Today’s topic was going to be AI but at 1800 words of quite negative views, I changed my mind.

Instead, I got my triangle finally on my Vulkan project.

Whoop.

Look at it.

Isn’t it a beautiful pixelated and jagged triangle.

Got an error to do with a semaphore being signalled when it’s potentially in use at the first frame but I’ll fix that another time. Just excited to finally have the triangle.

Roughly 3 or 4 weeks of on and off work... I’ve got too many hobbies.

It's currently about 1000 lines just for this triangle. OpenGL is so much easier to get started in.

But this has been brilliant. So many areas of customisation and control. I have full control over the graphics pipeline.

The first hurdle is done. I can now move on to importing meshes, sorting out a camera class and maths functions etc. This project is going to be longer than I thought. I’ve also got lots to consider because of how much there is to configure with the Vulkan API. I’m sure that I can do all that after I get meshes and textures in though because the shape of that data doesn’t really change. It’s the pipeline that’s going to change.

But I had another bit of inspiration the other day whilst at work to try making a machine learning project. I don’t know if anyone is familiar with Code Bullet, but he’s done some great fun projects to do with learning algorithms and it’s more relevant for my professional work at this current time. So, I’ve decided to switch back and forth between this renderer and the machine learning project. (I think the machine learning project will be easier). I’ve seen some machine learning stuff in Unreal Engine so I’ll use that and just make something basic to begin with. No plans yet.

Now a little on AI.

And brace yourself because I'm going on a rant.

Just going back to what I mentioned in that beginning about the AI post I’m working on. I feel like getting this negativity out of the way will mean the post on AI will be more positive.

I talked to peers recently about these things and realised that the negativity around AI is not to do with the technology. It’s the way that it’s sold, and the way it’s been made created from the hard work of others.

The technology itself is not bad.

Large Language models are trained by tokenising parts of words or whole words and weighting the context against other words. It becomes a generative form of predicting what word to say next in a sentence and so with enough context, appears smart.

The problem around the way it’s being sold:

It’s going to replace all jobs.

it’s so dangerous it can hack its way out of secure facilities.

It’s assassinating rival agents and not leaving a trace.

It’s AI. You’re thinking of AGI.

I love that AGI became a term out of nowhere to ''correct'' us. Everybody had the same understanding of what AI is and that it means artificial intelligence. But when it came to light that the current ''AI'' doesn't think, the term ''Artificial General Intelligence'' came into play. Most pathetic thing I’ve ever seen. But yeah, the sales tactics are the worst. The fear mongering with how dangerous it is and how it can replace everybody. It’s just disgusting because the real-world evidence contradicts it all.

The agents are dangerous because the agent that is 100 times faster than your senior developer can randomly delete your critical database and all backups and it’s your fault because you trusted it.

It’s so good at replacing jobs that anyone with years of experience barely utilises it at all... Why? Because they find that AI cannot give better than basic understanding. It has a weak familiarity with the basic concepts that it can put together with confidence which is enough to win over someone who has no familiarity of the basic concepts. You can try and prompt it with “You are an expert in blah blah blah” but all that does it make it pretend harder with more confidence and more jargon. It doesn’t make it smarter.

The problem with the way it’s built.

This slightly ties in with the sales in that it’s called AI but really, it’s maths models. Large Language models. They predict the next word in a sentence with weights. It’s hardly thinking at all.

But the way it’s built and the associated data clawing by these big companies that no one is doing anything about is nasty. It’s not built using a carefully selected bunch of material and text to learn from like any self-help material or a collection of learning curriculums that realistically, should be freely available. It’s built and trained on anything that’s been dumped on the internet, and anything that can be stolen, and not only that, but it’s also trained on our data that has been collected via any social media platform, and private conversations and all these things that we’ve been sharing with a company, when we use chat services. Every single conversation no matter how private, if it happened on a service, it’s been “Anonymised” and collected. And somehow, it's still worse than that.

If you buy a phone or a computer, you don’t have ownership of the systems running on it and that software is always asking for permission to enter and ransack your digital home. If you say no, well it'll just ask you again, and again, and again, and again. Until you click yes by accident and then good luck finding out what you agreed to.

This is the state of big tech really. It's an unregulated wild west of theft.

And this is the stuff that makes me mad. The hostile and criminal behaviour that we would never do to each other; we allow to happen to us because we’ve been given no other choice. You want a phone? You must accept our legal terms of service even though you own it. If you pay for a device to track your health, you are likely making a donation of your data as a little investment for a future product that you receive no benefit from financially, yet you paid to own the device gathering that data. It’s like paying an artist to paint a picture of you, but then that artists sells copies of that picture to other people quoting "You never said I couldn't".

Imagine you pay to get a kitchen refitted only to find it means you must share every recipe you make in it. Imagine you buy a home to find the doors locked unless you agree to let someone else come in first and set up cameras to watch what you do and then they say "If you don't want to share this, we'll just blur your faces, no biggie, but if you cover them then we can't let you in the house as it's against our terms and conditions". That’s the digital world of goods and services and it’s not right. Even more ridiculous is that these companies get fined but it’s just “cost of doing business”.

AI is like the perfect combination of everything wrong with the digital world as it stands and it’s been a steady downhill slope for decades.

It's trained on large amount of data that has likely been collected maliciously. And using these digital services is giving these companies more data even if you give no consent as the data gets "anonymised”. It's your IP you're handing them.

State of the internet now is sad. Bots trawl the internet spewing false claims or arguing in favour of bad actors powered by AI. Bots crawling, scooping up as much of our delicious data as they can to train the next generation of overly expensive models.

It’s just so gross.

Anyway. The technology is not the issue, the business models are. And with that rant out of the way. I think I can talk about the technology around AI in the next post without so much hostility.

I hope you enjoyed reading even if I it is a bit of a rant. If I got to you, please take a few minutes to just reset yourself. Mindfulness is something I'd do, thinking of 3 things that bring me joy and living in the memories of them for a moment to just recenter and take a breath. Just be aware of what you are doing online and what you are sharing. Sorry this turned into such a rant. I may even scrap this post.