Even in the original comic, that would have been appropriate, I think.

At one point the user linked to a rust-lang forum thread from 2016-2019 as evidence that Jai has “some of the tools to make the code language agnostic” or something like that. The thread started with a discussion of array-of-struct vs struct-of-array data layouts, which of course has nothing to do with making code “language agnostic.” The user also mentioned the coding influencer lunduke multiple times. So I think they are simply misinformed on a lot of points, and I doubt they’re in the closed beta for Jai.

(I read some of the comments simply because I had the same question you did. And, as it happens, the last post in the forum thread I mentioned was written by me, which was a funny surprise.)

Exactly: that’s tight feedback loops. Agents are also capable of reading docs and source code prior to generating new function calls, so they benefit from both of the solutions that I said people benefit from.

As an even more obvious example: students who put wrong answers on tests are “hallucinating” by the definition we apply to LLMs.

making the same mistakes

This is key, and I feel like a lot of people arguing about “hallucinations” don’t recognize it. Human memory is extremely fallible; we “hallucinate” wrong information all the time. If you’ve ever forgotten the name of a method, or whether that method even exists in the API you’re using, and started typing it out to see if your autocompleter recognizes it, you’ve just “hallucinated” in the same way an LLM would. The solution isn’t to require programmers to have perfect memory, but to have easily-searchable reference information (e.g. the ability to actually read or search through a class’s method signatures) and tight feedback loops (e.g. the autocompleter and other LSP/IDE features).

This seems like it doesn’t really answer OP’s question, which is specifically about the practical uses or misuses of LLMs, not about whether the “I” in “AI” is really “intelligent” or not.

Agile Meridian / Post Manager

Thanks for sharing this! I really think that when people see LLM failures and say that such failures demonstrate how fundamentally different LLMs are from human cognition, they tend to overlook how humans actually do exhibit remarkably similar failures modes. Obviously dementia isn’t really analogous to generating text while lacking the ability to “see” a rendering based on that text. But it’s still pretty interesting that whatever feedback loops did get corrupted in these patients led to such a variety of failure modes.

As an example of what I’m talking about, I appreciated and generally agreed with this recent Octomind post, but I disagree with the list of problems that “wouldn’t trip up a human dev”; these are all things I’ve seen real humans do, or could imagine a human doing.

Well now you’ve seen it elsewhere, too.

distributing relay knowledge among chatters (TBD)

This is the core reason that centralization is currently necessary. So admitting that it’s an unsolved problem for a federated alternative is basically reinforcing Signal’s point.

That’s because you haven’t unlearned it yet

String escaping sucks in bash and other posix-style shells too, though.

But that’s not actually true in general; there is a default branch concept in forges, and an integration and/or release branch in most recommended workflows. That’s the trunk.

Fair, but it’s one that the typical tools for finding bugs, tests and static analysis, cannot actually help with.

Believe me, whitespace-correct scripting is absolutely an issue.

You’re right that it’s annoying when filenames diverge right at a character that must be escaped.

For interactive use, tab-completion essentially makes this a non-issue, because shells add escaping in the appropriate places.

For scripting, where spaces are harder to deal with, unfortunately there’s just not much you can do; your two options are basically to learn all of your particular shell’s patterns for dealing with whitespace in filenames, or only write scripts in something other than a POSIX shell.

That’s fair; Python, Swift, and most Lisps all use or have previously used reference-counting. But the quoted sentence isn’t wrong, since it said no “garbage collection pauses” rather than “garbage collection.”

“Garbage collection” is ambiguous, actually; reference counting is traditionally considered a kind of “garbage collection”. The type you’re thinking of is called “tracing garbage collection,” but the term “garbage collection” is often used to specifically mean “tracing garbage collection.”

To be fair, the drop/dealloc “pause” is very different from what people usually mean when they say “garbage collection pause”, i.e. stop-the-world (…or at least a slice of the world).

This is hilarious. I’m not sure how often anyone would actually need to verbalize arbitrary binary data, but I do see an advantage over base64 since the English letter names are so often phonetically similar.