The public quickly became fascinated by the power of new artificial intelligence technology – ChatGPT – a chatbot developed by research and implementation company OpenAI and launched late last year. It has demonstrated its ability to provide detailed answers to complex questions while using the information it processes and feedback from users to improve its responsiveness.
ChatGPT has proven versatile as users use the technology to compose music, debug computer code, write restaurant reviews, create advertising copy, and answer test questions. It can offer feedback in a conversational way and has sparked excitement about its potential, along with some concerns about how it might be used.
But what exactly is ChatGPT and what does it say about the state of AI now and in the future? Today at Elon reached out to Assistant Professor of Computer Science Ryan Mattfeld, who has been using ChatGPT as a discussion point in the classroom and has insight into how it can transform our use of AI. ta.
When did you first hear about ChatGPT and what was your first reaction?
ChatGPT was launched on November 30, 2022, and I first heard about it in early December. The first sample I saw was a piece of software written by ChatGPT, including an explanation of how the code works . I was immediately skeptical about these possibilities and initially assumed that this example had been carefully selected. I continued to hear more about ChatGPT, so at the end of the semester, I created an account and started playing around with it.
I was immediately blown away. I gave it an assignment used in one of my 2000-level courses last semester, which included a specific programming task embedded in a story/scenario about creating out robots. These are many long paragraphs, with unnecessary information and class references that ChatGPT doesn’t know about. I think it will make it difficult for ChatGPT.
I’m wrong. It instantly gives an exact solution, includes an explanation for the code written and a description of how to solve the provided problem. I quickly emailed everyone in my department to inform them that ChatGPT was real and requested their attention.
People are used to search engines that can direct them to answers to their questions or provide resources to increase their knowledge. Why did ChatGPT make such a big leap?
ChatGPT and search engines have two different goals. The main goal of a search engine is to try to direct you to the correct resources. The main goal of ChatGPT is to generate reasonable responses to natural language inputs. The most important difference is that the main goal of ChatGPT does not include accuracy. That’s certainly a secondary goal, but it’s not a guarantee.
That said, there is definitely overlap. Part of the main goal of ChatGPT includes condensing multiple data sources into a useful response. So when ChatGPT provides accurate responses, it simplifies searching, allows for dialog, and provides answers in a conversational manner. Instead of looking for a list of websites, searching for those related to your topic, opening the website, finding the specific information you want, and regularly backing up and trying again, ChatGPT can provide feedback using easy to understand language. clear, direct and helpful.
In addition, ChatGPT also allows you to track. If you want to dig deeper into part of its response, you can. Just ask for a more detailed answer on the part you find most interesting.
Can artificial intelligence technologies like ChatGPT learn from the input they receive and respond to the responses?
Correct! In fact, this is one of the main mechanisms in its development and continues to be used for refinement.
Modern machine learning solutions rely on large amounts of raw data. The same is true for ChatGPT – its initial training is done on large amounts of data collected on the internet. However, as OpenAI, ChatGPT’s creator, describes on its website, humans are deeply involved in ChatGPT’s training process. After initial development of the massive data set, ChatGPT responses were fine-tuned with human labeling. The process begins by asking one person to ask a question and another person to provide what they believe would be a good answer. This information was compiled and another round of adjustments began. In this step, the AI gives several different answers to a question and the labeler ranks the AI’s answers from best to worst. This feedback has also been integrated into the network.
Now, when you use ChatGPT yourself, you will see options to give feedback that it generates “thumbs up” or “thumbs down.” If you select “dislike,” you can describe why you don’t like that answer. This is used to further refine ChatGPT.
What are some of the ways you see ChatGPT being put to use now, and how do you think those uses will expand as the technology becomes more advanced and integrated?
Of course, working in the field of computer science, my first inclination was to look at how ChatGPT started being used in software development. It is surprisingly effective at writing good code for relatively simple programs.
However, it’s not perfect. It still makes mistakes and can produce non-functional but beautiful code. So I expect that the most successful software developers will use it as a helper but will need to be able to review feedback to identify and fix errors. There is still a great need for knowledgeable people in the loop to verify that the solution provided is correct rather than convincing but wrong.
Another interesting perspective for computer scientists is that ChatGPT is an open API, which means software developers can integrate ChatGPT into new coding projects we develop. This means that we will soon see ChatGPT integrated into a lot of other programs and applications. I expect that ChatGPT will improve over time, and I hope that we’ll see it in more and more places. For example, ChatGPT is already integrated into Bing and will soon be integrated with Microsoft Word.
