Pinned toot

🔔 ☎️ 🔔 ☎️ 📞 hello qoto🎵 🎶 🎼 🎵 🎶 🎼

an example of the plausible-sounding bullshit (about the Venpa rhyming scheme in ). this one’s not in the “abcb” scheme, there’s no “third” or “fourth” lines, and I think this is not even a Venpa:

Show thread

is a great school-essay generator!

It always sounds confident and formal - regardless of whether its showing deep understanding and insight, or spouting plausible-sounding bullshit vaguely related to the question.

Exactly what students do!

Sundar boosted
Sundar boosted

@Fanua Here, just 2 proposals created by DALL-E, the #AI at labs.openai.com for the phrase " #JohnMastodon, the trumpeter of Mastodon". Scary, but cool 😉

Sundar boosted

Proposal: How about if we stop talking about #ElonMusk and #Twitter on here? 😁

Sundar boosted

What percent of published empirical papers in your field do you think are fraudulent (pick the closest number)?

Sundar boosted

It's looking like AI, like generations of teenagers before it, has learned to BS its way through a five paragraph essay years before it's safe behind the wheel of a car.

Free, light, autumn leaf
Sans home, drifts around, detached
Brittle, dry, fading.

@actuallyautistic

It makes a big difference to go from

“I feel like I wanna cry for no reason at all and I don’t know what to do about it” (- helplessness)

To

“I feel like I wanna cry for no reason at all. Why could it be?
It feels like I’m overwhelmed. By what?
Two days of constant masking in a strange new place is probably it (even if the people around were lovely). What can I do about it?
Quiet time, rest, giving myself permission to disengage from social obligations for a while.” (- making sense of it, understanding myself, and self-care)

It’s taken many years to arrive here, but I’m happy to notice that I’ve made progress.

Sundar boosted

I wish to use a database to store my experimental results, since I plan to obtain A LOT of data points and the good old csv table doesn't appear as a good solution anymore.

How would I go about making the structure of the database? Do you have any advise for this kind of things?
I will often have to change procedures according to the results I get, thus I would need to be able to understand what was the procedure used for each results and easily see which ones are the most recent and up to date.

Sundar boosted

Kids this is a reminder to collect physical copies of your favorite media. Corporations will axe your favorite show/movie/book in front of its family and make it a felony to retrieve it. Have a library.

Sundar boosted

If ever there was a reason to make , and mandatory in certain cases:

“Earlier this year, many people with Argus optical implants – which allow blind people to see – lost their vision when the manufacturer, Second Sight, went bust. […] There’s no ethical case for permanently attaching computers to people’s nervous systems without giving them the absolute, irrevocable right to nominate who maintains those computers and how.”

pluralistic.net/2022/12/12/uns

Sundar boosted

If you wanna obsess about anything other than 1) that other site, or 2) how this is *not* that other site, please come sit by me.

Bears? You like bears? Pre-Colombian construction technique? Why earmarking in books is a mortal sin? The Red Queen hypothesis? A really cool coprolite you found the other day? Hell, I'll give you 20 minutes even if it's just about how Star Wars is high art.

(If you're coming through the kitchen, grab me a beer, too.)

Sundar boosted

Basic Autism description 

@ratcatcher

I usually describe it as NT people have a brain that has many pathways already laid down : they can 'naturally' walk down a path without thinking, and just 'do stuff' easily.

My brain is like fighting through a jungle : the first time I have to force my way through a path with willpower, hacking my way through. This is very stressful and difficult, the second time still pretty tough, but it gets easier each time.

Many times it feels like I have to learn by trial and error, bumping into every single small obstacle along the way. I need clearer guidance and often ask 'stupid' questions, things that are obvious to NT I have to confirm.

This means small variations in what I do, coping with differences is much more effort - straying off the path is harder for me. I like doing things the same way each time.
:)
#ActuallyAutistic

Sundar boosted

What happens to your smartphone when it gets stolen?

I recently had my iPhone stolen while on a business trip in London and, through the wonders of Find My, have been able to track its journey in the past few weeks. I found it to be a fascinating insight to what appears to be a rather sophisticated phone theft operation and thought I'd share!

