides answers in a conversational way. Instead of finding a list of sites, searching for those that relate to your topic, opening the sites, finding the specific information you want, and often saving and trying again, ChatGPT can provide an answer in easy-to-understand language and of course, direct and useful.
Also, ChatGPT allows you to keep track. If you want to elaborate on any part of his answer, you are welcome to do so. Just ask for a more elaborate answer in the part that interests you the most.
Can AI technologies like ChatGPT learn from the inputs they receive and the reactions to responses?
Yeah! In fact, it was one of the main mechanisms in its development and is still used for tuning.
Modern machine learning solutions are based on vast amounts of raw data. This is also true for ChatGPT - its initial training was done with vast amounts of data collected from the internet. However, as OpenAI, the creators of ChatGPT, describe on their site, humans were deeply involved in the formation of ChatGPT.
After initial development of the huge data set, ChatGPT responses were refined by human taggers. This process began by asking one human to ask a question and another human to provide what they thought was a good answer. This information was incorporated and another round of modifications began. During this stage, the AI produced several different responses to a single question, and human taggers ranked the AI responses from best to worst. This feedback was also incorporated into the network.
Now, when you use ChatGPT yourself, you'll see options to give the responses that "thumbs up" or "thumbs down" generates. If you choose "Dislike", you can describe why you did not like the answer. This is used to tune ChatGPT.
What are the ways you see ChatGPT being used now, and how do you think these uses will grow as the technology becomes more advanced and integrated?
Working in the IT field, my first inclination of course was to consider how ChatGPT is starting to be used in software development. He is surprisingly good at writing good code for relatively simple programs.
However, it is not perfect. It still makes mistakes and can produce code that doesn't work, but it looks great. Therefore, I expect the most successful software developers to use it as a support tool, but they will need to be able to review the responses to identify and fix bugs. There is still a great need for informed humans to verify that the provided solution was correct rather than convincingly wrong.
Bing logoAnother interesting perspective for IT people is that ChatGPT is an open API, which means that software developers can integrate ChatGPT into new coding projects we develop. This means that we will soon see ChatGPT integrated into many, many other programs and applications. I expect ChatGPT to improve over time and I think we'll see it in more and more places. For example, ChatGPT is already integrated with Bing and will soon be integrated with Microsoft Word.
Of course, ChatGPT is starting to affect domains outside of IT as well. I created an assignment at Cornerstone of Technology and Society that I taught in January regarding ChatGPT. The task included the use and analysis of ChatGPT in various ways, including a contest in which students tried, and largely failed, to identify the articles written by their peers and those written by ChatGPT. The last question in the task asked them to consider how ChatGPT might affect a specific area of society with supporting evidence. In particular, law, IT, journalism, and education are some of the areas where several students predicted that ChatGPT would have a significant impact.
As with many technologies, there are pros and cons. Do you have concerns about how technology like ChatGPT will be implemented?
There really are many ways to approach this question. It has many advantages. This can make apps easier for everyone to use. Instead of requiring the peculiar and specifically formatted queries that computers often need, the use of a common language will soon be effective in a much broader way. It also has a positive impact on accessibility to help people with language disabilities. There are, of course, all the other cool things you can do too, from playing games to help clear writer's block, to explaining complex concepts in simpler terms, to simulating a conversation.
Of course, there are also disadvantages and risks. Generally speaking, if someone trusts ChatGPT too much, it could hinder its development. Making mistakes is essential in the learning process. Using ChatGPT to skip this step could lead to a false sense of capability and confidence.
Social Plugin