dilap

8 ms·

91.
by dilap 1y ago
I guess at a very high level, the question is, do you think the current system and what it incentives is fine/optimal (and are sanguine presumably then about things like the the Lesné Aβ*56 fraud or the OP article [the failure of over
92.
by dilap 1y ago
> Science selects actively against people who react strongly to incentives I don't think this is true at all! If it were true, we would not have the reproducibility crises and various other scandals that we do, in fact, have. Scient
93.
by dilap 1y ago
Of course people will follow their own internal values in some cases, but we really want to arrange things so that the common and incentived path is the happy path! And without the proper systemic arrangements, people with strong internal v
94.
by dilap 1y ago
Well, don't forgot I also said this! > With whatever sort of due process is needed to make this reasonable Is it not reasonable to not continue to fund scientists whose results consistently do not reproduce? And should we not spend
95.
by dilap 1y ago
Yeah "individuals do better" is never the answer -- you've got to structure incentives, of course. I don't think you want to slow down publication (and probably peer review and prestiage journals are useless/obsol
96.
by dilap 2y ago
We may catch 'em on a per-capita basis too, if current immigration, emmigration, and tfr trends continue.
97.
by dilap 2y ago
Yeah I gave Claude Code a try at about 5 different things, with miserable results on all of them (insult to injury -- each time it charged me about a buck!). I wonder if because it was C# with Unity code, maybe not so heavily represented in
98.
by dilap 2y ago
Heh, yeah. The protobuf people's expectation was if you had a really large dataset you'd wrap it up in your own mini-protocol of "sequence of protobuf messages". But of course that's way more friction, so in practic
99.
by dilap 2y ago
protobuf sure, but streaming libraries for json (and xml, as in the parent) are extremely common. not harder (maybe even easier) than non-streaming to write, tho more cumbersome to use, so something you'd reach for only if you specific
100.
by dilap 2y ago
They also have a website where you can see a bunch of misaligned examples, they are absolutely hilarious. Very fascinating result. It suggests that to respond to innocent code queries with malicious code, the model has something like a &quo
101.
by dilap 2y ago
I remember being a little kid playing with the landline, just punching the numbers randomly, and eventually the operator came on and told me to stop playing with the phone!
102.
by dilap 2y ago
Haven't used openai in a bit -- whyyy did they change "system" role (now basically an industry-wide standard) to "developer"? That seems pointlessly disruptive.
103.
by dilap 2y ago
Yeah, I agree it would be great to do some regional experiments.
104.
by dilap 2y ago
The only one I know of is Popeyes -- they famously never stoped frying their chicken in beef tallow. Probably unhealthy for about a million other reasons, but hey, at least it's not fried in vegetable oil!
105.
by dilap 2y ago
I'd take the other side of that bet. https://imgur.com/a/4b0O4AW Prior to early 2000s when sugar consumption started going down, it was probably a reasonble guess as one of the drivers of increasing disease. But s
106.
by dilap 2y ago
Looks amazing -- hopefully some day they'll add git-lfs support and I'll finally be able to try it. :-)
107.
by dilap 2y ago
Yeah, the devices were still new enough that they seemed kind of magical. I think it may all have been a terrible mistake though -- the world is more fun when people aren't glued to their phones.
108.
by dilap 2y ago
I was going thru a box of old things the other day and came upon an old iPhone 4. God, what a beautiful device. In my opinion, the peak of iPhone design. I would love to have a phone that small again.
109.
by dilap 2y ago
Plus even if you were using lots of deps, the Go code would almost certainly keep working, thanks to minimal version selection and the Go module mirror. (And of course for maximum reliability, could even easily vendor deps with "go mod
110.
by dilap 2y ago
Technical issues happen, but I wish they would've put up a YouTube stream or something (or at least asked YouTube to stop taking down the indie streams that were popping up). It seems like basically their duty to the boxers and the fan
111.
by dilap 2y ago
Personally I am convinced LLMs do have real understanding, because they seem to respond in interesting and thoughtfull ways to anything I care to talk to them about, well outside of any topic I would expect to be captured statistically! (In
112.
by dilap 2y ago
By intellectually understand, I just mean you can ask Claude or ChatGPT or whatever, "how can I recognize if a string is in a^n b^n? what is the language being described?" and it can easily tell you; if you were giving it an exam,
113.
by dilap 2y ago
Regarding your linked comment, my takeaway is that the very theoretical task of being able to recognize an infinite language isn't very relevent to the non-formal, intuitive idea of "intelligence" Transformers can easily in
114.
by dilap 2y ago
My understanding is that Tesla's approach is to not require hyper-detailed maps. So far it doesn't work well enough, but if they do manage to get it working, then I think it will actually be a "boom, suddenly it works every
115.
by dilap 2y ago
Depends on if Tesla succeeds or not. I'm betting they will.
116.
by dilap 2y ago
I did find codebuff a lot easier to install and get started with...usability can make or break a project. Just as a user, I think it's nice to have multiple projects doing the same thing -- exploring more of the solution space. (I&#x27
117.
by dilap 2y ago
Ha! Slightly more complex than that ;-) https://gist.github.com/erinok/4399bbb1ea535deac051781eb9177...
118.
by dilap 2y ago
I promise you this is not the case.
119.
by dilap 2y ago
The very first Go code I ever wrote, way back in 2011, still compiles and runs perfectly. It's glorious.
120.
by dilap 2y ago
gotchya. i do think that is a cool idea actually -- LLMs tiny enough to do useful things with formally structured output but not big enough to nail the structure ~100% is probably not an empty set.

More ›