Geek chat techno retro

[White-Noise]

I thought it would be fairer to have a proper thread for this sort of topic picking up from the jokes thread...

https://www.pistonheads.com/gassing/topic.asp?h=0&...

It was interesting looking at the technical and Hands-On aspects to technology and networking that many of you were talking about. I feel like I should have been born 20 years earlier because I used to enjoy being much more Hands-On, but I find these days getting further and further away from the detail and the Hands-On side of technology.

I've gotten into home lab in the last 5 years. Been great to actually understand more of the tech that I've owned and managed in my career. I find it difficult to find the headspace most of the time, I'd rather be running or working on cars than stuck in front of a screen when not working!

Edited by White-Noise on Thursday 7th August 17:43

[Clockwork Cupcake]

I'm definitely up for this and have bookmarked / watched it.

I also started a thread a while back on Retro Computing if anyone is interested.

https://www.pistonheads.com/gassing/topic.asp?h=0&...

[Lucas Ayde]

White-Noise said:

I thought it would be fairer to have a proper thread for this sort of topic picking up from the jokes thread...

https://www.pistonheads.com/gassing/topic.asp?h=0&...

It was interesting looking at the technical and Hands-On aspects to technology and networking that many of you were talking about. I feel like I should have been born 20 years earlier because I used to enjoy being much more Hands-On, but I find these days getting further and further away from the detail and the Hands-On side of technology.

I've gotten into home lab in the last 5 years. Been great to actually understand more of the tech that I've owned and managed in my career. I find it difficult to find the headspace most of the time, I'd rather be running or working on cars than stuck in front of a screen when not working!

Edited by White-Noise on Thursday 7th August 17:43

I have a feeling that the age of AI is, over the space of a few years, going to really lead to most people not having the faintest clue about how any information technology actually works. Not that most people have a clue now.

The nice thing about older tech is that you can actually grasp it and it is the basis of modern tech. I've been getting back into my old Atari 8bit micro and programming it, optimising it etc. Ironically, AI is really handy for programming 6502 assembler...

[Clockwork Cupcake]

Lucas Ayde said:

I have a feeling that the age of AI is, over the space of a few years, going to really lead to most people not having the faintest clue about how any information technology actually works. Not that most people have a clue now.

On the one hand I hope not. On the other hand I am a C++ guru and made passing reference to a young colleague about a constructor and he was like "oh! What's that?". This was a grad who had been sent on many C++ courses by the employer and who had just returned from an "Advanced C++" course. And when I elucidated, instead of saying "oh ha ha silly me! Of course!!!!" he was "Oh wow. Thanks for the info"

[Lucas Ayde]

Clockwork Cupcake said:

On the one hand I hope not. On the other hand I am a C++ guru and made passing reference to a young colleague about a constructor and he was like "oh! What's that?". This was a grad who had been sent on many C++ courses by the employer and who had just returned from an "Advanced C++" course. And when I elucidated, instead of saying "oh ha ha silly me! Of course!!!!" he was "Oh wow. Thanks for the info"

C++ might still have a place - the LLMs (as the name implies) actually depend innately on actual, written, language which means that they work best at coding in higher-level 'english-like' languages. Plus, a lot of their training with source code is with higher level stuff like Python, Rust etc. So whilst they can do C++ using pure reasoning, they are much more effective at, say, using Python to implement what natural languge prompts describe.

Luckily for 80s retro afficionados, stuff like 8-bit assembly language is pretty straightforward for them to actually reason from scratch.

[Clockwork Cupcake]

Being a C++ guru will see me out to retirement. But I do worry about the younger generation. Mind you, I have had young padawyns who have scared me with their C++ knowledge. The most recent being young enough to be my son. So I do have hope.

Note: Not tooting my own horn here. But if I had spent 35+ years specialising in C++ and wasn't a guru in it, then that would be even worse.

[Lucas Ayde]

Clockwork Cupcake said:

Being a C++ guru will see me out to retirement. But I do worry about the younger generation. Mind you, I have had young padawyns who have scared me with their C++ knowledge. The most recent being young enough to be my son. So I do have hope.

Note: Not tooting my own horn here. But if I had spent 35+ years specialising in C++ and wasn't a guru in it, then that would be even worse.

Hmmm, I just asked Gemini AI and it said that C++ is one of the more suitable languages for AI coding

I assumed that something like COBOL would be 'easy' but C++ not ... but not according to Gemini (Python was the top language however).

--

That's an excellent clarification. The question isn't just about what C++ is good for, but rather, why is AI good at using it. The answer lies in the nature of both C++ and how AI models are trained.

