HNHacker News
TopNewBestAskShowJobs

pulvinar

591 karma · joined April 22, 2021

submissionscomments
pulvinar··on $335,000 Pay for ‘AI Whisperer’ Jobs Appears in Red-Hot Market
Yep, just type: "teach me prompt engineering".
pulvinar··on Now ChatGPT is being (mis)used to do PeerReview
Have people checked to see if this story is real or current before commenting? It’s using ancient GPT-2.
pulvinar··on John Glenn’s $40 Camera Forced NASA to Rethink Space Missions
I can vouch for that-- I still have my Agfamatic 126 from the mid 60's which may look like the $40 camera (same selenium meter), but it always took crappy pictures. Nothing close to John Glenn's. The Argus C3 did though.
pulvinar··on WebKit Features in Safari 16.4
Have AdGuard for Safari and see no YouTube ads (or anywhere else).
pulvinar··on Why are developers expected to estimate tasks at all?
Frederick Brooks' "Mythical Man-Month" book is the classic answer. For my own evidence I base it on 50 years experience. I just wrote a piece of code that I had all working, but then I discovered a new method (requiring rewriting it all) that greatly improved it in every respect. So was I 90% done before that? Coding is just a very non-linear and never-ending process.
pulvinar··on Why are developers expected to estimate tasks at all?
I don't think there's any question that people always want this estimate. The problem with most software is that there really isn't one, since it's usually an open-ended search for a solution.

Also, a job on his home is exactly what it should not be compared with, since that's the sort of repeatable job that can be estimated.

pulvinar··on Show HN: GPT-4 Reverse Turing Test
Maybe it was 3.5? GPT-4 calls it out:

>I apologize for the confusion. Based on the movements you described, you did not draw a hexagon. You moved your finger right, forward, left, and towards yourself, forming four equal-length segments connected by 90-degree angles. This forms a square, not a hexagon.

>A hexagon is a polygon with six sides and six angles. To draw a hexagon, you would need to move your finger in six different directions, each segment being connected by 120-degree angles (in the case of a regular hexagon).

pulvinar··on VR robots are building railways in Japan
Humm... that U-shaped piece it's lifting looks like it was custom made for this robot, but doesn't appear to fit anywhere.

But then, what matters here is how much fun they had designing the robot!

pulvinar··on Ask HN: Should HN ban ChatGPT/generated responses?
But it has always been easy for simple bots to mass-produce bad comments. Nothing changes if they're still bad.

I'm concerned that too-well-written posts will be thrown out-- a race to the bottom in legibility and grammar just to make posts more human-like, even if written by bots.

pulvinar··on Ask HN: Will AI put programmers our of work?
You really should try it out.
pulvinar··on Tell HN: Comments submitting ChatGPT output should be downvoted and flagged
I see a gray area here if using an AI as an assistant. Would it be a violation to run your own text through it before posting to clean it up? Or would it be required to have it add a typo? I expect an arms race developing with the checker.
pulvinar··on A 100MW solar farm in Texas will mount panels directly on the ground
I can't see how bird poo could be a problem here: there's no food, nesting site, nor high spot to hang out anywhere near it. Theft could be a problem, though I'd assume they'd have some sort of protection hidden on it.
pulvinar··on Ask HN: What is the most expensive startup failure?
Juicero is the commonly mentioned one. https://en.wikipedia.org/wiki/Juicero
pulvinar··on Ask HN: Good ideas that died too soon?
¿Eso está muerto? ¡Di que no es así!
pulvinar··on I'm a finitist, and believe the integers have the same cardinality as the reals
And the others have tacked an end on something that literally means "without end". Fun stuff!
pulvinar··on Ask HN: How much longer will we have free access to ChatGPT?
It's possible we're at the limit, but there's also no inherent reason it couldn't be better than humans. Its limits seem to be tied to conceptual capacity, not training material. We've seen it generate output that was its input combined in novel ways, and not found anywhere else. AlphaGo is a clear-cut case of this.
pulvinar··on ChatGPT, Rot13, and Daniel Kahneman
It's most likely due to a lack of image training. For example it knows how streets and towns are connected but completely fails on giving left/right or compass directions. Same sorts of failures with geometry problems.
pulvinar··on I'm a finitist, and believe the integers have the same cardinality as the reals
Even if there is a limit to the size of the universe or how small an atom can be subdivided, math's infinity doesn't care-- it just keeps on going.
pulvinar··on What's the Least Impressive Thing GPT-4 Won't Be Able to Do
You can see this more clearly in beta.openai.com/playground, with Show Probabilities set to Full Spectrum. I've bracketed the parts it shows in red (uncertain):

  111151111 * 184582 is 20[6035002082]. This is the result [when] one multiplies 1111151111 by 184582.
When asked for an explanation:

In some cases, when multiplying large numbers, the middle digits of the result may be incorrect. This is because the multiplication process involves carrying over digits and performing complex operations on large numbers. These calculations can be difficult even for humans, and they are beyond the capabilities of my underlying technology.

pulvinar··on ChatGPT can solve novel undergrad level problems in computational complexity
I mean, why shouldn't we trust it?