After playing with for a bit, I’m still impressed by what it’s capable of achieving (sometimes), but I can see that it still has an ELIZA side to it - a tendency to go on random tangents,
to repeat answers,
to ignore chunks of the question and make up its own unique interpretation of it.

ChatGPT stuff 

Zooming out to generalize, it seems likely that this’ll be a significant issue with ChatGPT/other LLM responses - mistakes like this are subtle and insidious, and could easily pass muster at a first glance. Especially if you don’t know the answer already, which will be the case when people actually start using it beyond this trial.

Information about things that are already popular and has a lot of text on the Internet - pre-established languages, technologies, the currently dominant cultures, ideologies, etc. - will be well known to ChatGPT, and could be commingled in subtle ways with newer upcoming ones, and presented as part of that newer entity, perpetuating those old influences.

This type of misinformation will be insidious, and can easily fly under the radar in a lot of situations.

Show thread

ChatGPT stuff 

tl;dr: it seems likely ChatGPT took advice meant for Python or MATLAB, and substituted Julia’s name in their place because it considers them “close enough”.

1 isn’t wrong, but makes no mention of type stability, which is at least as important as its (more generic, applicable many languages) suggestion.

2 is also generic advice that’s good to have, but isn’t specific (but I consider this one a success)

3 rates as kinda okayish advice. One could argue that for a beginner, the base functions offer a solid place to start, if we assume they’re gonna write badly optimized code (but then teaching them how not to do that - as the prompt asked - is a better way to solve that). But since this is Julia, it’s not uncommon that simple custom code you write beats the obvious ways using built-ins. So I’d consider this potentially misleading.

4 is likely the smoking gun here - the repeated mentions of vectorization and asking to use techniques to “vectorize your code” seems to suggest that this whole thing was taken from guides written for Python or MATLAB, which are at this point more numerous than those for Julia, and then the Language Model substituted Julia in the language name’s place because it considers them similar. (A previous answer said “there are a number of programming languages that are commonly used for numerical and scientific computing, including Python, Julia, MATLAB, R, and others” - so it knows they’re in the same category.)

Show thread

ChatGPT stuff 

[ChatGPT’s full response below:]

There is no single secret to making Julia code fast, as the performance of Julia code can depend on a number of factors. However, there are a few general strategies that can help to optimize the performance of Julia code and make it run faster, such as:

  • Choosing the right data types and structures: Julia is a dynamically-typed language, which means that the type of a variable can change at runtime. This can make it difficult to optimize the performance of your code, because the compiler cannot always determine the best way to represent and manipulate your data. To make your code run faster, you should choose the appropriate data types and structures for your data, based on the operations you want to perform and the performance characteristics of the different types.
  • Avoiding unnecessary computations: Inefficient code often contains unnecessary computations, such as calculations that are performed multiple times or operations that do not affect the result of the program. To make your code run faster, you should avoid performing unnecessary computations, and try to structure your code in a way that minimizes the amount of work that the computer has to do.
  • Using built-in functions and libraries: Julia has a rich set of built-in functions and libraries that can help you perform common tasks efficiently. Rather than implementing your own algorithms or functions, you should use the built-in functions and libraries whenever possible, as they are often highly optimized and can save you a lot of time and effort.
  • Exploiting parallelism and vectorization: Julia is designed to support parallel computing and vectorization, which can help you to make your code run faster by utilizing multiple cores or processors. To make the most of these features, you should try to structure your code in a way that allows it to be executed in parallel or vectorized, and use the appropriate techniques and tools to parallelize or vectorize your code.
Show thread

ChatGPT stuff 

I asked “What is the secret to make Julia code fast?” (just to see if it picks up anything from Jakob Nissen’s excellent guide viralinstruction.com/posts/opt or the many other optimization guides online).

The answer turned out to be good example of the model’s ability to be subtly misleading while not being entirely wrong.

(Thread, with ChatGPT’s reply as the next post.)

Show more
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.