I’m actually quite tired of the current AI hype. And in my circle of communication researchers, there are many experts in the AI discourse. We now even have a subfield called human-machine communication (HMC). I have no place in that area. Please go to twitter and seek for threads written by those experts.
This article looks like it’s about AI, but it’s not. The entry point here is that with the recent advance in large language models (LLMs) and image generators, people are worrying about their jobs. A very common angle in covering AI is let some AI agents try out our jobs and then “prove” that AI can’t replace our jobs. This video by WIRED is a typical example.
The arc of all these stories is almost the same: asked an AI agent to do x, the human expert said the AI made several mistakes, so the job of the human expert can’t be replaced by AI. However, if someone asks me the same question: Can AI replace my job? I don’t need to think and I don’t need to evaluate the performance of AI. My answer is extremely definite: Yes, AI can 100% replace my job. Absolut!
The common AI journalism makes a mistake about how AI should be approached. We are not talking about what unique things that human can do and what AI can’t. Also, we are not trying to show the unique value that human workers can generate that AI can’t. And all the things that the current AI journalism tries to put on the table miss the whole point. The million-dollar question is: can AI replace our jobs? What we want to know is the replaceability of our jobs by AI.
Think about it this way: Suppose I run a fashion company. DKNY, for example. Can the product of my company be replaced by a cheap brand, let’s say Primark? As I run my fashion company, I can say my fashion has x, y, z values and therefore it deserves to be more expensive than Primark. However, try to ask an ordinary person how often they buy things from my company and Primark. It is extremely unlikely this person would have brought anything from my company, but rather from Primark. In this consumer perspective, Primark does replace DKNY economically. Most of the consumers don’t care about your added x, y, z values. What they care about is the Price-Performance Ratio. Both DKNY and Primark provide T-Shirts. And Primark costs 10 times less, even though the T-shirts were made in some dangerous sweatshops in Bangladesh with no fire safety.
I remember when I was young, counterfeit was rampant. My parents brought me a “DHNY: Donna Haren New York” T-shirt. In a movie by Fruit Chan, there is a scene that parodies Hong Kong as a counterfeit city. A guy wears a “PKNY” T-shirt. PK is a common curse word in Cantonese and that guy doesn’t care. Again, from the consumer’s perspective, “DHNY” and “PKNY” replaced “DKNY”.
We need to remind ourselves honestly how we actually behave when there is a sloppy but cheaper clone of the same thing. To give more examples: illegally download MP3 files from Napster or buy CDs; eat cheap junk food or nutritious whole food; buy greenhouse-gas-emitting energy from warmongering dictators or invest in green energy. Let’s face the truth: there is a market for sloppy product even when that sloppy product is killing you, having a risk to put you behind bars or making you a complicit in crime against humanity, because we don’t have an unlimited amount of purchasing power. Again, can AI replace my job? As long as I am way more expensive than AI: yes, AI can 100% replace my job even AI might do a sloppier job than I do. Or even, AI is doing a sloppy job that can harm the consumers.
Like in the AI journalism, I can regurgitate what x, y, z values I can generate and therefore my service should be more expensive than AI’s. From the DKNY example, we know that the market value of your x, y, z depends on 1) whether my consumers care about those x, y, z and 2) if they care, whether or not they can afford to pay me for my x, y, z added-values. From this analysis, I know for sure for most consumers they don’t care about the x, y, z values a researcher can generate. Most consumers would say they can do their own research on YouTube and listen to those YouTubers. They can also get their PhD from the University of Facebook.
Let’s back down a little bit. LLMs are like people who can speak confidently about things that they know shit about. And they don’t care about the factual accuracy, integrity or how they can harm you. In my language, we have a slang for that: 9upper (Shit-speakers). An English translation would be bullshitter. For example, a bullshitter without any knowledge about labor economics can write confidently about labor economics in a blog post. Let’s face it, even a bullshitter who doesn’t know how to read can have more audience than an expert on Telegram. An LLM has at least consumed the whole Wikipedia, code from GitHub and all of the racist comments from Reddit.
If you really need to prove you are not replaceble by an AI, what you need to do is not to show you can provide x, y, z values than AI can’t provide. Instead, what you should do is to show how cheap you are willing to sell your added x, y, z values. Let’s say you are a voice actor and you sell a service of voice acting for a movie for USD 1000. Your added value is the cadence. You can voice act 100% like a human. The hottest AI can do it for less than 1 USD. Maybe the AI can do it only with 60% of your cadence. If your customers don’t care about that 40% difference (and the gap will only get narrower, not wider), what you need to do is to reduce your charge so that your customers are more willing to pay for your added 40%. Are you willing to do that? I can’t say for you but I can’t sell my added x, y, z values for cheap because I have mouths to feed. And because of my price, again, AI can 100% replace my job for cheap.
Having written all these: DKNY still exists. In this sense, we can say that the existence of Primark and the counterfeit PKNY does not threat the __existence__ of DKNY. There are still consumers who are willing to pay for the x, y, z values that Primark and counterfeit can’t provide. On one hand, the market is segmented into high-end and sloppy fashion and therefore both DKNY and Primark can peacefully co-exist. On another hand, the highly lucrative counterfeit threats the financial interest of DKNY. Selling of counterfeit is illegal.
And interestingly, these are also the two paths we can take. The first path is to segment the market into human service and AI service. Luddites can choose human service as a high-end service. The second path is to treat AI services like counterfeit because they are against our financial interest. We should outlaw those AI service providers so that we can still work gainfully, pay for our student loan and mortgage till 60 years old and then die. Or, we should tax those AI service providers heavily and use those money to provide universal basic income to all of us.
If AI does replace my job, I don’t mind working for AI. For example, working in a Uranium power-plant to generate nuclear energy for its consumption (or even generate bioelectric, akin Matrix) or the extremely low-end job like debug the Linux kernel manually, the software that all AI agents probably run on.
BTW, this whole mumbling blog post could have been written by an AI agent, given the less than 100% correct factual information and English grammar.