Summary: I believe that Chat GPT-3 like systems eventually would destroy the incentive system that encourages content creation on the Internet. We probably require a new kind of business model in order to empower writers, companies, community members & other types of stakeholders to actively create and promote their content on the internet.
With the recent advancement in large language models like Open AI’s GPT-3 & Google’s PaLM, AI-based chat agents are getting to the “sentient” realm. Google’s PaLM model is built with 540 Billion parameters and is said to outperform OpenAI’s GPT-3 model, which is trained with 175 Billion parameters. So the obvious question here is, why has google not integrated their model with the current search engine to provide similar or better results for the search? I quite honestly do not think it is a technical problem that they are facing, especially since a recent whistleblower from google claimed their AI has reached “Sentience”.
My hypothesis: I believe it is more about the business model and about the need to maintain the status quo of the Internet. If you look closer, GPT-3 like technologies fundamentally disrupt the way the Internet currently works. Currently, people/companies generate content to build thought leadership, build an audience, get potential buyers on their website, or generate revenue using ads. Chat GPT-3 based models crawl through millions of such websites, extract contents, understand them, and aggregate them on an event of a relevant question and serve it to the user in a short paragraph.
Related article – Getting started with GPT-3 model by OpenAI
This makes the technology extremely useful for end users. An average research on the internet that requires 1–2 hours of browsing 100s of websites and reading can now be replaced by a tool that can curate and summarise the required information and serves the user on a silver platter. This convenience is absolutely amazing for the end users!! But what about the content creators?
Since the chat agent uses information from 100s or 1000s websites to curate one response, it is impossible to provide references to the source websites. So, eventually, the traffic to the websites would decline drastically, resulting in no visibility, no ad revenue, and potential buyer acquisition opportunity for the content creators. So the obvious question is, why even create content in the first place?
Consider a scenario with forums like websites — Quora or StackOverflow. GPT-3 makes it easy for users to ask a programming-related question and get the response right in the chat screen; the user takes that code or solution and goes back to his IDE to continue his task at hand. Whereas traditionally, the user would have searched for a technical question on google, and google would have shown them the relevant StackOverflow page that discusses that particular question. The user goes into the page and finds answers from different people; suppose our user just quickly got a new idea for a more optimized solution, and for the rewards from the tribe, he just adds his answer to the forum as well; voila!! We just got a new, better solution. This wouldn’t have been possible with the chat GPT-3 based solutions in the market, as users may not even see those threads going forward.
However, on websites like Reddit, Linkedin, etc., we may see a temporary surge in content as people may feel empowered to write content with the support of GPT-3 like services. But, I do think it is temporary as the content quality gets saturated, and as everyone would realize, they could find the same thing just with a prompt. Alternatively, this could also lead to the development of prompt-driven interactable insights/content.
Let us think about another content creation source: Companies, Individuals, thought leaders etc. Content marketing is one of the best ways to build thought leadership in your domain of expertise and helps potential buyers find you on the Internet easily. In a world with Chat GPT-3 like interfaces, there is very little visibility or credit creators get for their content. GPT would never correlate the number of articles a company has on their domain to verify they are experts in the space. It is smart enough to look at other parameters like company size, type, previous case studies, news articles, references across the website, discussions across various forums, etc.
For, e.g., ask Chat GPT, which are the best blockchain development companies in the world, and it would give you the following big names like IBM, Consensys, and Accenture. An average user would be pretty content with this response and will get back to his work. So, the real question is, what is the incentive for startups and SMEs to generate content?
As an example, consider TechCrunch: People usually land on the TechCrunch website via google when trying to find more about a specific tech event, startup, or any related topic. For, eg. let us ask Chat GPT the question, “what happened to FTX? why did they close down?” Chat GPT-3 would give you a response of 200–400 characters, clearly explaining what happened without any fluff; 20 mins job reduced to a few seconds!! But from TechCrunch’s perspective, they lost a visitor and potential revenue-making opportunity. So, why would Techcrunch want to create any content in the future?
