On Thu, Dec 2, 2021 at 11:52 PM Nil Geisweiller <[email protected]> wrote:
> > > These days, I don't like NARS or PLN, because I think I know how to > > learn common-sense reasoning "from scratch". That includes automatically > > learning PLN-like or NARS-like rules, whatever they may be, as > > appropriate to whatever setting they've been learned for. Doing this is > > on my TODO list, but at the current rate of progress, this is at least > > a decade away. > > Interesting. I gave a talk on that at AGI 2021. My todo list is to convert it to a blog post. I guess though in the end you still need a metric of > comparison. > I don't mean to discourage you. I would still like to see a faster URE/PLN and also an easier-to-use/comprehend URE/PLN. I'd like to see more users, so the easier-to-use part gets higher priority. In the past, I've glimpsed how a certain subset of openpsi is a lot like the URE, and I'd like to see that explored some more. It would be interesting if both could be made to have very similar or the same API, allowing comparisons. As long as it is just you and I (and occasionally Ben) in these conversations, it remains hard to pursue these ideas. Sometimes I wonder if NAL isn't actually based on probability theory > with a very unconventional prior. > Huh. Maybe? The axioms of probability theory are very generic; it would indeed be hard to evade them. Perhaps the underlying set on which NAL acts is not what it seems to be. --linas -- Patrick: Are they laughing at us? Sponge Bob: No, Patrick, they are laughing next to us. -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/CAHrUA35Zv5Pvmsz6%2BY4_BrEACRNHrezfev%2BfANGdiKFn6du6Aw%40mail.gmail.com.
