#How to adapt to AI replacing software development jobs?

26 messages · Page 1 of 1 (latest)

earnest otter
#

With the recent release of gpt 4, it seems like it has gotten better at programming compared to gpt 3 in Nov 2022. With the current rate of progress, in the next 5 years, LLMs could replace a lot of coding tasks.

So what can software engineers do to adapt? I entered the industry in 2020 and was looking forward to a 20 year career as a software engineer, as I like problem solving and building stuff and making lots of money, but it seems like this is gonna be a distant dream, as AI is getting better at coding 🥲

So what skills should us software engineers focus on developing to prevent us from sinking in this jobs maelstrom that is AI automation? Will software engineering go the way of the horse buggy?

karmic totem
#

There have been a lot of similar AI promises in the past. They don't really pan out. Or at least when they do, they only create more software engineering jobs.
There's no guarantee that's how this will go, but generally it is young programmers who are worried, not senior ones.

#

I would be extremely cautious of using two data points to extrapolate a predictive curve. That's rarely how it actually goes (for stuff like this, it tends to be more logarithmic) and that's how companies sell unwitting investors on hype

#

You might also find this helpful: https://lethain.com/forty-year-career/

A forty-year career.

The Silicon Valley narrative centers on entrepreneurial protagonists who are poised one predestined step away from changing the world. A decade ago they were heroes, and more recently they’ve become villains, but either way they are absolutely the protagonists. Working within the industry, I’ve worked with quite a few non-protagonists who experi...

earnest otter
#

I gave it some thought, and I think the nature of software engineering will change with the AI revolution.
Here is a few of the current skills mentioned for a software engineer

1. Proficiency in programming languages

2. Problem-solving skills

3. Understanding of data structures and algorithms

4. Technical writing skills

5. Communication skills

6. Collaboration skills

7. Continuous learning
#

Do you think proficiency in programming languages would still be important in the future? Given GPT's capacity to program...

karmic totem
#

Yes, because it is bad at programming correctly. But also, the actual programming part is a small section of what makes software engineers valuable. Understanding code and being able to fix it is still going to be super important, and you don't really get that without doing programming in the first place.
And then all the other stuff that isn't even programming.

jade night
#

This is a question I'm intensely interested in. Too many variables right now to make predictions, which is what makes it so anxiety inducing!

#

Like, the core of my value in a moneymaking sense has always been "somewhat intelligent, knows lots of things, has good facility with language and deductive reasoning, and can be lightly adapted to any task requiring those skills."

Seems like these tools are pointed squarely at automating that.

The specific skills listed for SWEs above are good advice, but when you start automating the "synthesizing information and composing language" parts of the job I worry that will make the landscape much more competitive.

earnest otter
# jade night Like, the core of my value in a moneymaking sense has always been "somewhat inte...

It's no question that AI makes us more productive, so I think this could go two ways. The first possibility is that with the amount and scale of projects done by x workers go up. The second possibility is that the amount and scale of projects stay the same, and the amount of x workers needed go down.

The second scenario is the worrying one, so I chatgpted this, asking

In the past, like industrial revolution, what did people do when machines took over? For example, say it took 10 diggers to dig a ditch. But then with the invention of a excavator, it took 1 digger to work that excavator. What did the other 9 do then?
#

This was the answer it gave

During the Industrial Revolution and subsequent periods of technological advancement, many workers faced displacement as machines took over tasks previously done by humans.

Here are some ways that workers adapted to these changes:

1. Reskilling and Retraining

2. Shift to New Industries

3. Increased Specialization

4. Entrepreneurship

5. Education

It's important to note that the transition was not always smooth, and many workers faced hardships, unemployment, or underemployment during these periods of change. However, over time, technological advancements have generally led to higher productivity, economic growth, and improved living standards.
#

I asked what value can humans bring to the software development process that AI can't, and this was the answer it gave

1. Emotional Intelligence

2. Creativity and Innovation

3. Domain Expertise

4. Collaboration and Teamwork

5. Critical Thinking and Problem Solving

6. Adaptability

7. Oversight and Quality Control

In summary, humans bring a unique combination of creativity, critical thinking, emotional intelligence, domain expertise, and adaptability to the software building process.
#

So from this I think/it seems like, the best way forward would be to focus on improving the skills mentioned, like interpersonal skills (collaboration, communication, emotional intelligence), business skills (creativity, entrepreneurship, vision), high level skills (design, architecture, process) , specialization skills, and problem solving.

