I don't think AI will entirely replace tech workers in the near future, but it will drastically reduce the numbers of tech workers needed for a particular task. AI will help a lot in coding/designing/testing, etc., which will reduce the number of workers needed for a particular task. I can already see this at my work where we automated lots of basic stuff using LLMs, but those are just basic stuff right now. The pace of these replacements is much higher than what people actually think. If you graduated in the last 10 years, you must know the exponential advancement happening in this field over a decade. In just a year, there will be 10+ research paper with better results than the year before on benchmark dataset. Just a decade ago, the journey of deep learning started with the AlexNet architecture with 62M parameters with basic image classification taks(I know some people will argue on this). Now, we have GPT4 with 1.76 trillion parameters. In the next few years, technology-wise, AI will have the capacity to replace tech workers, but companies will weed out engineers slowly, replacing most of the systems with AI. This will drastically reduce demand of tech workers. TC: 420k YOE: 5 years (3 job switch and masters)
I am totally hooked to copilot, in 4 out of 5 cases copilot has been effective in writing codes, few examples from this week: 1. Code to replace websocket with SSE 2. Shell script to build dockerfile by fetching aws credentials (8 manual steps were automated by this) Both these are at least 1-2 days worth of work each, using copilot I did both the work in 30min including testing
By copilot do you mean it helped you write code in your companies repo or write new code from scratch. Latter works for me. I have never found it useful for the former.
I am trying to use copilot and chatgpt everyday as a SRE. My life has not changed. My productivity has not increased.
A slightly better Stack Overflow. The major problems with complexity will still exist.
I disagree. AI co-pilot tools have the potential to really increase the productivity of engineers on certain tasks, but the whole history of computing has been a story of ongoing dramatic increases in productivity. The increased productivity makes new types of software and systems possible, which expands the market, and increases the need for experts in the new areas. This happened for virtual machines, for big data, for cloud computing, and so on. Each has revolutionized what is possible in the industry but only resulted in more software to be written. Saying that AI will put software engineers out of work is like saying that Photoshop has put all the graphic designers out of work.
It wonโt push SWEs out of work, but it will definitely change the way we code now
Well of course, that's what I'm saying. When I started in this industry, if you wanted more computing power, you placed a hardware order and waited a month. Now you make an API call and scale up some ec2 nodes in a few seconds. That didn't put anybody out of work, it just made it possible to build much bigger and cooler things. Things which, by the way, needed a ton more engineers to support them ๐
The question is when this is going to happen. Iโm guessing 5 to 10 years.
We might become verification engineers. Ai will write code we need to approve and fix bugs. Ya pay and number of jobs will reduce drastically. ๐
To a great extent, that is the job you do today
I can see that. But it seems this will have similar effects on many other white collar roles as well, not just tech.
LLM != AGI
This^ . We are not even close to a true AI able to replace us. We dont even understand how our brain works and even worst we cannot make something similar because of first reason.