Generative AI Won’t Revolutionize Search — Yet

Let's see our search history. 

• Search 1.0 required people to enter a certain keyword or combinations to query the engine. 
• Search 2.0 allowed users to type natural phrases as if they were interacting with a human. 
• Search 3.0 (already between us right now) wants an answer instead of certain websites. Users want deeper insights and understanding. 

ChatGPT's first problem: it's not a search engine, because it doesn't have access to real-time information, just like the search engine does. ChatGPT was trained on a massive dataset with an October 2021 cut-off. This process gave ChatGPT the ability to understand and produce human language. 

Probably yes and probably no. Relying on LLM (Large Language Model) as the source of information is extremely difficult. The most challenging part is the tremendous amount of processing power needed to continuously train an LLM and the financial cost regarding these resources. 

Google covers the cost of search through selling ads, and it provides the service free of charge. If it aims to process queries at the rate Google does, then it will need a higher cost to pull off. 

However, there is still problem of actual information it will deliver: What exactly are tools like ChatGPT going to learn and from whom?

Even the curation that ChatGPT does in a massive online dataset, doesn't mean it is actually correct and free of bias. 

In fact, a study by Emily Bender, Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell (credited as “Shmargaret Shmitchell”) found that “large datasets based on texts from the internet overrepresent hegemonic viewpoints and encode biases potentially damaging to marginalized populations.” 

Bias is a problem with traditional search engines too, they can lead users to websites that contain biased, racist, incorrect information.

However, google simply guides users toward sources, it bears less responsibility for the content. Users can apply their judgment to differentiate the information they want to use. This judgment step is removed with ChatGPT because it is directly responsible for the biased and racist results it may deliver. 

Also, users have no idea what sources are behind every answer with ChatGPT and it won't provide them when asked. This creates even more dangerous situation where a biased machine may be used by certain users as an objective tool that must be correct. 

Because of this issue, ChatGPT and generic LLMs have to overcome a major challenge to be of use in any serious situation to find information or produce content, especially in academic and corporate applications where the smallest misstep could have catastrophic career implications. 

LLMs will likely sharpen certain aspects of traditional search engines, however, currently, it seems incapable of dethroning Google search. They could play a more disruptive and revolutionary role in changing other kinds of search. 

LLM stands for OpenAI’s Large Language Model

Search 3.0 will be the rise of purposefully and transparently curated and deliberately trained LLMs for vertical search, which are specialized, subject-specific search engines. 

In this rapid technological transformation, it's not a question that everyone got anxious about what the future would bring for us.