1. The Abundance of Training Data

While Python is the most-used language for developing AI applications, C++ is one of the foundational languages for the AI and machine learning libraries themselves. The underlying engines of frameworks like TensorFlow and PyTorch are written in C++ for performance. This means there is a vast amount of high-quality, professionally-written C++ code available for AI models to learn from. This includes not just the core libraries but also thousands of real-world applications in robotics, gaming, and high-frequency trading that are built on these frameworks. AI is good at using C++ because it has been trained on a massive and well-structured C++ codebase.

2. C++'s Predictable Structure

C++ is a statically-typed language, which means that the type of every variable is known at compile time. It also has a more verbose and explicit syntax compared to Python. While this can make it harder for humans to write, it makes it easier for an AI to parse and understand the code's structure and intent. The clear, less ambiguous nature of C++ syntax helps AI models generate code that is more likely to be syntactically correct and logically sound.

3. The Need for Performance-Oriented Code

AI is often tasked with generating code for a specific purpose. When that purpose is high-performance, low-latency, or embedded systems, the AI knows that C++ is the most suitable tool. AI models are trained to recognize patterns and choose the best tool for the job. Because C++ is the de-facto standard for performance-critical applications, the AI's training data would have established a strong association between these requirements and the C++ language. The AI is therefore "good" at using C++ because it understands the context in which C++ is the superior choice.

In short, AI's ability to effectively use C++ stems from two main factors:

A vast and high-quality training dataset of C++ code, including the very libraries that power AI.

The clear, structured, and predictable nature of the C++ language itself, which is well-suited for AI's pattern-recognition capabilities.

[White-Noise]

I studied Java at uni but once I got into the real world I did some vbscript and that was kind of it given how much outsourcing and offshoring to Asia took place.

With that said my degree did help a lot at work. I recall asking what normalised form they went to and the guys in the office didn't have a clue

My career needs a reset at some point it would be nice to have a core skill rather than being a generalised fire fighter.

I worry about the younger generation too. Some of them are very very impressive as you mention above but for many it's going to be so tough.

Still we can all give up work and let the machines do everything for us right... Right!?

I've been enjoying the retro computing thread too CC.

[Clockwork Cupcake]

Lucas Ayde said:

Hmmm, I just asked Gemini AI and...

Well that was your first mistake.

[Clockwork Cupcake]

White-Noise said:

I studied Java at uni but once I got into the real world I did some vbscript and that was kind of it given how much outsourcing and offshoring to Asia took place.

I am a little older than you. In my first degree (Applied Physics) I studied FORTRAN. When I did my MSc in Software Engineering in 1993 I specifically went for C++ which was new then. I built my whole career on applying the Scientific Method to debugging with huge success. I would like to consider myself an Enola Holmes of debugging, It's something I am good at. I remember one time I instantly solved a bug that had stumped some colleagues for a week, but fair play I analysed their findings and when a colleague said it worked in Debug and not Release and it seemed to be calling the wrong methods and I looked at the code and saw #ifdef statements round virtual method declarations I instantly knew... but you got to know your st to know that. It's like £5 to hit it with a hammer and £195 to know where to hit it and how hard.

White-Noise said:

I've been enjoying the retro computing thread too CC

Thank you so much

[White-Noise]

Clockwork Cupcake said:

White-Noise said:

I studied Java at uni but once I got into the real world I did some vbscript and that was kind of it given how much outsourcing and offshoring to Asia took place.

I am a little older than you. In my first degree (Applied Physics) I studied FORTRAN. When I did my MSc in Software Engineering in 1993 I specifically went for C++ which was new then. I built my whole career on applying the Scientific Method to debugging with huge success. I would like to consider myself an Enola Holmes of debugging, It's something I am good at. I remember one time I instantly solved a bug that had stumped some colleagues for a week, but fair play I analysed their findings and when a colleague said it worked in Debug and not Release and it seemed to be calling the wrong methods and I looked at the code and saw #ifdef statements round virtual method declarations I instantly knew... but you got to know your st to know that. It's like £5 to hit it with a hammer and £195 to know where to hit it and how hard.

White-Noise said:

I've been enjoying the retro computing thread too CC

Thank you so much

Yeah youre much further down the geek route than I managed. Owning technical solutions Ive always had a want to understand the tech better and its something Im working on. Its that balance I had to strike. Having the right techie has been an art in itself finding the right person who can cut through all the noise and save so much time and money. The number of times people try and roll out the wrong solution or miss things, when I can point at a screen with some of my basic knowledge and ask whats this doing and when we get to the answer often it solved our problem.

Was it you that was using truenas?

[Clockwork Cupcake]

White-Noise said:

Was it you that was using truenas?

Yes, I run a TrueNAS built out of a used Supermicro rack server, re-housed into a desktop case.