It is true there were massive strides in the last 8 years. But fundamentally, the tech is still the same large language model it was before, just bigger and better optimized.
It's like going from an ancient, slow, Ford Model T that topped out at 45mph to a Bugatti that can do 260mph in 8 short years. It's impressive, it boosts productivity, it is a marvel of modern technology, but that's not my point of contention.
The issue is that AI companies have been funded on the promise that with enough advancements and upgrades this tech will achieve AGI. Which is an absurd statement to anyone in the field actually developing these things. That is the equivalent of promising that this 260mph Bugatti, with a few years of upgrades and advancements, will become a Harrier Jump Jet!
It's just not happening, a fundamental shift in model architecture or technology used is needed. And from what we've seen so far, no one has discovered any.
Hey there buddy, pretty nice post. But you forgot to pay the recipe tax. You can't post something that looks delicious without explaining how you made it.
I don't. I build new every time. previous PC was 1080Ti Ryzen 1700 16GB, current one is 9070XT 9850X3D 64GB. There's something about being able to build a new PC while knowing you have a fallback ready.
And also - most critically. When introducting a predator to hunt a specific populous prey, ensure that one of them isn't diurnal while the other is nocturnal.
I was going to suggest solvent welding, unfortunately PLA's solvent is extremely toxic. Were this ABS or ASA you could mix acetone and some raw fillament to get this gloop-like mixture. Fill the gap, let it dry and you have a solid bond as strong as if it was printed in one piece. Can also be used as a glue and is just as strong.
I tokenmaxx as much as I can. Every report, design doc, or piece of code I prompt I keep bragging how it's written purely with AI. Nevermind that for each single-line fix I have to write a whole paragraph explaining how and where to apply the fix. And how for reports and docs I feel like a teacher grading a student that winged the test.
At this point I just find it amusing and try to see how far I can push it.
Disagree.
It is true there were massive strides in the last 8 years. But fundamentally, the tech is still the same large language model it was before, just bigger and better optimized.
It's like going from an ancient, slow, Ford Model T that topped out at 45mph to a Bugatti that can do 260mph in 8 short years. It's impressive, it boosts productivity, it is a marvel of modern technology, but that's not my point of contention.
The issue is that AI companies have been funded on the promise that with enough advancements and upgrades this tech will achieve AGI. Which is an absurd statement to anyone in the field actually developing these things. That is the equivalent of promising that this 260mph Bugatti, with a few years of upgrades and advancements, will become a Harrier Jump Jet!
It's just not happening, a fundamental shift in model architecture or technology used is needed. And from what we've seen so far, no one has discovered any.