We could argue that OpenAI could pay TechCrunch for the content the AI agent uses, but do remember that this response from the agent is a result of content from 100s or 1000s of different websites. It wouldn’t even make sense for Open AI to pay all of them. So it is just a matter of time; before Techcrunch either goes to a completely paid subscription model or completely stops its free content development efforts.
This category of websites survives either with donations (like in Wikipedia), ads, or member payments. With ChatGPT answering user questions straight away, traffic to these websites could reduce by up to 80–90%, which may lead to a lot of these websites shutting down, not being able to meet the basic burn to keep the servers & staff up.
I believe these issues would lead to the fall of overall content development on the Internet by 60–70%. This drop in content creation will eventually lead to a catastrophic situation for the netizens and the GPT-3 like systems. Internet perhaps would turn into this slow-moving, niche, content-focused boring highway of Information for the netizens with very few choices on different new topics & discussions.
Whereas AI systems like ChatGPT require an extremely large amount of data/content to find & understand new topics better, with less content, the AI learning process would slow down and may eventually stop. Now we are in a situation where we have a dysfunctional internet with our information highway almost shut, and the ChatGPT that came as the savior is almost outdated and does not understand the new topics/questions well because of the lack of relevant content.
This is perhaps the worst-case scenario that can happen if we scale GPT3-like technologies to the masses without carefully architecting every minor detail around it. I do think that GPT-3 like technologies are nothing but true innovation, and there is no way to stop it, just like CRISPR or blockchain. However, It is very important for us to come up with a business model or user experience that shows justice to all the content creators of the Internet to ensure that the Internet strives with new content, discussions, and content creators popping up every single second.
Related article – How to implement Artificial Intelligence in your business?
So, coming back to the initial hypothesis, I believe google already has something similar or even better than ChatGPT. Still, they have to figure out an appropriate viable business model before integrating that into the search engine and making it accessible to the masses. Otherwise, this system may eventually turn into the Frankensteins monster and kill google itself by destroying its number one revenue source — Google Ads. It will eventually affect the well-being of the Internet itself. I hope Microsoft & OpenAI have considered these potential problems while they seem to be preparing to integrate ChatGPT into Bing.
So, is the solution a hybrid search? With AI answering questions & a bunch of results being listed below that explanation? I do not think so, an average user would mostly not want to go into the lower-level details, and the overall drop would still stay at 50–60%, and the user experience would also suck. It is like we are compromising so much on both that we are only left with the worst of the worst — a bad user experience with no specific utility.
Related article – How to tackle biases in Artificial Intelligence?
But this is not the time to get disappointed over the future of the Internet or the future of this technology, rather get excited about the kind of disruption we are going to witness and the opportunity of a lifetime it is going to present us with: To figure out a business/technology/user experience model for implementing general AI-based question-answering search technologies that empower content developers to create & share more content on the Internet.
Idea 1: (Partial solution)
Currently, Businesses rely on companies like google for marketing — to be able to reach potential buyers and catch their attention, and somehow get them to visit the digital storefront/website, which eventually gets added to the top of the sale funnel for a salesperson to pick it up and convert the deal. Current ML systems can understand a buyer better than a sales guy with the wealth of data they have at their disposal like demographics, context, intent, previous purchase history, email exchanges, internet activity, etc., and GPT-3 is more sensible and intellectual than 90% of the sales guys out there (including me), so what if they are able to provide sales as a service?
Companies can go into a store backend, add their services or products, and basically bid for deals — could be an e-commerce product, website views, a service, a hotel, or whatever, and an AI agent does nurturing & sales on behalf of the company/website. Already, for a lot of who are into technology, 90% of our reality is modeled by the internet. If AI agents like GPT-3 becomes our single gateway to the Internet, It can very easily nudge us towards an action that is beneficial for the system, like a future purchase, a vacation to a specific destination etc.