Is ChatGPT the new Google?
There have been numerous speculations about how ChatGPT could replace Google. To that point, this blog is a non-technical analysis of how the two tech companies with great technical prowess differ and whether ChatGPT will replace Google, if at all.
Did you know that the first chatbot was created in 1962 by MIT professor Joseph Weizenbaum? It was called ELIZA.
From the 1960s to the introduction of Siri in 2010, Google Assistant in 2012, and Cortana in 2014, it took roughly 5 decades for us to interact with AI-powered personal chatbots. The major boost has been the normalization of smartphones.
However, it only took less than a decade and we are already interacting with a chatbot that is usable by any normal human being with the arrival of ChatGPT 3 aka GPT-3.
While chatbots have become a norm in the recent past, the scope of the interaction had been limited to the business it was serving. Think of customer service chat you might have had with Lyft, Spotify, Whole foods…etc. On the other hand, Siri and Google Assistant have been more powerful but were still limited to assisting with the day-to-day tasks of the user.
So, how is GPT-3 different?
You can use GPT-3 to write a poem, use it to debug your code, write a blog, etc. It is like interacting with a superhuman who can help you do pretty much anything you could imagine doing using a modern smart device.
GPT-3 is the latest ChatGPT version and it comes with enormous capabilities. While the data set used for its training has not been publicly disclosed, it is reported that GPT-3 has a whopping 175 billion parameters and was trained on about 45TB of text data from different datasets.
That is enormous! Maybe we can call GPT-3 a super-chatbot and it is only going to grow stronger. So, will it, ever, replace Google search?
It is highly unlikely that GPT-3 will replace Google search.
A few more basics before the conclusion. Let us understand how Google Search and ChatGPT work as in to help us with the comparison.
Google Search
Google, with its mission to organize the world’s information and make it universally accessible and useful, is used to discover a broad range of information from a wide variety of sources over the internet.
Each Google search goes through 3 major stages
- Crawling
- Indexing
- Serving search results
Google does not directly answer your search query, well it does in its own fashion, but in the context of this discussion, it actually fetches the information that answers your question such as web pages, images, videos…etc.,
It is important to understand that the Google Search index contains hundreds of billions of web pages and is well over 100,000,000 gigabytes in size. Apart from the 3 major stages, Google also Ranks the results and detects spam.
The Google search engine uses complex algorithms and ranking systems to search through billions of web pages in its index and return the most relevant results. Since 2015, RankBrain, which is a combination of AI and Machine Learning, has been part of the search algorithm.
Have you ever done some vague search on Google, and been surprised when the search engine still returned accurate results?
That is because of RankBrain. If I search vaguely for ‘the movie with plane dropping action sequences’ the search returns ‘Furious 7’.
Pretty cool!
ChatGPT: Generative Pre-trained Transformer 3
GPT-3 does not have a specific mission or purpose in and of itself. GPT-3 is a large language-processing artificial intelligence model developed by OpenAI. It is not a standalone product or service, but rather a tool that can be used by developers to build natural language processing applications such as chatbots, machine translation systems, and other language-based technologies.
It has the ability to generate text that is highly human-like in style and structure.
In simple terms, GPT-3 is an AI chatbot that is pre-trained on enormous datasets. It relies on the information and knowledge that it was trained on to generate responses to user inquiries. Let’s read it again and note that “it relies on the information and knowledge that it was trained on”.
Also, note that GPT-3 does not have the ability to browse the internet or search for information online and It does not have the ability to access new information or browse the web like a typical search engine.
GPT-3 is all ‘text-in-text-out’. You won’t get results in formats other than text, unlike Google search.
Now, let us try to search the same vague term ‘the movie with plane dropping action sequences’ in GPT-3.
GPT-3 does not get what the user might be looking for, hence the text results in this case do not exactly render the expected results.
In the other similar cases that I tried on both Google and GPT-3, it was evident that GPT was not built with the purpose to become a search engine.
Conclusion
The users of the likes of Developers, Engineers, Creative writers, Analysts, etc., look for more definitive and comprehensive answers and would love to avoid browsing through an endless list of search results. GPT-3 serves this exact same purpose.
We have already read on the https://openai.com/blog/chatgpt/ that it helped debug a code snippet, but to prove the aforementioned, I asked it to help me with a terraform script to spin off a Kubernetes cluster on the Google Cloud Platform.
This task is purely technical and requires Terraform coding but GPT-3 aced here.
If the current capabilities are extrapolated with the potential capabilities that GPT may acquire in the future, it still seems like ChatGPT will not replace Google Search.
However, It will certainly take away a portion of Google’s users.
Critical Differences
There are two major differences between Google and chatGPT
- GPT Can’t connect to the internet hence the results are somewhat limited
- While the current usage of chatGPT is free, it appears like at some point one will need to pay for it
Finally, let us ask GPT-3 itself if it intends to replace Google 🙂
As ChatGPT puts it, it is not designed to function as a search engine and is not a replacement for Google search.
Points to Ponder
It is astonishing how almost no internet user thinks the actions taken on a computer or smartphone to browse the internet could create a carbon footprint!
In a 2019 study, researchers at the University of Massachusetts estimated that training a large deep-learning model produces 626,000 pounds of planet-warming carbon dioxide, equal to the lifetime emissions of five cars. As models grow bigger, their computing needs are outpacing improvements in hardware efficiency.
A 2021 study estimates that training of GPT-3 produced roughly 552 metric tons of carbon dioxide. This is about the amount that 120 cars would produce in a year of driving
Based on the assumption that Google.com “processes an approximate average of 47,000 requests every second, which represents an estimated amount of 500 kg of CO2 emissions per second.” That would be about 0.01 kg per request. These numbers are approximations.
In fact, in a 2009 estimate, Google said each query causes 0.2 grams of CO2 emissions.
It is also worth noting that Google has committed to using 100% renewable energy to power its operations by 2030, and has already reached an average of 52% renewable energy across its operations.
References:
