- cross-posted to:
- pcgaming@lemmy.ca
- cross-posted to:
- pcgaming@lemmy.ca
So we are pretending. This is the truth, Rob. We as the industry are pretending we’re not using AI, but the reality is, it’s almost impossible to code now without using AI. Almost impossible. I mean, the efforts you have to go to extract AI from line completion stuff and from generative stuff is very challenging. So everyone, every title that’s come out in the last year has been using that, but we’re still able to not tick the AI box. AI is going to be, when it’s sorted out and not producing AI slop, it is the most… This is very controversial. It is the most, the biggest advance in creativity for smaller studios, for medium studios and bigger studios. The biggest step forward in creativity."
Uh huh. Sure there guy who keeps overhyping things and never releasing what was promised.
There’s a huge difference between “It makes things a lot easier/better” and “Coding is impossible without it”… One of them you can argue is true. The other is an outright lie, and even he knows it.
Yeah of course it’s not impossible and I think we should keep developing personal coding skills because we need those to have proper judgement about what the clanker spat out. It’s just my 2c on the topic.
The main issue is. If companies enforce AI use (and many do), and/or they stop recruiting junior developers (a lot have), then we’re going down a road where soon (for values of soon) we just won’t have people with the skills to evaluate what they produce.
Worse than that, they can only improve if we provide the input for them to use to improve. Any future models are going to be consuming a mixture of human work, human curated AI work and AI slop, ever skewing towards the latter. I am not convinced the way these models work, they are able to tell the difference.
We might well have the best models we’ll ever have, right now.
Well yeah companies need a healthy mix of experienced and young employees and what they’re trying to save in expenses now will hit them straight back in their face a few years down the line. But CEOs don’t get paid to think about the future, the shareholders want their revenue right now. It’s stupid and it won’t change soon enough.
For training, I think you’re mistaken about how this works nowadays. Training doesn’t happen just from ingesting human written text anymore. For stuff like programming the models are trained by reinforcement learning with difficult tasks. This allows for much more targeted tuning without the need for a huge corpus of pre-written data.
Well if they are using that method it’s only going to work as a refinement technique on existing or new traditional models. At the end of the day they can only tackle difficult tasks using knowledge they have vacuumed up by their (often from dubious sources) large datasets. This reinforcement training still won’t produce innovation. It will most likely just specialize the models, possibly to work better for specific tasks while requiring less memory or other resources. A good thing for sure, but it still doesn’t mirror innovation.
Now, I’m going to be honest. On one hand, I realise if we got AGI the best case scenario is most of us would have ended up with no jobs, the rich would get richer until we’d have a violent, bloody revolution and have AI design us a more efficient guillotine (for any future legal trouble, it’s important to note this was a hypothetical based on us achieving AGI, we didn’t so you cannot legally constitute this as a threat, call to action or plan to commit a crime).
Nonetheless I did kinda hope that with the huge investment the tech bros got for AI, that they would spend it going down the avenue of trying to achieve AGI, rather than just buying all the GPUs and making more expensive models where we’d already know there was a diminishing return before they did it! Yeah it might go full skynet on us, well apparently current models are perfectly capable of going skynet on us if we let them near our nukes. But, still this would have a chance to just bypass us and do it anyway. But if it didn’t, such issues of innovation might have been solved.
As it is, they squandered the money and here we are now. Which I suppose is still a good thing, in a weird way.
Anyway I’ve gone off path, but I’ll leave it all there. My main point is, what we have is ultimately a dead-end failing some great discovery. It should be used as a tool to assist skilled writers, artists, programmers and anyone else. Not to replace them, but to augment them. Oh and it wouldn’t go amiss to not take all the returns to the top, maybe the little guy could get a tiny tiny piece of the pie for a change.