I must admit that like most people these days, I am quite busy thinking about the impact that AI will have on our lives. Imagining the extent to which it will transform our life is quite difficult because there are many variables that even those who develop it can’t envision yet. While admiring the capabilities that are already on display, I decided to try to discern the impact that LLMs (ChatGPT) are going to have on education based on what we already know and the things we fail to see.
Since the release of ChatGPT in November 2022, districts, schools and teachers have been panicking about the threat it poses to education as we know it. ChatGPT or other tools like it can deliver an entire piece of writing based on simple questions or requests and teachers are concerned that students will be handing in AI-generated essays as their own. The easiest solution that quite a few districts decided to immediately adopt was to ban it from the district/school, but that would not be a lasting solution because students can circumvent the ban by using their cell phones or computers at home.
If You Can’t Beat Them, Join Them
As it turns out, the ability of a teacher to identify whether submitted work was produced by a Chatbot or a student is quite limited and its accuracy is in doubt. A more feasible approach lies in the saying “If you can’t beat them, join them.” Attempts to see how ChatGPT can help teachers proved to be much more beneficial for students and teachers alike. Teachers who spent some time thinking how they could use ChatGPT to improve student learning came up with very creative solutions that students enjoyed. Using ChatGPT or other tools like it successfully requires a commitment to change the traditional method of teaching and adopting a new method that requires time and effort that not all teachers are willing to undertake.
At first it may be deceiving. ChatGPT is very good at answering questions, almost like you’re talking to a person that has spent hundreds of years absorbing knowledge. Its output is fluid and grammatically correct. However, ChatGPT’s answers are not always correct. In fact, it often hallucinates and states completely wrong facts. It behaves like an advanced autocomplete engine. It takes your prompt and chat history and tries to predict what should come next. And it doesn’t get things right, even if its answers mostly look plausible.
And while it is very tempting to adopt the answers and the content of essays that ChatGPT or other tools like it generate, we have to remember that there is currently no way to tell hallucinations from truths in ChatGPT’s output, unless we verify its answers with some other source, such as Google Search. This is in complete contrast to the opinion of those who claim that A.I. chatbots such as OpenAI's ChatGPT could make typical search engines obsolete within two years. Even the Gmail creator Paul Buchheit predicted in a Twitter thread that the future of web search as we know it right now could be completely upended by 2025.
Will It Become Obsolete
There are several reasons why search will not become obsolete in the future. ChatGPT provides plain text with no reference to the actual websites. To solve this problem, some companies are experimenting with adding a mechanism that links different parts of LLMs output to actual web pages, which seems to be a complicated solution because it would require access to another source of information, such as a search engine index database. This is one reason why classic search engines are not likely to lose their relevance anytime soon.
Another challenge that ChatGPT or other tools like it face is updating their knowledge base. Search engines have the tools and software to constantly index new and modified pages, which is also a very cost-effective operation. In comparison for a Large Language Model (LLM) to be updated, adding new knowledge requires retraining the model multiple times a day, which will be much more expensive than adding and modifying records in a search engine database. Also the operation of determining which web sources are reliable sources that should be given priority relies heavily on tools and mechanisms of search engines.
There are also speed challenges. Google has created a highly optimized database infrastructure that can find millions of answers in less than a second because they don’t need to browse their entire dataset for each query. Google’s indexing, sorting, and search algorithms can pinpoint the right records at very fast speeds. So even though the body of online information is growing, search engine speed is not dropping. On the other hand, LLMs run information through their entire network every time they receive a prompt, requiring a substantial amount of computation that is a lot more than querying indexes. Therefore, in comparison to Google Search, LLMs such as ChatGPT take several seconds to compose their responses.
LLMs which are in their infancy lack a business model and the cost of operating them is huge. In comparison, there is a solid business model for search engines such as Google, whose share of the online search market is so large that even with very low click-through rates, it makes billions of dollars yearly.
The Impact On Education
When we try to translate all this to education talk, the bottom line is that when students are using tools like ChatGPT, they are not able to certify that the information it presents is accurate. They will also not be able to point to the sources that ChatGPT took the information from or be able to cite their sources. Because LLM tools scrape content from a wide variety of sources, the material produced is a mixture of many other people's work, and there's no credit for creators. Students may have to risk accessing the latest information because the cost of updating its huge database will make it infeasible to do it frequently. Not to mention that LLMs can only learn from its source(s), so it takes on the biases, misinformation, and problematic content of the original material, which may be reflected in the student’s work.
To remedy all these issues, students and teachers may need to turn to the old search engine. Not everything that Google or other search engines provide is truthful either. But at least they provide you with links to sources that you can verify, reference its sources, have speedy access to information and rely on getting the most updated information out there. Google Search is flawed. It shows a lot of useless ads. It also often returns a lot of useless results. But it is an invaluable tool and it will not become obsolete anytime soon. It is more likely that ChatGPT or other similar LLMs will become complementary to online search engines.