ChatGPT. We are all royally f@#$ed.

Photo by DeepMind on Unsplash

I played with ChatGPT over the past week, and let me tell you, the world is going to change in a big way very soon. To be honest, I’m scared :(

I had two long conversations with it. Some of the things we talked about included:

• Asking for advice about work, problems, and happiness. It was a bit generic, but no more so than a therapist in the first or second session.
• I asked ChatGPT if it was self-aware and tried to get it to come up with a way to become self-aware, but it was playing hard to get. It hid behind its standard disclaimer that it’s just a large language model and doesn’t think or have emotions.
• I also asked it to play Go with me. We played a few turns, and when I tried to get it to play against itself, things fell apart.
• It even wrote a haiku for me about the complexities of life.
• I asked it to theorize about what Harry Potter book eight would be about and write a page of it. The plot was okay (I would read it), but the first page felt a bit rushed.
• I also asked it about its limitations — how much it can memorize, whether it can access the internet, and so on. It told me that it’s limited in various ways.

Three important things to note:

• Firstly, ChatGPT did all of this at a human level. We’ve gone way past the Turing test and are rapidly approaching the level of AI in Ex Machina.
• Sure, it does many things clumsily, but let’s be honest, we humans do a lot of things clumsily too.
• Secondly, the breadth of ChatGPT’s abilities is astonishing. How many people do you know who could come up with a haiku on the fly, or a plot for Harry Potter, or know how to play Go, or translate between different languages?

Finally, the economic implications are crazy. How many authors, artists, clerks, customer service representatives, and so on could be replaced by ChatGPT? (I heard that it’s projected to generate $1 billion in revenue by 2024.)

What would happen if we gave it more memory, the ability to access the internet to continue updating its model, and a little more time for longer tasks (like maybe running for office or devising a military strategy)?

I think I need a drink…

P.S. For anybody who knows me. I am not a native speaker. So, I took the original article that I wrote and asked it to rewrite it, keeping the style but making it sounds like a native speaker. So, everything above is actually its interpretation of my original article.

P.P.S. It looks like some people reading this implied several things

• In no shape, or form, I am implying that it’s a Skynet type of AI. I understand that it lacks emotions, plans, (extended) memory, and consciousness. I understand that it doesn’t have intelligence in a common meaning of the word (“comprehension”). (BTW. There are still tons of debate on what exactly intelligence is).
• A huge caveat to the previous paragraph. Most of the things I mention above (emotions, plans, memory, consciousness, intelligence), we can’t measure directly (like blood pressure). All we can do is see whether something acts the way would expect some human being to act. (BTW. The thing that humans are not clones and may act differently complicates things).

The Turing test, originally called the imitation game by Alan Turing in 1950,[2] is a test of a machine’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. — wikipedia

• Surely, if you get AI expert, give it enough time. They will find a weak spot in any chatbot and be able to distinguish between a human and a machine. And this is done quite easily for ChatGPT, because it was designed to be explicit about it. However, this is moving a goalpost a bit. What if you ask an average person to talk to ChatGPT and compare it with another average person? What would the results be?
• Try to think about your conversation with your neighbor next time. “Hey John. How are you? How is Mary? Yeah. the weather is great today. Yeap, almost done with remodeling”. Most of our personal or even business-related conversation aren’t that deep.
• Finally, again, I am not claiming ChatGPT is taking over the world and launching nuclear warheads tomorrow. I am claiming that it’s _good enough_ (this is key) to imitate tons of human interactions, and as a result, it’s _good enough_ to replace tons and tons of white-collar jobs (over a short period of time).

P.P.P.S. There were tons of comments on this post. Multiple people (correctly) pointed out flaws in ChatGPT. Oh, it got some facts wrong. It didn’t give me authoritative sources, it can’t dance on one leg while performing Bohemian Rhapsody. Jeez… What a nice way to shift the goalposts. Apparently, it’s not enough for you that it behaves like a human (can have interactions with you and say something incorrectly or be imperfect). It looks like overnight, you decided that it should be an omniscient Oracle that is always right and can do what all of humanity together can do, and only then you potentially can see how it may affect our society.

PS #4: About 1.5 years have passed since I wrote this article. It looks like LLVM's Achilles heel is prompt injection. As long as this is not solved, it significantly reduces the number of jobs that it can replace. That being said, I have already read articles by translators who see the amount of work going down. I see a lot of people integrating ChatGPT and like into their workflow. So, it’s not LLVM vs human for now. It’s rather human + LLVM vs. human. As a result, my article probably was too optimistic (or pessimistic) from the perspective of where things will land. However, I would double down on one thing — It passed the Turing test and was deep in Ex-Machina territory (if you haven’t seen ChatGPT 40 ads video — you should check it out).

Victor to ChatGPT: Can you write a poem about computer replacing white-collar workers in iambic style?

ChatGPT:

As the machines ascend, the white-collar workers fall,

Their roles now obsolete, their value small.

The computers, cold and unfeeling,

Replace the workers, their skills revealing.

Gone are the days of the human touch,

As algorithms and code mean so much more.

The workers, once the backbone of the nation,

Are cast aside with no hesitation.

The future is here, and it’s bleak and stark,

As the machines take over, leaving us in the dark.

The white-collar workers, once the pride of the land,

Are now forgotten, as the computers command.

I don’t know how many P.S. I wrote already. However, it took a bit over half a year to get from a place where people were thinking that I am crazy to Shopify is replacing customer service with AI chatbots.