Both the recession and the development of AI (probably the first a way more) has had an impact on the number of “entreprise” kind of Software Engineer jobs in Switzerland (Java, etc. or alternatives).
Jobs that are still showing up are AI/VR/Security/Database/Architect kind of roles.
Ofc it’s just my observatio and thus a bit subjective.
What skills an “entreprise” kind of Software Engineer should be picking up next to stay relevant?
AI is a hype, for example, today but I am not sure if soon it’s not to become just a niche thing, ie. it’s hard to tell whether it’s going to take all jobs away or remain at an “driving assist” level for various kind of jobs including development.
It’s going to take many jobs away and also change the way we work.
AI skills are the new scripting skills. Yesterday you would prefer “enterprise” developer who also has python/js scripting skills as they are much more productive scripting things out rather than typing everything. Today you would choose the developer who can use AI aid effectively. So many people don’t know yet how to prompt AI to get the right answer.
1 Like
You could say the same thing about doing a Google search
1 Like
Anyone else have this curse where instead of typing something like "ls -l /dirname" you hit up arrow 15 times to find the command in the history? Well, I’ve been experimenting with using AI to help code some functions and am experiencing the same thing.
Instead of writing out a function, I write what I want in English to ChatGPT and get it to do it for me. When it creates examplefunction(xyz). Instead of copying and pasting it and renaming it to customfunc(xyz), I type out ‘please rename it to customfunc(xyz)’ and wait for it to generate it out again and then copy and paste it! 
2 Likes
Hi! I’ve started to use ChatGPT as soon as it has come out before a use of AI became a norm where I work. I agree it is a convinient alternative to Google, ie. Google is still valid for precise factual source based search, while ChatGPT (or other AI based solutions like Github Copilot) removes the white paper problem and overall it’s good for lazy discovery.
Speaking of the job market, I feel there’s a drop in the number offerings, ie. it feels like there’re surviving jobs in fragmented niches like ML, VR, Security, Networking etc. yet the “mainstream” seems to have reduced in size, and I wonder what hard skills could be picked up next to stay in it besides working with AI products, or maybe the very idea of staying in the “mainstream” is wrong and something more life/career changing needs to be done at this point.
I’m not really sure what you have in mind. Demand for generalist developers, if that’s what you call “mainstream”, are always subject to IT boom cycles, hired/fired as the tides come and go. IMO we’re still in cool down period, there are lay-offs around. Demand for AI skills we can see nowadays is a bit different thing, it’s a rush for securing investments. Companies are jumping into the AI bandwagon because that helps to boost the stock (temporarily IMO), and maybe it’s a to-be or not to-be for the future of many businesses. It’s like having a website in 90ties, having a mobile app in 2000++
I’m sure once the economy recovers, IT will flourish again
Thank you, I had exactly this in mind
There‘s always been fads. A bit like now with digitalisation. Haven’t we always written computer systems to replace manual processing. Such is the way it goes.
Certainly, we always are writing systems to replace some manual processing, even if it’s just replacing one digital system with another requiring fewer clicks to get the job done.
Yep, I agree it’s totally normal. I’m just thinking if it is the moment when we hop from a mainstream language like Java to something new, and what’s the new? As if people where moving on from Delphi before (and maybe it’s not a language at all but a paradigm shift…)
In that sense, I think Java is already dead, used only in legacy projects. Everything cloud related (and that’s 90% of all the enterprise software these days, no?) tends to be written in Go.
Things like ChatGPT and Bard are useful for generating code but in my experience they don’t provide a full or always correct solution to a problem so developers will continue to be needed. Give them the stack or log of an issue and they read it very well. So I see them as useful aids at the moment.
And by the way Tesla’s self driving technology has been approved by the EU so is coming here soon. But it’s imminent release has been stated many times. My opinion is it will never be fool proof so deaths will be inevitable. They just want less than normal driving. Use of statistics will be interesting.
As an aside to the general topic - I was wondering who would be responsible, in law, if:
- A self-drive car hit a pedestrian?
- Would it make a difference if the car was empty on its way to pick up customers, or with customers?
- Who would be responsible if the car performed an illegal road maneuver?
Would it be the company managing the vehicles, or perhaps a software engineer working freelance to write some of the self-driving code?
In the work relating to this forum, I’ve been using ChatGPT extensively. But I ask it only for small functions or short snippets e.g. write a query to show the biggest topic starters per category
To which it gave:
WITH UserTopicCounts AS (
SELECT
u.name AS user_name,
c.name AS category_name,
COUNT(t.id) AS num_topics_started,
ROW_NUMBER() OVER(PARTITION BY c.id ORDER BY COUNT(t.id) DESC) AS rank
FROM topics t
JOIN categories c ON t.category_id = c.id
JOIN users u ON t.user_id = u.userid
GROUP BY u.name, c.name
)
SELECT user_name, category_name, num_topics_started
FROM UserTopicCounts
WHERE rank = 1;
Now for sure, I know a bit of SQL and maybe with 20 minutes of Googling and trial and error, I could have come up with this query
But I wrote a single question in plain English and the whole process took less than 10 seconds. I doubt even an experienced database person could have done it as quickly.
I’ve done the same with other languages and it’s come up trumps but not always so you need to be technically savvy to use it. And to add you say simple question but you knew the simple question to ask and that’s not always so simple
Being cynical I guess all parties would blame each other. But seriously there will be initial cases and then their decisions will determine the future decisions.
That’s the main benefit of Co-pilot, it can explain code snippets quite well, or generate comments. Code generation is far from being really useful. When using chat-gpt I usually have to ask about 5 questions to get some code close to what I was looking for.
Exactly! Sometimes it gets it right first time, but for more complicated stuff you have to fine tune your request multiple times so that it would have been quicker to just write it yourself. That said, somehow re-asking the question feels less mentally taxing than having to think deeply to write code.
AI also never reacts aggressively which is kind of interesting in today’s social media etiquette