zahlman

7 ms·

151.
by zahlman 2mo ago
> If LLMs get better at writing They definitionally cannot, for the definitions implied in the argument. The point is that good writing is a thing humans are capable of doing because they are human. > It's a sort of a claim t
152.
by zahlman 2mo ago
> absolutely targeted at modeling Not everyone accepts a simulationist view in which modeling something accurately enough inherently results in creating the actual thing.
153.
by zahlman 2mo ago
> The parent poster isn't inventing anything or deciding anything. Stating something authoritatively is, in fact, "deciding" it for others. > They're just sharing what the common understandings are. No, it is not a
154.
by zahlman 2mo ago
Indeed; but I've only heard of them being used for images rather than text. Why?
155.
by zahlman 2mo ago
GP's point is that Socrates' argument is not fairly comparable to the one made in TFA.
156.
by zahlman 2mo ago
Why should it die?
157.
by zahlman 2mo ago
It is consistently no longer working for me on desktop, FWIW.
158.
by zahlman 2mo ago
No, sorry. Political activists do not get to decide for everyone else what words mean.
159.
by zahlman 2mo ago
And people ask me why I complain about web pages failing to display basic content without JS. (I keep WASM and WebGL disabled in Firefox settings, too. Yes, I had another post ITT questioning why those things are any less secure than the Ja
160.
by zahlman 2mo ago
Okay, but we're still talking about running machine code inside the sandbox , where that functionality is still not available.
161.
by zahlman 2mo ago
Okay, and why is that more of a security risk than executing arbitrary JavaScript in the sandbox?
162.
by zahlman 2mo ago
Out of curiousity, do you ask your editing system for diffs? Seems to me like the best way to notice and review whether your "voice" is degrading. Personally I would never let an LLM touch my prose (although I'd happily use i
163.
by zahlman 2mo ago
I did read it. People are allowed to disagree with your conclusions. Comment guidelines ask us all not to make such accusations. To say that a statement is incomplete, but not strictly speaking wrong, is perfectly compatible with describing
164.
by zahlman 2mo ago
> We know exactly how LLMs work I think you imply a rather loose standard for "exactly" here. I wouldn't even say this of major deterministic software projects that are orders of magnitude smaller than frontier LLM weight-
165.
by zahlman 2mo ago
> or perhaps a human who has been reading too much Claudish recently To be fair, is there anyone who hasn't been "reading too much Claudish recently" who is also qualified to write on the topic?
166.
by zahlman 2mo ago
> What is true, however, is that in order to generate the next token autoregressive LLMs produce internal/hidden state about future tokens far past the next token so that it's not like the entire machinery of the algorithm depr
167.
by zahlman 2mo ago
> I'm not sure what you want me to do with that information For example, you could cite specific things that you believe to be "AI tells" or "admissions".
168.
by zahlman 2mo ago
Even so, one might wonder why we don't try making systems that take different approaches. For example, after a traditional first pass of output, they could do sliding-window "optimizations" considering each token in the con
169.
by zahlman 2mo ago
> generated an article that continually undermines its own main point. I disagree that this accurately describes TFA.
170.
by zahlman 2mo ago
What exactly does "RCE inside sandbox" describe that goes beyond "the webpage can supply arbitrary JavaScript and the JavaScript engine executes it", but is still isolated from the system?
171.
by zahlman 2mo ago
Well yes, that's the point. Burry had his success early and was not notable before that. Zitron has been outputting a continuous stream of bearish sentiment that hasn't come to fruition, from the moment anyone recognized the name.
172.
by zahlman 2mo ago
Anything where you use a state machine that has to remember the parent state is effectively an intrusive linked list (even if they rarely have more than two elements); similarly the most obvious implementation of undo/redo. In these us
173.
by zahlman 2mo ago
What are you even talking about?
174.
by zahlman 2mo ago
> That is: rather than saving either a light or dark theme preference permanently in localStorage, Dr. Verou’s implementation stores your preference only if that preference is different than your device setting. In that case, the site sa
175.
by zahlman 2mo ago
> as if you're supposed to buy like 8 H200s or something or code by hand is the solution People like GGP who have been coding for 20 years presumably are perfectly capable of doing so by hand. Or should be.
176.
by zahlman 2mo ago
Many replies are calling this out. Though it hurts to see one of them literally screen-capping and circling an em-dash and acting as if it were impossible for humans to produce this glyph in text on the Internet. It took me hardly any effor
177.
by zahlman 2mo ago
I assume the intent is to teach new programmers about less-than-obvious things that teachers might easily gloss over. But it comes across largely as FUD about things that are well understood in the field (and have been for decades) and are
178.
by zahlman 2mo ago
Is this sort of thing no longer part of a standard computer science or software engineering undergrad curriculum?
179.
by zahlman 2mo ago
I'm pretty sure the author there means to compare the intrusive linked list to a non -intrusive linked list (such as C++ std::list), not to vectors etc. that aren't "linked" at all. As described e.g. in https:/&#x
180.
by zahlman 2mo ago
>"Intrusive" got popular with C++ intrusive pointers It got popular with C++'s attempts at type safety. In particular, std::list lets you accomplish the machinery without macros, and allowing for polymorphism (heterogeneou

More ›