@pfau The “building machines that learn and think like people” paper radicalized me. I also really like the follow up work extending this sort of genre of work to LLM’s and probabilistic programs
https://t.co/FsYzM5LxNN
@anshulkundaje Fully agree. Do you think this can be addressed by just generating a lot of distinct GRN’s? Like try to infer the whole space of hypotheses that’s consistent w/ data?
I think the latest decision by Musk to reinstate an account that shared CSAM has cemented my decision to leave. Fuck that and fuck Elon too. I’m leaving my account but I’m not going to check Twitter anymore
I hope i can find many of you on other platforms
@KaijuUltimax @ChombaBupe I’m sorry if an interpreted language with reference counting memory management is high performance then i think we have different definitions of the term.
@KaijuUltimax @ChombaBupe Here’s a fun one: number of questions asked on SO about about one javascript framework versus both the C and C++ languages in their entirety
https://t.co/hfqfWX2RID
@kahlildozier@ChombaBupe Go to stack overflow and look at the newest questions. It’s pages and pages of questions like https://t.co/Tb0BetsBsm
I found one, maybe two questions that out of >100 were algorithmic. What does that tell you about the day to day work of most programmers?
@o_guest@ChombaBupe Ok some simple problems are hard for LLMs. That’s abundantly clear. It’s also clear that many simple problems are easy or even memorized by LLMs. That’s why they’re interesting/useful
The data requirement for training don’t have any bearing on the present utility of LLMs.
@o_guest@ChombaBupe Yeah i don’t have data to back that up
I think the point is that simple tests are much more likely to be in the subset of problems that LLMs can solve. So complex LLM code won’t always be correct but LLM test code can at least verify with high probability of being correct
@ChombaBupe Virtually no one is writing complex or performance critical code like you are. Copilot is quite good at boilerplate and thats what most programming is and probably what most stackoverflow questions are about: variations of trivial framework issues.
@o_guest@ChombaBupe I mean sure, but tests are often so simple that any reasonably competent programmer can verify that they work as intended. e.g. checking numbers are in order is trivial, quicksort is not
@vrgldh Ok that makes sense. I just think Nashville has a lot of room to grow. Plenty more people can live here. The whole “not everyone needs to live here” thing is backwards. Let people come and have opportunities like you did.
@vrgldh Ok I’m confused. You said not everyone needs to live in Nashville or LA and described how you lived in both Nashville and LA.
So I’m asking why you think not everyone deserves to live in Nashville or LA like you.
@vrgldh Those are good reasons, similar to mine. Presumably many other people have similar or better reasons to want to live here.
But your post implies that they don’t actually need to live here and should settle someplace less desirable. Why is that? Why doesnt the same apply to you?
What you are seeing on the radar are purple martins leaving their roost in a circle as they go out to look for food during the day. Once again, they have made a roost in downtown Nashville and will be there until late August before they migrate to Central America for the winter.