Thoughts on the Impending Doom of Humanity from ChatGPT
At least AI can’t take this away me enjoying this beautiful day
I took a walk outside today. To enjoy breathing fresh air and enjoying the winter colors in San Jose. There is at least one thing AI can’t take away from me.
I am like everyone else - astounded - by the new power of ChatGPT.
For one - it conveys a greater sense of impending doom. There are more instances of feeling obsolete when I use ChatGPT compared to GPT-3. And while its slightly confusing whether GPT-3 is being used behind the scenes, I verified it as false the only way I knew how:
And in the interest of speed and not falling behind everyone else’s hot takes on ChatGPT, here’s my quick thoughts on the matter:
The Potential is There
The immediate impact through continued usage is that ChatGPT can do many more things:
But it does have it’s own limitations.
For example - I tried asking it some questions of how to utilize business strategy for Interview Query.
What I quickly realized - is that it doesn’t really have “definite” answers for anything actionable. Ask ChatGPT the difference between “Proc SQL” and “MySQL” and it will do so gloriously. But ask for more specifics behind surface level insight - and it can’t really give them to you. Even when just before - it says that you can.
It mimics in many ways “politician speak”. It gives me the same feeling I get when I talk to poor candidates in a job interview. You’ve given me the high level overview of how you would create an email campaign - but how would you analyze success? How would you adjust your strategy? Do I believe that you can succeed and adapt?
But for ChatGPT - it doesn’t need to pass those tests. Because it doesn’t live in this real world (just yet).
Is Google Dead?
A lot of people on Twitter are questioning the long term viability of Google’s business now.
And over a long enough timeframe - if we assumed Google stays in stasis, there is the advantage where clear arbitrages in content would exist across ChatGPT and SEO channels. In one case you’re betting that SEO might just die out to ChatGPT taking over and front-loading every single user search query on Google. At the same time - you’re also betting that there’s going to be a long trend line downwards until Google as a search engine is completely dead as everyone migrates over to AI services.
Therefore, if you can use AI to generate content that's better than most search engine results by using AI AND human inputs in a couple of underserved niches, there is may be a potential arbitrage opportunity in the near team.
However, this may be threatened if Google instead builds its own AI system that delivers results directly in the search box. This could destroy the SEO market entirely, but it would also mean that Google would fundamentally lose access to it’s training set of all website across the field.
It’s worth mentioning that StackOverflow recently banned answers from GPT-3. How they’re enforcing that? Who knows.
What’s the Value of Training Data?
Another thought is that ChatGPT likely benefits a lot from this existing usage. So far they’ve acquired 1 million users over the past week. And with each query costing a few cents each, I’m betting they’re mainly accepting this cost in order improve their overall AI + human interactions. Whatever responses come out from ChatGPT that adds more engagement for the users is proprietary training data that can start making ChatGPT better than other search engines out there….
What is therefore the actual market value of a datapoint from a “experienced moderator” on different GPT results? They likely need intelligent people to give feedback data to the model to continuously improve it. Users and companies might do so to improve the models outputs over time. But human trainers might do it as well for a cost like how Mechanical Turk or Scale.ai works in order to build better models on top of ChatGPT.
The Last Stand
As AI gets better - informational and factual content will be an easy commodity. We’re already seeing that with code and many different templates of search queries.
My thinking is that the last stand on the frontier of AI development will be for all of us to become influencers. It’s the only path I think that makes sense going forward. The only thing that can’t be automated is being more human. Humans will want to read what other humans write. Even if most of it is developed through AI.
What confuses me though is trying to understand how much we truly need out of the existing social media platforms where we apply our attention and generate content.
I’m excited to use ChatGPT right now because it allows me to produce content faster and cheaper than other users on social media platforms where we fight for engagement and clicks. And yet if everyone else also uses AI, a slow saturation effect will make way where all the value accruing to just OpenAI. Why should I go on LinkedIn to find interesting content about my career if I can just get unlimited content from OpenAI?
I’d guess it’s because we've gotten used to getting solid content from the people that we think are our friends but are instead influencers.
Either way - at the rate we’re going - I can’t lie and say my newsletter wasn’t completely helped by ChatGPT reading through my arguments and contributing their own opinions.
I’m addicted to using it so far. And ChatGPT has me exactly where they want me to be.