My observation is that the models are better at evaluating than they are generating, this is the technique used in the o1 models. They will use unaligned hidden tokens as "thinking" steps that will include evaluation of previous attempts.
I thought that was a good approach to vetting bad ideas.
This is not totally my experience, I've debated a successful engineer who by all accounts has good reasoning skills, but he will absolutely double down on unreasonable ideas he's made on the fly he if can find what he considers a coherent argument behind them. Sometimes if I absolutely can prove him wrong he'll change his mind.
But I think this is ego getting in the way, and our reluctance to change our minds.
We like to point to artificial intelligence and explain how it works differently and then say therefore it's not "true reasoning". I'm not sure that's a good conclusion. We should look at the output and decide. As flawed as it is, I think it's rather impressive
I mentioned in another comment, I enjoyed the structure and found it much easier to process. This is despite mostly reading long form articles and never having been a huge twitter user.
On the contrary, I have never been a heavy twitter user, yet I would dearly love all articles I read to be broken down like this. I definitely find it easier to process a list structure like this.
I am currently tinkering with this all, you can download a 3b parameter model and run it on your phone. Of course it isn't that great, but I had a 3b param model[1] on my potato computer (a mid ryzen cpu with onboard graphics) that does surprisingly well on benchmarks and my experience has been pretty good with it.
Of course, more interesting things happen when you get to 32b and the 70b param models, which will require high end chips like 3090s.
At the risk of sounding a bit dumb, I shorted my motherboard with a USB A, it felt like I got it right, and pushed it in, at which point it shutdown. Fortunately there didn't seem to be any permanent damage