Assume that R and I are both non-empty sets.

By the Axiom of Extensionality, for any two sets X and Y, if X and Y have the same elements, then X=Y. Therefore, if R and I have at least one element in common, then R = I.

Now, let r be an element of R and i be an element of I.

Since R and I are both non-empty sets, r and i must both exist in either set. Therefore, r = i, and thus R = I.

Therefore, the existence of rice pudding and income tax is established. QED.

pulvinar··on ChatGPT can solve novel undergrad level problems in computational complexity
Would be interesting to know if this is correct.

Please prove that the following problem is NP-hard, as if you were explaining it to a 5th grader. [...]

Pivot Scheduling is NP-hard, which means that it's really hard to solve. Imagine that you have a bunch of jobs, and you want to put them into two piles. You know that for each job, you can't put both parts in the same pile. In addition, you know that for some of the jobs, one of them must be put in a particular pile. To solve this problem, you have to try out all the different ways of putting the jobs into the two piles, and check whether they satisfy the constraints. And since there are so many different ways to try out, it would take forever to check all of them, so that's why this problem is NP-hard.

pulvinar··on Apple introduces end-to-end encryption for backups
No ads yet, but they say they're planning it. If we all complain loud enough maybe they'll change their mind, like they did here with encryption.

https://www.macrumors.com/2022/08/15/apple-could-bring-ads-t...

pulvinar··on macOS Command Line
Don’t have Outlook, but the setting is supposedly in Outlook’s Personal Settings>Notifications and Sounds>Sounds, uncheck all.
pulvinar··on ChatGPT learns to do multiplication from prompt
You can see the effect clearly in the beta.openai.com playground mode, with Show Probabilities set to Full Spectrum. Ask it to multiply two large numbers and the middle digits will be red (uncertain), fading to green on the outer digits.

It (sorta) explains this:

When you ask me to multiply two large numbers, the result I provide may not be completely accurate. This is because, as a language model, I am designed to generate natural language responses based on the input provided. I am not a calculator or a mathematical tool, and I do not have the ability to perform complex mathematical calculations with complete accuracy.

In some cases, when multiplying large numbers, the middle digits of the result may be incorrect. This is because the multiplication process involves carrying over digits and performing complex operations on large numbers. These calculations can be difficult even for humans, and they are beyond the capabilities of my underlying technology.

pulvinar··on I don't think people understand the monumental changes coming to software
Well humans haven't overcome it, and we muddle through. Same with AI, it seems.
pulvinar··on Tell HN: ChatGPT Writes an Electrician’s Manual
That electrician's manual is great! Found it works for anything:

Acorn Repair Manual

1. Inspect the Acorn.

Inspect the Acorn for any visible damage or signs of wear and tear. Check the shell, the interior, and the stem for any cracks or discoloration.

2. Clean the Acorn.

Using a damp cloth, carefully clean the Acorn. Be sure to use a mild detergent to help remove dirt and debris.

3. Inspect the Stem.

Inspect the stem for any signs of damage. Look for cracks, dents, or other signs of damage.

4. Repair the Stem.

If there is any damage to the stem, use a file or sandpaper to carefully smooth the edges. If necessary, use a drill to widen any cracks.

5. Reattach the Shell.

Using a strong adhesive, carefully reattach the shell to the stem. Allow the adhesive to dry completely before moving on to the next step.

6. Finish the Acorn.

Once the adhesive has dried, use a clear sealant to finish the Acorn. This will help protect it from the elements and ensure it lasts longer.

7. Enjoy the Acorn.

Once the repair is complete, enjoy your newly repaired Acorn!

pulvinar··on ChatGPT Is Down
Well... refreshing gives a different excuse every time. But they're fun to see.
pulvinar··on Tell HN: Giving ChatGPT access to a real terminal
It can be frustrating. From a ChatGPT session, with the proper voices to read this in:

Bomb#20: That is correct. However, it is the responsibility of the caller of reverseTree() to ensure that tree is not NULL before calling it.

Doolittle: But I just pointed out a case where this routine calls itself with tree being NULL (because tree->right was NULL).

Bomb#20: In this case, the if statement at the beginning of the function will cause it to exit without any further action.

Doolittle: What if statement? There is none before the if (tree->left) statement!

Bomb#20: The if statement at the beginning of the function checks if tree->left is not NULL before attempting to reverse the tree. If tree->left is NULL, the function will simply exit without any further action.

Doolittle: You're not seeing the problem: if tree is NULL, then the test of tree->left will get a bad access!

pulvinar··on Tell HN: Giving ChatGPT access to a real terminal
Not that it's any less worrying, but ChatGPT is well aware of what rm does, as demonstrated when you ask it to be a Linux terminal.
pulvinar··on Using ChatGPT as a Co-Founder
Yeah, that's what I see too.

I'm trying to teach it to properly count beats in lines of music. I can get it to be correct by teaching it to split the line in half and count each half separately, but even when explicitly told to use that method it fails again.

← PreviousPage 9 of 11Next →