This really is just insanely cool. What a genius idea — take a machine-learning algorithm that can produce images from text, train it on images of spectrograms, let it interpolate between them, and convert the spectrograms back to audio. I could honestly listen to this for quite a while.
I’ve been thinking a lot about confabulation lately, because large language models do it too. And in particular, last night I read an interesting study (https://arxiv.org/abs/2305.04388) that shows that LLMs, asked to explain their decisions, come up with a plausible story that doesn’t necessarily reflect their actual decision process. This is very much something humans do; for examples, see the post quoted below, particularly the section on choice blindness.
We have so far explored confabulation in patients with brain damage. Do neurotypical, everyday people produce “honest lies”?
We confabulate all the time.. We just don’t realize that we are.
In Telling More Than We Can Know: Verbal Reports on Mental Processes, Nisbett & Wilson (1977) review hundreds of studies, across dozens of disciplines. Their evidence admits a theme: people’s attempts to explain their behavior is almost always unhelpful in identifying the important factors influencing their decisions. Let me briefly review four example findings.
This is one of the very few things I’ve ever read that made me feel like I understand the Standard Model just a little bit better.
Although in full honesty he really lost me when he said, ‘…if I gave you the Standard Model, you wouldn’t come back to me and tell me about the existence of a giraffe.’
(I’ve only read the transcript, no idea whether it’s better or worse in it’s original audio form)
https://www.quantamagazine.org/are-there-reasons-to-believe-in-a-multiverse-20230517/
^ check out the 15 seconds following 5:38 (direct link to that start time) for some really useful mnemonics for the polyrhythms for 2:3 (‘Nice cup of tea’), 3:4 (‘Pass the goddamn butter’), and 4:5 (‘I’m looking for a home to buy’).
Some others can be found at these links (although you really have to hear them, or feel them out against the polyrhythm, to get them right):
[EDIT]
A couple more I figured out by using https://4four.io/ploop:
3:5, emphasis on the 5 — “HEAT UP the CHEESE PIZza PLEASE”
2:5, emphasis on the 5 — “I’VE BEEN CEL-e-BRAT-ING”
3:4 emphasis on the 3, as mentioned above, is “PASS the GODdamn BUTter”. With emphasis on the 4, you can use “I’M LOOKing FOR a CAKE”. It’s pretty interesting to put on a 3:4 rhythm (using eg this excellent tool) and switch between those two.
Wow. If you know you’re going to read this fascinating essay, which I highly recommend, avoid reading my spoilers below.
Galton, Ehrlich, Buck – by Scott Alexander
…still here? Today I learned:
In 1975, India had a worse-than-usual economic crisis and declared martial law. They asked the World Bank for help. The World Bank, led by Robert McNamara, made support conditional on an increase in sterilizations.
…
In the end about eight million people were sterilized over the course of two years. No one will ever know how many were “voluntary” by standards that we would be comfortable with, but plausibly well below half.
The West didn’t just tolerate this process, they supported it and cheered it on. The Ford and Rockefeller Foundations provided much of the funding. Western media ranged from supportive to concerned-for-the-wrong-reasons…
Maybe now you’ll go read it? I was pretty shocked to learn about this. In the mid-70s! Despite the enormous post-WWII backlash against forcible sterilization for eugenic purposes!
Wow.
In 2010, Ratcliff started working with brewer’s yeast, the single-celled fungus we use to make bread and beer. He repeatedly grew the yeast in liquid-filled tubes, shook them, and then used the cells that sank fastest to start new cultures. By favoring cells that stick together and settle faster, this simple procedure radically changed the yeast within just 60 days. Now whenever a cell divided in two, the new cells didn’t drift apart as they normally would; instead, they remained attached, creating beautiful, branching snowflakes that comprised dozens of cells. The yeast had evolved multicellularity in just two months.
The evolution of MREs is surprisingly fascinating.
…the Metabolic Kitchen, the lone culinary research and development center for the world’s most technically advanced military, is what I’m after. This is the place that developed pizza that can survive for three years without refrigeration; a strangely convincing approximation of key lime pie that can be sucked through a tube at 9Gs, or nine times the force of gravity; dehydrated chili that can endure an Antarctic expedition; and ice cream that can go into outer space. I’m here to find out how.
https://www.atlasobscura.com/articles/natick-lab-military-food-science
In December 2018, Thierry Brunfaut and Tom Greenwood published an article in Fast Company where they coined a new word: Blanding.
“The worst branding trend (…) is the one you probably never noticed. I call it blanding. The main offenders are in tech, where a new army of clones wears a uniform of brand camouflage. The formula is sort of a brand paint-by-numbers. Start with a made-up-word name. Put it in a sans-serif typeface. Make it clean and readable, with just the right amount of white space. Use a direct tone of voice. Nope, no need for a logo. Maybe throw in some cheerful illustrations. Just don’t forget the vibrant colors. Bonus points for purple and turquoise. Blah blah blah.”
I got curious about the current state of research on criminal deterrence. Caveat: I don’t know whether this summary of the research is broadly shared by those in the field or whether it’s controversial, nor am I clear on the reputation of the NIJ; I just wanted a first approximation.
- The certainty of being caught is a vastly more powerful deterrent than the punishment.
Research shows clearly that the chance of being caught is a vastly more effective deterrent than even draconian punishment.- Sending an individual convicted of a crime to prison isn’t a very effective way to deter crime.
Prisons are good for punishing criminals and keeping them off the street, but prison sentences (particularly long sentences) are unlikely to deter future crime. Prisons actually may have the opposite effect: Persons who are incarcerated learn more effective crime strategies from each other, and time spent in prison may desensitize many to the threat of future imprisonment.
- Police deter crime by increasing the perception that criminals will be caught and punished.
The police deter crime when they do things that strengthen a criminal’s perception of the certainty of being caught. Strategies that use the police as “sentinels,” such as hot spots policing, are particularly effective. A criminal’s behavior is more likely to be influenced by seeing a police officer with handcuffs and a radio than by a new law increasing penalties.- Increasing the severity of punishment does little to deter crime.
Laws and policies designed to deter crime by focusing mainly on increasing the severity of punishment are ineffective partly because criminals know little about the sanctions for specific crimes.
- There is no proof that the death penalty deters criminals.
According to the National Academy of Sciences, “Research on the deterrent effect of capital punishment is uninformative about whether capital punishment increases, decreases, or has no effect on homicide rates.”
https://nij.ojp.gov/topics/articles/five-things-about-deterrence
Time Magazine, of all media outlets, again publishing an excellent piece on existential risk, this time from MIT researcher Max Tegmark.
https://time.com/6273743/thinking-that-could-doom-us-with-ai/