Why not offshoring our brains?

AI, AI, AI, AI, AI

Whichever meetup I have attended lately during the past, “How to use AI efficiently?” and “What’s your AI workflow?” has been a pivotal topic; to some extent that it has became a cult to shove AI into anything relavant or even remotely irrelavent. That made me take a step back, make a sparkly circle like Dr. Strange and freeze time to rethink what’s going on here and what are we trying to accomplish there.

How it [might have] started?

AI has been proven effective in performing many tasks such as coding, drafting docs, writing tests, brainstorming, and research. It’s really fast, has relatively unlimited access to information created by mankind (or other LLMs 😉✨), and comparatively getting intelligent over time. Hence, given a good context (as skills, prompts or whatever) they can walk us over untamed lands and unstructured instructions with an acceptable pace and quality that seems tempting to offload our tasks upon. This being the motive, people has been taking the leap into unmatched productivity and started exploring use cases for this magical being X — Tanya the evil reference, thanks to AGHAYE ROSTAMI — that has initiated this cult.

How I got onboard?

As most of you have guessed (must have been tough since I have complained a lot now and then 😂), I got into the tracks with a bit of push from my workplace. The push and urge for being more productive and deliver more made me explore this beast and try to incorporate it into my workflows. However, I have not been as “progressive” as this might sound, testing the waters little by little and navigated my way there warily, skeptical. Anyways, after a wild ride I got here and started to trust this being; letting it to write my tickets, used it for brainstorming and writing actual production code — doing my job for me!

This being my villain arc, I started to deliver more and more, being tempted to do more and more turning me into this (ignore title):

That felt really impressive in standups, I was contributing to the team and shipping projects that usually took months to production in a few weeks. Even the dopamine you get from receiving personal shoutouts from the CEO for the projects you delivered is unmatchable. Let alone working on multiple projects simultanuously 🚀.

Taking a step back from AI

Being that “productive”, created a new mindset of “building more good, the more velocity the better” and I became self-aware when I noticed this magical being X making some random assumptions in the codes I review (either it wrote for me or others’ PRs). One day, I was trying to implement a feature on gRPC that my AI was making it up. I was going insane until found out who the dumb one is 😐. This was the trigger in my arc, when I noticed I’m just letting the model steer my thoughts and I became the passenger prince.

On another note, the AI Bros putting so much emphasis on the hype that “job X is already solved” made me rethink my sanity because I felt like “job X” (X being software engineering for the sake of this post) is more than some hallucinations of a statistical model on steroids (or silicon, depending on how you/your agent sees it).

All in all, my neurons started firing in a critical direction. For example, the velocity that we deliver started to triumph the core problems that we solve, resulting in many debates between me and friends. So what is the issue here?

The issue(s)

Here, I try to spare you folks reading this the random details of my thought process. The main issue I found was that AI has become an Intelligence for rent that takes over doing the whole job, rather than being a tool or companion for people actually doing their jobs. That is, people started to use those tools as a brain and sign off what it does. That is a huge problem IMO! What we, as humans, bring on the table in this environment (e.g. society) is that we are RESPONSIBLE, therefore can be held ACCOUNTABLE for doing (or not doing) the job. If we’re giving up that implicitly, then are we actually doing any job or only act as a naive operator?

Moreover, when you shut your brains down it’s like you stop going to the gym. When you stop hitting the gym normally (each person at their own pace), you lose muscle mass! Not the exact same thing literally happens to your brain (fortunately) but it might wire back to factory settings after a long while. Imagine hitting the slot machine (LLMs here) mindlessly for a quite time hoping you hit the jackpot (does the job). This definitely numbs the critical thinking functionality and sends it to an indefinite vacation. For example this study published by Microsoft (and CMU folks) or AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking provide some crisp arguments for what I said (I’m not writing an academic paper, just bringing some evidence for the sake of the argument).

The ones I mentioned above are the root causes — in my own and honest opinion —, and let’s see some of their adverse effects on our lives based on my limited (yet valid) observations:

Losing [language] skills

I know of a bunch of people who write even the simplest emails (e.g. communicating minor issues with their landlords) using AI, and gradually lose their ability to write or communicate in a foreign language. This applies to any skill too; such as critical thinking that already was mentioned. Those skills are not promised nor burnt into your brains, but kept familiar in your memory by being actually used.

Rendering some “communications” meaning-and-soul-less

When you don’t read (and care) LLMs — or your agent’s — output, why should someone else do so? Each person has a soul, and it is reflected on their comms (medium being talking, writing, …). As soon as some soul-less entity (being X) takes on sorting words one after another it kills the vibrant soul behind it, making up some robotic neutral thing that exists out of spite. Thus, back to the main point, who would care about that? Other NPCs or Agents?

Getting vendor-locked!

Imagine one day, the evil guy on a bridge with half-framed glasses — random soul-less anime reference again — (any similarities to any CEO like Dario is just a coincidence and I deny any intentions behind that) decides to lock you out of their LLMs. Given they have become a crucial part of your workflows, are you able to move on with your work?

False sense of knowledge/confidence

I have seen so many examples of this case in day-to-day life, people asking their “AI agent” a random question, and without verifying the results or even thinking if it makes sense, confidently post it everywhere making a 🤡 of themselves. Please don’t be that person! In the aforementioned case the implication is just being looked like a stupid person, though it might come with more dire consequences such as others’ lives (including children) on the line in military use cases.

There exists definitely a longer list as per your own experiences if you look around. Consider this as an invitation to look with a new lens next time 😁.

So, is this Silicon-devouring monster doing any good?

So far, I was just making a villain out of this soul-lacking being X — yet those were valid arguments 🔪. However, it would be ignorant if I just end the post there. Of course a tool that has become so integrated within our lives must be beneficial in some ways. The key here is that it is a tool and shall be treated as one. It is not a real, but an Artificial Intelligence that is not neccessarily accurate in any manner. It can handle simple routine tasks that it has seen tons of it within its training, but it still needs due verification upon task completion as the person operating it is signing it off and taking responsibility.

It definitely increases the speed things flow and decreases the cost associated in knowledge work — like software engineering. Let’s take this blog as an example: I was procrastinating starting it again for obvious reasons; it would have taken a lot of time to revive it from 2016 tech, and was not worth putting so much effort on it. However, I managed to do a quick research, verify Claude’s ideas and refine them, create the repo and customize the theme within a night or two, and here I am.

Other examples include (but not limited to):

• My work projects (planning and coming up with initiatives)
• My university projects
• Learning and development
• Cooking and menu planning

All I’m saying, is that it falls back on the user and their responsibility to use it in a constructive/beneficial way. Who am I to say this is the only correct way, I just can say some ways might be or not be working in long term. For example, as per the CMU folks suggested in their papers, using AI as in Tutor mode rather than one-shotting tasks in Answer mode seems to be skill-building while boosting productivity.

My last word closing this article is just please (pretty please, with :cherry: on top) stop offshoring your brains and renting the intelligence from LLM providers, you are not a NPC, but a real character with real intelligence and AI should be only a tool in your toolkit, nothing more!

P.S.: You might think em-dashes are put there by my mysterious AI tool, though I should disclaim that they’re purely hand-crafted by a real human (being myself) for you. They would love being appreciated.

Disclaimer: I may or may not (in this post not) use LLMs for proofreading so I gain more confidence (again) in writing. I might stop after a few posts.