Archive for March 15, 2023

Wednesday, March 15, 2023

GPT-4

OpenAI (Hacker News):

GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks. For example, it passes a simulated bar exam with a score around the top 10% of test takers; in contrast, GPT-3.5’s score was around the bottom 10%.

[…]

We are releasing GPT-4’s text input capability via ChatGPT and the API (with a waitlist). To prepare the image input capability for wider availability, we’re collaborating closely with a single partner to start. We’re also open-sourcing OpenAI Evals, our framework for automated evaluation of AI model performance, to allow anyone to report shortcomings in our models to help guide further improvements.

Hartley Charlton:

Apple is testing generative AI concepts that could one day be destined for Siri, despite fundamental issues with the way the virtual assistant is built, the New York Times reports.

Employees were apparently briefed on Apple’s large language model and other AI tools at the company’s annual AI summit last month. Apple engineers, including members of the Siri team, have reportedly been testing language-generation concepts “every week” in response to the rise of chatbots like ChatGPT.

Previously:

Update (2023-03-20): Gary Marcus (via Hacker News):

Chomsky co-wrote a New York Times op-ed the other day, and everyone is out there once again to prove they are smarter than he is, in the smuggest possible language they can muster.

Update (2023-03-22): Bill Gates:

In my lifetime, I’ve seen two demonstrations of technology that struck me as revolutionary.

[…]

I thought the challenge would keep them busy for two or three years. They finished it in just a few months.

In September, when I met with them again, I watched in awe as they asked GPT, their AI model, 60 multiple-choice questions from the AP Bio exam—and it got 59 of them right. Then it wrote outstanding answers to six open-ended questions from the exam.

Update (2023-03-24): DV (via Hacker News):

You might know that MSFT has released a 154-page paper on #OpenAI #GPT4, but do you know they also commented out many parts from the original version?

A thread of hidden information from their latex source code.

Aiko 1.0.3

Sindre Sorhus:

High-quality on-device transcription. Easily convert speech to text from meetings, lectures, and more.

The transcription is powered by OpenAI’s Whisper model running locally on your device.

It’s free. There’s even an option to translate to English.

Previously:

Printing Lazy Web Images

Adam Engst:

Mail generates an offscreen web view, waits for the “load” part of the document to arrive, and then “prints.” But loading="lazy" causes images to load only when scrolled into view, which never happens with printing.

[…]

I did testing and found that if I changed the scaling percentage to 99%, some images would load—but they’d be fuzzy—and others would remain blank. However, if I changed the scaling percentage more seriously, such as to 87%, and then scrolled through the entire document preview, all images would appear crisply in the resulting PDF, even if I later returned the scaling to 100%. Presumably, asking the Print dialog to scale the output forced WebKit to load all the images, allowing them to appear.

This is a longstanding problem that also affects pages saved as PDFs from Safari and third-party apps. WebKit should just handle loading the whole page automatically.

Why a Separate App for Classical Music

Jessie Char (via Nick Heer):

Now here’s The Nutcracker. The composer is in the album title, the performer and conductor are listed as the artists, each track has a byline for the conductor, and while all 17 tracks are unique you can’t see the bit that sets them apart. And the play preview displays even less.

In fact when you tap to full screen the song it still doesn’t display the full title, you have to wait a few seconds for it to start scrolling.

And it doesn’t show it at all in CarPlay. Will that get Apple Music Classical, too?

The year refers to the year this particular version was was recorded, not the year the piece was composed. Year of composition AND year of recording are both important.

The “artist” tag is also really tricky, and often this’ll be tagged with composer, ensemble, soloist (if relevant), and conductor. It’s a lot of names, and these often get clipped or some just don’t get added to the mix.

[…]

Take Apple News vs iBooks- they’re both reading apps, so why is it necessary to have two separate apps? Because they’re extremely different reading and writing experiences and it doesn’t make sense to lump them into a one-size-fits-all approach.

Previously: