Ask a LLM: SQL vs. CSS What’s the Difference? Which Is Better?
Following Dave's attempt to get ChatGPT to implement SQL, I bring you another burning SQL-related question:
Has any human been involved in authoring this? (I'm wondering what "Liam Frady" is an anagram of. The history-computer site seems rather unforthcoming.)
Why did any human even think it was a sensible comparison?
There seem endless opportunities for lazy journalism on the pattern of 'XXX vs. YYY: blah blah blah?'
Here's a trickier one: Haskell vs. PureScript: The difference is complexity Those are at least two relatively comparable programming languages, one deriving from the other. If you read the second half first, and didn't know that language too well, you might think the article has something to say. But the first half is more or less word-salad.
Again, was a human involved?
It's certainly possible that ChatGPT or similar were involved, though the articles don't have the typical easily-recognised structure. They remind me more of student essays I used to mark by the bucketload.
Therefore, I'm inclined to avoid attributing to AI what can be adequately explained by bad human writing.
Or bad writing by humans. Etc.
Quote from Dave Voorhis on July 27, 2023, 9:31 pm... bad human ...
Heh, heh, we can all write badly. I guess for a long time there's been 'reportage' on science topics that merely scraped stuff off a google search. Are the 'ask a LLM a question's just a more sophisticated form of that?
I think LLM's are now able to not only produce sentences that flow like sentences, but also a paragraph of a few sentences that manage to remain on-topic. What's beyond them is to put together a whole article with a beginning, middle and end that examines a topic coherently.
The backstory is I was looking for something to compare Haskell closely with Purescript. And came across the second article I linked; and thought it contained some applicable keywords, but failed to connect them coherently, or use the concepts to measure how one language compared to the other. So I asked on a Haskell form for opinions on the article. There were no responses on the substantive topic, merely commentary this sounded LLM-generated. And somebody linked to the SQL piece.
They remind me more of student essays
So: who set that as an essay topic? Why did they think anything meaningful could be said? There'd be an infinite choice of topics of the form: TLA1 vs TLA2 which is better?/what is the difference? (And were your students scraping stuff from Google in to their essays without it passing through the mind of either? -- to mis-quote Peter Ustinov.)
The other shibboleth is these articles are embedded in advertising. Some are for generic retail offerings -- but why would the generic public be looking at SQL?; some are industry-specific offerings -- but why would a decision-maker with a budget 'trust' a product appearing alongside such drivel? (Also the tech-savvy would have suppressed the ads.)
Quote from AntC on July 28, 2023, 4:01 amQuote from Dave Voorhis on July 27, 2023, 9:31 pm... bad human ...
Heh, heh, we can all write badly. I guess for a long time there's been 'reportage' on science topics that merely scraped stuff off a google search. Are the 'ask a LLM a question's just a more sophisticated form of that?
I think LLM's are now able to not only produce sentences that flow like sentences, but also a paragraph of a few sentences that manage to remain on-topic. What's beyond them is to put together a whole article with a beginning, middle and end that examines a topic coherently.
They can put together whole short articles that examine a topic superficially, though they have an almost obviously templated structure.
The backstory is I was looking for something to compare Haskell closely with Purescript. And came across the second article I linked; and thought it contained some applicable keywords, but failed to connect them coherently, or use the concepts to measure how one language compared to the other. So I asked on a Haskell form for opinions on the article. There were no responses on the substantive topic, merely commentary this sounded LLM-generated. And somebody linked to the SQL piece.
They remind me more of student essays
So: who set that as an essay topic? Why did they think anything meaningful could be said? There'd be an infinite choice of topics of the form: TLA1 vs TLA2 which is better?/what is the difference? (And were your students scraping stuff from Google in to their essays without it passing through the mind of either? -- to mis-quote Peter Ustinov.)
The other shibboleth is these articles are embedded in advertising. Some are for generic retail offerings -- but why would the generic public be looking at SQL?; some are industry-specific offerings -- but why would a decision-maker with a budget 'trust' a product appearing alongside such drivel? (Also the tech-savvy would have suppressed the ads.)
They look like content created for the sole purpose of raising Google search rankings.
Imagine you set up a Web site selling some rubbishy product. Nobody will find it because it will be at the bottom of Google search rankings, so you need some Search Engine Optimisation to artificially inflate the search ranking of your Web site. That means people Googling for something else will be more likely to find it listed toward the top of their search results. The idea is that they'll hopefully then buy whatever it is you're selling, though that almost never happens.
A typical way to optimise (?!) Google rankings is to fill your Web site with "content," i.e., stuff that humans might want to read. Since writing stuff people want to read is really hard, almost nobody bothers to try, and because Google ranks pages by inbound links and some automated (and largely undocumented) notion of content quality that has little to do with its actual appeal, humans game the system by filling their otherwise-rubbish Web sites with AI-generated or human-generated content. It's also rubbish, but that doesn't matter.
The important thing is that there be a lot of it, because that can fool the Google content-quality detectors into thinking your rubbishy Web site is a goldmine of genuine, worthwhile, worthy-of-search-results content.
If it's just good enough to fool people into linking to it, even if only to critique it on obscure technical forums, that's a bonus that will push it even further up the search rankings.