ChatGPT vs Search Engines

If you’ve read a bit about ChatGPT and other generative AIs, you’ll probably know a little bit about what they can do: write articles, help with school, and generally look stuff up for you. In this last way, they’re remarkably like search engines, but can we compare ChatGPT vs search engines? Will AI one day replace Google, Bing, and all the others?`

ChatGPT vs Google Search: how they work

Before we can answer any questions about their efficacy, let’s first see how the two systems work. Google is the one you’re bound to be most familiar with: practically anybody who has ever used the internet in any way has entered a search query into the blank box and then gotten a page of results to choose from.

How Google works is worth a few dozens of articles all by itself, but the short version is that you enter a search term and Google then starts looking for web pages that have the same term in them, known as a keyword. Using a specialized algorithm, a very secret one, Google then ranks these pages in relevance to your search.

That’s pretty impressive if you think about it for even just a second: billions of pages, indexed in just a few milliseconds. Still, though, when interacting with Google a lot, like say when doing some research on trying to find out more about a tricky topic, it can get frustrating.

If you don’t type in exactly what others are writing, you’ll not get the pages you want. What’s worse still are all the sites that try and game the system, offering sub-standard info, but using the right keywords to rank well.

How ChatGPT works

Generative artificial intelligence like ChatGPT works very differently: when you ask it a question, it doesn’t search the ever-changing web, but instead a cordoned off set of data which contains nothing after September 2021. The team at OpenAI, the people behind ChatGPT, decided to do so to make sure they can control their chatbot better.

This is likely due to what happened when other chatbots were exposed to the web at large before being trained. Micosroft’s Tay, for example, became a racist nightmare after just a few hours. By training their chatbot on a limited set of data, OpenAI can prevent this from happening, or at least minimize the chances.

Besides this limited data pool — though limited only compared to the entire expanse of the internet — there are a few other key differences between Google and AI, the most important of which is likely how you get your information. Where Google presents you just a massive list with what may be relevant, ChatGPT actually answers your question, and in an almost conversational tone, to boot.

It does so pretty fast, too, and, better yet, can remember what you asked it before, meaning you can “chain” questions, zooming in on the answer you want by increments. Google has no memory, or at least not one you can use it this way. As you can imagine, this gives ChatGPT a lot of added utility. Still, there is a strong case to be made for Google when compared to ChatGPT.

Google Search vs ChatGPT: which is better?

There’s no doubt that ChatGPT is a very attractive option: ask it a question, and information pops out without you having to go find it like with Google. However, this convenience comes with a price, and a massive one at that. This is simply that ChatGPT cannot be trusted, and for two reasons.

Hallucinations

The first is that ChatGPT makes stuff up. Called hallucinating, whenever the chatbot doesn’t know the answer to something, or even when it’s just not sure, it’ll just spit out a random answer with massive confidence. Its lies get detailed, too, like making up the names of people who founded a company, or lying about how many wins a certain sports team has.

To its credit, OpenAI has been hard at work on this issue, and when comparing GPT-3.5 to GPT-4 the number of hallucinations has been brought down considerably. Still, there is a non-insignificant number of them, and you never know when the AI is lying to you. You have to double check everything it says.

Poor Data

The constant rechecking of what the AI is telling you isn’t just because of hallucinations, either: even when it’s telling the truth, ChatGPT sometimes spouts nonsense. This is usually due to the sources it was trained on: selected by humans, they often reflect the biases of those humans.

For example, ask ChatGPT a question about history, and you’ll get the “standard” version of events, usually either taken off Wikipedia or some other sources. Though there isn’t something directly wrong with this, it can often bypass secondary reasons, or disregard narratives that are outside the mainstream. It being history, the mainstream often feels like it’s a decade or two behind at the best of times.

Of course, this isn’t limited to history, and plenty of other academic disciplines, as well as a lot of far less lofty ones, suffer from this issue. While the answers ChatGPT gives you may be correct, they often aren’t complete. Since it doesn’t give you its sources, there’s no way to check what it’s saying directly; you need to utilize your own research for that.

Making the Choice Between Google vs ChatGPT

As handy as ChatGPT may seem, you really need to check its work. While asking a question of an AI may seem like an amazing shortcut, the result may be far from what you need and may actually create more problems than it solves. Between its occasionally poor sourcing and its penchant for lying, you never know what you’re going to get.

At the same time, the issues with Google remain unsolved. Still, though, it’s the devil you know: while Google’s answers to your questions may be lacking, at least you get to pick and choose what you want to use, and you can check sourcing and reasoning for yourself.

That said, there are some interesting ChatGPT alternatives that aim to fix these issues: for example, Google Bard sources what it says fairly well, though of course you’re just getting summarized results from Google’s own web search. Other bots aren’t even that far yet, though with the way AI has been going, that may change very soon.

Until that day, though, we recommend you stick with Google for your factual information, or at least always fact check what ChatGPT is telling you: it’s just safer that way.