Of course, ChatGPT is also starting to influence fields beyond computer science. I created an assignment for the Technology and Society core that I taught in January that involved ChatGPT. This assignment involved using and analyzing ChatGPT in a variety of ways, including a contest in which students tried – and largely failed – to determine which articles were written by their friends and which Which article is written by ChatGPT. The final question in the assignment asked them to consider how ChatGPT could influence a specific area of society with supporting evidence. In particular, law, computer science, journalism, and education are some of the fields where many students predict ChatGPT will make a significant impact.
Like many technologies, there are advantages and disadvantages. Do you have any concerns about how technology like ChatGPT will be deployed?
There are actually a number of ways to look at this question. It has many positive aspects. It can make apps easier to use for everyone. Instead of requiring the specific and specifically formatted requirements that computers often need, the use of a common language will soon have a much more widespread effect. This also has positive accessibility implications in supporting people with language difficulties. Of course, there are all the other cool things it can do — from playing games to helping eliminate writer’s block, to explaining complex concepts in simpler terms, to simulate a conversation.
Of course, there are also some disadvantages and risks. More broadly, if someone relies on ChatGPT too much, it can hinder their growth. Making mistakes is very important in the learning process. If ChatGPT is used to skip this step, it can lead to a false sense of ability and confidence.
In addition to this indirect but very important concern, ChatGPT can also be used directly to cause harm. Previous versions of ChatGPT and other synthetic AI (the broad term for this type of technology) have been in development for many years. Previous iterations were taken directly from information available on the internet without filters against hatred, lies and prejudice. Part of the reason ChatGPT is successful is because it is currently the most successful tool at reducing harmful instructions, misleading content, and inaccurate information. However, none of these disadvantages can be completely eliminated. ChatGPT is very effective at creating convincing stories, even if they are not true.
I am most concerned about how ChatGPT could make life easier for those who seek to spread misinformation widely. The notion that we are or will soon be in a “post-truth world” scares me.
Google has now announced a competing technology – Bard – and others could enter the space. How much attention are tech companies paying to developing this type of AI?
In particular, Google is extremely interested in ChatGPT. Their business model is almost entirely based on collecting advertising revenue, which is driven by the number of people using their services. They have diversified into several areas across Youtube, cloud computing and other areas. However, if ChatGPT replaces even 50% of Google searches then I would guess that Google will lose about 30% of their total revenue.
The tech industry knows that Generative AI is coming soon, but ChatGPT has changed from “soon” to “now,” so other companies working on these technologies are now extremely focused on its release. their version. It would be interesting to compare the effectiveness of each synthetic AI option. I suspect that many will be released before they are ready.
You have used ChatGPT. How have you used it or have you seen others use it that surprised you?
I was first surprised by its coding capabilities, but this is a narrow application. The more I use it, I have to say that it surprises me most with its versatility and ability to hold a dialogue. For example, I used it to create a text-based adventure very similar to the choose-your-own-adventure books I remember loving as a child. You can choose a theme or theme either for your adventure or let ChatGPT choose. Then you can make decisions and change the way the story develops. Additionally, you can ask it to explain quantum computing in a way that a 10-year-old can understand, and it does quite well. You can provide your own writing sample and have it answer questions in your own writing style. You can ask it to write a short story in the style of your favorite author.
I think it shines in its ability to support struggling writers or other situations where you’re struggling. For example, my three-year-old daughter was invited to her five-year-old friend’s birthday party. I was having trouble coming up with gift ideas for a 5 year old so I asked ChatGPT for suggestions. It offers a great list of diverse ideas and even adds:
“Remember that every child is different and what may impress one child may not happen to another, so it’s always a good idea to consider your child’s interests and personality when choosing a treat. gift.”
I knew nothing about this five-year-old’s special interests, so I asked for more choices in the category of Lego and building blocks, which seemed to me the most interesting. For my next request, ChatGPT provided specific options including classic Lego sets, Duplo Blocks, Mega Blocks, magnetic building blocks, and themed building sets. It describes each option in detail and when one set may be more suitable than another.
This two-minute exchange probably saved me 20 minutes of searching websites for options, and helped both provide more original and detailed options for gifts. Specific donations that interest me.