Chegg Inc. (NYSE: CHGG), an education company in California that provides homework help to students and online tutoring services, could be another company facing an existential threat from ChatGPT and other language models.
Chegg is most known for operating a question-and-answer (Q&A) site where students can post their exam and homework questions and wait for expert answers.
Subscriptions for the service cost between $9.99 to $19.95 per month and are capped at 20 questions per month for its popular Chegg Study plan.
But with ChatGPT, students can get answers to their questions instantly and for free. Students type in their questions, and the chatbot delivers an eloquent and comprehensive answer, sometimes better than what a human tutor could conjure.
Due to ChatGPT’s popularity, service disruptions are the norm, leaving users locked out from having their questions answered. It costs OpenAI, ChatGPT’s creator, around $3 million per month to keep its language models online, which could partially explain why a session count is enforced.
This raises the question of what other platforms ChatGPT could threaten as a viable substitute.
ChatGPT is a formidable competitor but with its share of problems
ChatGPT’s obvious threat is to other Q&A sites, such as Quora and Stack Exchange.
There’s evidence that users are already starting to favour these tools to answer questions instead of pouring over a 300-word answer by hand.
If you look closely enough and are familiar with the tone and cadence of ChatGPT’s distinguishable writing style, you can see that many users are already copying and pasting the output from these tools as answers to questions asked on these platforms.
Many of those answers receive applause in the comments and hundreds of upvotes from oblivious readers, partially due to how confidently it constructs its prose. But that confidence may also beget humans from fact-checking the machine’s sources and sometimes its dubious thought process.
It’s not uncommon for ChatGPT to make up imaginative facts and statistics to suit its arguments or famously output code that’s brilliantly written but contains fatal errors.
ChatGPT’s coding errors are so common that Stack Exchange enforced a temporary ban on users posting output from the tool on its platform last December.
A threat to search giants
This gatekeeping against artificially intelligent (AI) content has become necessary for many platforms due to the velocity of spam content being churned out by these tools on every topic imaginable.
Google took a step with its helpful content update to protect its index from language models last year.
One of its pillars is prioritising content “written for humans, by humans,” which in practice means penalising AI content and rewarding human content that delivers a satisfying experience.
The search giant’s play of devaluing ChatGPT’s output could be as much of a quality-control process as warding off one of the biggest competitive threats it has seen in the last few decades. Google is essentially a middleman or conduit between the searcher and their answer.
By contrast, ChatGPT gives the user their answer directly and instantly. ChatGPT’s modality reduces user friction and works within people’s rapidly declining attention spans and higher demands for instant gratification.
Microsoft has quickly jumped on OpenAI’s disruption of the online search and advertising market with a $10 billion investment announced earlier this week.
Part of the agreement would see both companies commercialize their technologies, which could mean a joint platform on the cards that could threaten Google’s supremacy in the search space.
However, ChatGPT’s problems with users needing to trust that its often-unreliable output is correct will stand in the way of becoming a threat in the short term.
It’s also unsuitable for performing many types of searches. Still, it excels at the Q&A types, which could evolve into becoming a behemoth, regardless.
Around 8% of Google’s total search volume is in the form of questions, with around 700 million daily queries.
By even capturing a fraction of those questions, this would see ChatGPT become one of the most popular sites on the internet and will likely evolve into its own ecosystem, as we’ve previously observed with social media sites such as Facebook and YouTube.
As the parameters of these language models expand exponentially, they will become more accurate.
Parameters are how much storage the language model has to work with and is largely what determines the intelligence of its neural network or synthetic brain.
In the very short term, sometime this year, GPT-4 is expected to release which has 1 trillion parameters. GPT-3.5 currently powers ChatGPT, which has around 175 billion parameters.
So very soon, we could work with an AI content tool that’s around 80% stronger than what we have today, and this is still in the very early stages of these language models.
They will one day be indistinguishable from expert writing, with more creativity, accuracy, and ability to perform tasks previously only thought possible for humans, such as qualitative analysis and empathy, accomplished through deep and machine learning models and the astronomical leap in processing power via quantum computing.
General AI could one day be the civilization-defining moment where code improvements don’t need to be made by humans but by the AI improving itself.
This then leads to the popular but eerily achievable tropes found in science fiction of the emergence of machine consciousness, and ultimately leading to a singularity event when the ‘robots take over,’ so to speak.
Whatever may happen, AI’s rapid and accelerating progress should always be appreciated as one of the most disruptive forces many of us will see in our lifetimes.