Akseli Palén

A web developer and designer with too much inspiration.

[Akseli Palén @AkseliPalen]

@DeltyThe73rd Finally people can look at their phones safely during walking.

[Santiago Ortiz @moebio]

Interactive visualization of chatgpt completion probabilities, I ask it to to complete the prompt "When " hundreds of times.

[Tim Holman @twholman]

Coded out this optical effect I'd seen last week... you should be able to pause the video any time, and the cube will disappear! Which kinda makes sense, as its not flipping the pixels, so you don't see the movement, but still its a cool one!

[proper @ProperPrompter]

Sora.

New videos just dropped and they're wild.

9 new clips:

1)

[Akseli Palén @AkseliPalen]

@jamiegraydesign As a ZUI dev, I can agree that sidebars appear no matter what. I think it boils down to differences in human perception and preferences. Some of us prefer maps (non-linear layout), others prefer list (linear layout). Sidebar provides the list.

[Lex Fridman @lexfridman]

Journalism seems to increasingly optimize for drama over truth.

It uses everything from misrepresentation of cherry-picked quotes/events/data to downright fabricated info and lies, which bait the "other side" to respond and thus generate even more drama, creating a vicious cycle.

Plus, Wikipedia cites it, giving the drama more legitimacy and permanence.

This is a broken system. It adds to cynicism, distrust, and division in the world.

I've been increasingly the target of it 😔

But I maintain hope that truth & compassion wins out in the end.

[Akseli Palén @AkseliPalen]

@lexfridman "Eventually" is a strong word with repeated trials of small probability events. The question becomes: how likely AI kills all humans before something else does.

[Lex Fridman @lexfridman]

Propaganda machines are hard at work stoking the fires of outrage and division, while more and more innocent people suffer.

[Akseli Palén @AkseliPalen]

@housecor Maybe comments should not be so structured as JS Doc forces them to be. Structure in code is enough. I created Yamdog to create API docs without such structural constraints. Maybe Yamdog + TypeScript types is something I should look into. https://t.co/HNHD49JtpX

[Lex Fridman @lexfridman]

This is a time for great leadership from US that helps unite the entire Middle East not divide it.

I pray for peace ❤

[Yann LeCun @ylecun]

Very interesting paper: using generative AI to produce text or images emits 3 to 4 orders of magnitude *less* CO2 than doing it manually or with the help of a computer.

https://t.co/ErIPs4jCpM

[Lex Fridman @lexfridman]

"The two most powerful warriors are patience and time." - Leo Tolstoy

[Santiago @svpino]

Here is why developers don't trust research papers anymore:

I ran a poll with more than 3,500 answers, and 80% of the respondents find ChatGPT helpful when writing code.

But last week, a new research paper came out. They compared ChatGPT with Stack Overflow using Software Engineering questions.

Their conclusions:

• 52% of ChatGPT-generated answers are incorrect.
• 62% of the answers are more verbose than human answers.

Some prominent people have already echoed these results. To them, ChatGPT is nothing else than a parrot unable to write any useful code.

But if you dig deeper, you will see what's going on:

The paper used ChatGPT 3.5-Turbo. Why would they use a worse model when GPT-4 and Code Completion have been available for a while?

Right from the start, this research feels disingenuous. Why didn't the authors use the best version of ChatGPT?

The methodology to evaluate the model follows what others have already done: either the model answers flawlessly or is a failure. This is a massive disconnect between studies like this and real life.

I've used ChatGPT a lot since it came out. It makes mistakes. While these answers might be incorrect for research experiments, they are still helpful.

I’m much better using ChatGPT than without it, and any research paper saying otherwise is disconnected from reality.

Here is the paper: https://t.co/awotZzKWGy.

80% of developers find ChatGPT helpful.

Why do you think so many people don’t want to accept this?

[Akseli Palén @AkseliPalen]

Bugs in graph layout algorithms can sometimes produce novel results. Let's call this the pincushion layout 😅 #graphics #DataAnalytics #visualization #algorithm

[Akseli Palén @AkseliPalen]

CIELAB is a cool color space. Explore this infinitely zoomable color explorer for starters. Made with Tapspace.js #javascript #zui #zoom #cielab https://t.co/TpeEZuWapc

[AK @_akhaliq]

Meme legends 3 photoshop generative fill AI by @SavvyDone

[Akseli Palén @AkseliPalen]

@anatudor Fishes swimming in a shallow lake, dashing freely around rocks and water lilies, versus a tiny aquarium with a cheap plastic bridge and a bubbling treasure chest.

[Moshe Vardi @vardi]

For nerds :-)