In your case Danica I think 'know lots of things' and 'good facility with language' can be automated, but intelligence (the action of thinking), deductive reasoning, adaptability won't be able to be automated as easily. If that happens that'll probably be AGI.

#

Of course if all these don't work out, I'll just be a beekeeper that makes honey, as that is unlikely to be automated by AI. And it seems like a leisurely job.

polar axle
#

I don't think software engineering will go away all together because AI is here. AI will automate lots of tasks, so engineers who are just following a pre-defined script in their work will get replaced. But there will be a need of engineers who are doing creative work.

#

We can;t predict for sure how future will look like, but one way to navigate AI is be part of it. Learn how AI works, learn about OpenAI and similar technologies. See what fine tuning a model means. See how vector embedding works. Keep up to date with latest advancements in AI applications. A boom of tech hiring is coming, where engineers with AI skills will be needed. Take this paradigm shift as an opportunity.

earnest otter
#

Just writing my thoughts here to see what people think.

I gave this topic abit of thought, and I was thinking of how things evolved throughout history. In the past, we have gone from an agricultural economy, to a industrial economy, to a knowledge economy currently.

The agricultural economy was characterized by pure human production like farmers, the industrial economy was characterized by humans working with machines like factory workers, and the knowledge economy is where humans work with knowledge.

If you see the trend, you can see we have slowly gone from lower order processes to higher order processes. From pure hand labour (agricultural), to hand with tool/machine (industrial), to knowledge labour.

So in the future, I think we will be moving to an economy with even higher order processes.

#

Here is a list of general low level, medium level, high level processes

Lower-order processes:
-Sensation
-Perception
-Attention
-Memory encoding and retrieval
-Basic motor skills
-Habituation
-Classical and operant conditioning

Medium-order processes:
-Pattern recognition
-Language comprehension
-Basic problem-solving
-Emotional processing (identifying emotion)
-Implicit learning
-Basic decision-making

Higher-order processes:
-Abstract reasoning
-Critical thinking
-Creativity
-Complex problem-solving
-Planning and organization
-Metacognition/Reflection
-Social cognition/Emotional Intelligence
-Moral reasoning
#

As can be seen from a glance, the majority of tasks that AI like Dalle and ChatGPT can do are medium order processes. This also concides with the fact that in our knowledge economy, the majority of current blue collar work (lawyer, graphic designer, marketing etc) consists of tasks that use medium order processes.

So what happens when AI takes up the majority of medium order processes tasks? From what I can see, there'll be two direction of workers flow.

There'll be a flow downwards, to tasks which use a mixture of medium level processes and low level processes, things like mechanic, plumber, hairdresser, electrician, construction worker. This is due to currently robots aren't as sophisticated at their fine motor skills and adjusting to different scenarios. However once robots get better this will get automated.

There'll be also another flow, a flow upwards to tasks which use higher order processes. I don't excactly know the job titles of the future yet, but the tasks would probably be higher order processes tasks. Things like abstract reasoning, critical thinking, metacognition, emotional intelligence etc. It would be tough for AI to automate these higher order processes tasks, as it'll be AGI if it can do these tasks.

So like my point is I think we should hone our higher order processes to prepare for the future economy.

#

Specific for a software engineer, these are the processes used.

Lower-order processes:
-Sensation and perception
-Attention

Medium-order processes:
-Language comprehension
-Basic problem-solving
-Simple reasoning
-Working memory

Higher-order processes:
-Abstract reasoning
-Critical thinking
-Creativity
-Complex problem-solving
-Planning and organization
-Deductive and inductive reasoning
-Analogical reasoning
-Cognitive control
-Reflective judgment
#

Just like for general blue collar workers, we can see for software engineers what LLM is automating is the medium order processes. So I think this would then shift the nature of our work to the type of work that involves more higher order processes. So I think that's where we should look to improve our skills on.

This is also a natural progression in sync with history as throughout time, our economies have consisted of tasks which required higher and higher order processes.

So I think that, just like the industrial revolution automated lower level processes resulting in the shift to medium level processes tasks, now the AI revolution will automate medium level processes resulting in the shift to higher level processes tasks. (Of course this shift usually happens over lots of years).

Interested to hear people's thoughts on this view.

strange wing
#

Enjoying reading the answers to this question