I wonder did they add that flag before or after the day when they accidentally made people see only Elon's tweets on their timeline: https://www.theverge.com/2023/2/13/23598514/twitter-algorithm-elon-musk-tweets

It will take time, but I'd imagine it should be possible to derive a method of determining this by observation.

Algorithms like this will have fingerprints.

What is the definition of "industry research" you are considering here?

This already happened, splitting into dozens of niches - it's just the niches didn't reform on Reddit. The ML community gradually migrated from here to twitter a few years ago.

I always find it curious that lot of these "godfathers of AI" seem to be a bit like this. It gets draining to listen to them, as they have a tendency to reframe any debate or definition just so they can be right.

They undeniably did copy them for training, which is the allegation. Not even Stability would deny that.

The question is whether doing that is legal. Plain reading of the US law suggests it is legal to me, but Getty will argue otherwise.

Have you tried finding a way to speak to an experienced product specialist at your EMR software provider about this? It may be something that could be achieved in your existing solution with an elegant configuration change.

Did she lose the ability to speak entirely, or is it more that it's become unintelligible? If the latter, and you haven't seen this already, I came recently came across a google project aimed at speech recognition for non-standard speech https://sites.research.google/relate/

And best wishes to both of you. Hope you eventually find a way for her to more easily communicate.

Not necessarily. This kind of thing will also happen if you chain upscaling, quantization, smoothing and sharpening techniques.

What's the video link?

It was just such a perfect match for the vibe of the time.

It was a really optimistic time in a lot of senses. A world of primary colors (before everything gradually got darker with 9/11 and the resulting war on terror propaganda), and the music mirrored that perfectly.

Smash Mouth's All Star video is almost like a Rembrant painting of what the mood was as a teenager back then. Ha.

https://www.youtube.com/watch?v=L_jWHffIx5E

It truly was. God the gigs were so much fun.

Yeah, I'm in the same boat actually - they sound vaguely familiar...

OP, there were a lot of pop-punk bands around that time. A LOT. You'd see entire festival line-ups that were pretty much just 100 pop-punk bands, a bit of ska, and maybe some nu-metal thrown in.

These guys probably had a following that any guitar band today would die for, but due to the sheer level of competition, didn't really capture wide attention.

There are food delivery drones in my area. You don't hear them. Their rotors are too big, and they fly too high, unlike infernal ratty photography drones you normally see.

You might hear them gently as they do a drop-off, but it's quieter than the car or motorbike that would otherwise be delivering the food.

> don't you hear the person's voice & the sound of the voice makes you think of the person & if you think of a person who makes you sick, it's gonna remind you how much they make you sick. so yeah, you can say you separate it, but realistically how much do you actually separate

I find this depends on what the artist is singing about. If it's something completely incongruent with their behaviour or beliefs, like say Chris Brown singing about women, I have to switch the song off.

But if it's something entirely unrelated, I don't really tend to get those thoughts as much.

Just in case you miss my other comment - chatgpt seems to actually be particularly expensive to run in comparison to their other apis. Altman says "single digit cents per chat".

Step 1 definitely explains why its responses often feel so similar to SEO waffle-farm content. I had been wondering where that aspect was coming from.

> Or just pay .004c per api query?

"average is probably single-digits cents per chat; trying to figure out more precisely and also how we can optimize it"

https://twitter.com/sama/status/1599671496636780546?lang=en

To confirm, 4k tokens is indeed what their FAQ says. https://help.openai.com/en/articles/6787051-does-chatgpt-remember-what-happened-earlier-in-the-conversation

I wonder if there is a significant difference between unskilled and skilled work once the threshold of versatility to do "most unskilled work" is met.

Spotify's collection of recorded classical music is vast, with many of the best ever recordings of pieces on there. Some famous pieces, such as beethoven's works, have 100 different versions.

It can be worth googling the piece, to see what comes up about it, and its history, and the different interpretations.

In terms of the range of difference between what the original might have sounded like, vs a modern interpretation, you could check out the Voices of Music version of Pachelbel's Canon, vs say the Neville Marriner version. The VoM version is performed on vintage instruments, and they try to adhere to what their belief of how it would have been performed, whereas Marriner's version is the more familiar version to our ears.

This is probably at the more extreme range of variance in interpretation, as the piece is so old, and little of Pachelbel's music survived. Interpretation differences for something like Paganni's work would be more subtle.

But what's most interesting, is how different interpretations can be of a composers works while they are still alive. Recordings exist of Gershwin's Rhapsody in Blue, featuring Gershwin himself, in 1924 and 1927.

https://youtu.be/VGvuUOtHGkk

https://youtu.be/PA3ryT2HJK4

There are some quite noticeable differences in interpretation, even between these two performances.

https://en.wikipedia.org/wiki/E_Street_Band

This is a classic example. It looks hilarious on stage. The timeline at the bottom now suggests five for tours.

To take an example where it's a fair fight, and the computer doesn't win by virtue of having more input bandwidth: RL models applied to narrow physical tasks.

These will often exceed human ability after just a couple of hundred attempts - Cart Pole would be an example.

Narrowly targeting very specific species of mosquito shouldn't cause major issues. Certainly compared to existing approaches to deal with them during severe disease outbreaks, which is basically to carpet bomb entire cities with insecticide.

Yes, exactly.

A USB mic is a normal mic that has a hidden built-in adc.

It sends you the digital signal over the USB cable.

Yes. Or I believe you can use a USB mic that is compatible with the pi.