I was linked to the blog post Don't be a meat proxy, written by Niklas Gruhn, by a coworker on Slack. It finally gave a name to something that has bothered me for the longest time: LLMs' impact on human-to-human interaction.
While it can be contested whether Gruhn actually came up with the term "meat proxy", he can definitely be credited with popularising it.
Have you ever been on the receiving end of a wall-of-text that is clearly written by an LLM? Or maybe you have had to look over some work where the "creator" of said work clearly hadn't paid any attention themselves?
That is a meat proxy!
A meat proxy is simply a person who adds no value to the output of the LLM. They just relay whatever the machine spits out. No critique, no opinions and no thought. Or... a proxy... that happens to be made out of meat.
My first experience with a meat proxy happened a few years ago. It was the first ChatGPT boom, and people were freaking out. At work, we had a pretty complicated discussion about how to do something technical. Not angry or anything, just a discussion. We didn't agree during the meeting, so we decided to pick it up the next day.
During that same evening, my half-technical manager sent us a long-winded email with the output from ChatGPT. His prompt was clearly biased in his favour, so ye ol' GPT totally agreed with his stance. After the slop, the manager declared that ChatGPT agreed with him, so the "yes" side had one more vote than the "no" side. There was therefore nothing more to discuss.
There were a few more of these instances, and it was clear that ChatGPT had been allowed the same voting rights as a person.
Consider how this makes someone feel. Someone who cares about what they're doing and has opinions they truly believe in, just to be shut down because an LLM said otherwise. What does that do for morale? What does that do to collaboration?
The title above refers back to this blog post by Tom Bedor, and the title is spot on. I love that phrasing so much.
There's a reason LLM output can feel so unpleasant to receive. Well, there are many reasons, like accuracy, verbosity and so forth. One reason is the knowledge that the person who sent the output didn't bother to write it themselves. Heck, sometimes it is unclear whether they even read the output!
Or to put it like Tom Bedor: "If reading this wasn't worth your time, why is it worth mine?"
To take it a step further: I believe this is bigger than just an interaction between two people. We see the same in the developer space when it comes to code. If verifying your code wasn't worth your time, why is it worth mine? If you cannot be arsed to do the most basic of checks, why should I be burdened with it?
This becomes a bigger problem if meat proxy behaviour is allowed to fester and become a part of the team culture. If a team consists of meat proxies, you no longer have a team. You have a group of isolated developers, working on stuff in isolation, where no real checks happen between them. Code reviews might occur on paper, but no real review will be conducted. A team culture where meat proxying/slopwork receives no active pushback will see their solution slide towards the tech-debt death spiral.
Even if you are not asking for human attention, you might be trusted to answer questions or help out. How would you feel if you asked a question and got a bunch of LLM slopped at you (or, grenaded at you)?
It feels bad.
Not only does it feel bad, but it also makes you trust that person less. First of all, they made you feel stupid by giving you the modern equivalent of a lmgtfy link, but now we're also left wondering whether the answer is even correct. Being pointed to Google will at least force me to do some basic source-checking before doing something, but with an LLM, who knows? Did the person who copied the text look over it?
If someone asks a question, answering them like a human builds trust. It shows that you either put in past effort to know the answer, or that you found the answer and took the time to give a proper reply (regardless of your AI usage). In other words: Trust is built by you showing effort.
The same is true for your work. If your work shows effort, it builds trust in you as a person and your output.
Taking inspiration from the term "meat proxy" and websites such as "no hello" and "Don't ask to ask, just ask", I decided to create my own site: No Meat Proxy.
Is the idea original? No.
Is it clever? No.
Is it technically impressive? Not even close.
But it is my little protest thrown into the void against anti-social behaviour that is enabled by AI.
The site itself is tongue-in-cheek; I don't actually recommend that you throw it at a meat proxy (unless your goal is to make enemies). Instead, use it to put a spotlight on the problem. Don't target individuals, but share it internally as a general comment on behaviour. Hopefully, the meat proxy near you will take the hint without the shame (If not, there's always the meat proxy emoji you can download on the page).
Use it at your own discretion.
I'm sure we've all heard the stories that AI is coming for our jobs. I don't think that is the case... unless you're a meat proxy.
Here's the thing: Meat proxies do not add any value. One can argue they subtract value, as now I'm reliant on someone else's context, which I cannot control. Or their work is dangerous and will cause issues that will hurt the company, culture or worse... cause more work for me. Work would be better if meat proxies weren't there.
If you made it so far and you feel targeted by this post: Good. I hope it resonates. But it is also not too late.
Maybe you feel addicted to LLMs. Maybe you feel that you have no value to add. Maybe you have imposter syndrome or don't feel that your skills are good enough without an LLM. All of which are totally understandable, and I sympathise with all of them.
People value effort. It is easy to just assume that results are the only thing that matters, but most people value that you try. And equally important is that you learn and grow.
Even in the unlikely case that you are truly terrible (you're probably not), there's hope. If you show consistent effort and growth, then people will start trusting the fact that you will one day become great at what you do.
That trust will be rewarded through investment. If people see that you grow, they also see someone worth investing their time into, to mentor and help. They see it as a good use of their time because they trust that you will become great.
Being a meat proxy slows growth because you spend less time learning and looking into something. It makes people trust you less, and it demonstrates laziness. It does the opposite of what you want.
My advice is: Don't be a meat proxy. Learn to learn. Learn from feedback. Show consistent effort. Demonstrate growth.
