Wish there was a benchmark for ML safety? Allow us to AILuminate you...
Very much a 1.0 – but it's a solid start
AI + ML
5 Dec 2024 | 8
Brainwaves rock! Scientists decode Pink Floyd tune straight from the noggin
A group of scientists say they are the first to reconstruct a recognizable song from data collected directly from the brain by monitoring electrical activity and modeling the resultant patterns with regression-based decoding models. That song is Pink Floyd's "Another Brick in the Wall, Pt. 1" and the question on the mind of …
-
-
Thursday 17th August 2023 14:43 GMT that one in the corner
Re: Sounds like...
Oooh, thank you. Clearly I'm not as much of a Pinkie as I thought, as I had no idea that existed :-(
The track list alone is intriguing. Even if it turns out to be (to these ears) ghastly, it's got to be worth listening to at least once, if only to find out what the more than seven dwarves are doing.
And, yes, if the tech existed back then then they'd have used it.
-
Thursday 17th August 2023 19:07 GMT TRT
Re: Sounds like...
I noticed it in the CD section of Brighton library whilst browsing one day, way back in 1987. Saw the names... borrowed the disk... was enchanted by the utter lunacy, humour and charm of it. Over the years, whenever I remember, I try to get the DVD of the film. Haven't succeeded yet.
-
-
-
-
-
-
Thursday 17th August 2023 17:43 GMT Steve Davies 3
Re: Visceral response
The first time I heard 'Eclipse' as it was called then at Brighton Dome, 20th Jan 1972 was out of this world. That was the premier of what later became DSOTM. It was a very early version of the iconic piece but all the main parts were there.
Then they played Echoes at the opening of the second half. Even then, all it took was one note for it to be recognised by the crowd.
Yes, I'm getting old but the sheer technique that they brought to their music makes most of the modern [cough][cough] stars look like 3yr olds with their toy piano. Having to Lip-sync at Glasto? get outahere.
-
-
Thursday 17th August 2023 14:16 GMT NoneSuch
This Happy Thought Experiment
has dark undertones. With a stable technology how long do you think it will be before road side brain scans will be implemented by the Home Office?
"Oh, we know you're not a terrorist, sir. Just stick you noggin in this helmet to confirm."
Every other technology ever invented has ended up in the hands of the irresponsible or dictators, so why would this be any different?
-
-
Saturday 19th August 2023 03:21 GMT M.V. Lipvig
Re: This Happy Thought Experiment
Not needed, eventually. All they will need to do is figure out an induction method, then just being in a room will be enough to do a full brain scan. You may not even know it happened. But it is disturbing that they have successfully recreated someone's thoughts via a scan, as the hard part is always making it work the first time. Once you make it work, the rest is just refinement. You know that once large government spy agencies get wind of this that they will take the lead on development. We may only have 10 more years of privacy between our own ears left.
This icon has never been more relevant.
-
-
-
Thursday 17th August 2023 14:45 GMT Anonymous Coward
This appears to be remarkably similar to the study re-constructing doctor who scenes from EEG recordings and (as far as I can tell) suffers from the same problem with its methodology.
i.e. The recorded data was from people listening to a single Pink Floyd song. An algorythm was used to associate brain activity as recorded by the EEG with sounds from parts of the song. So far so good.
However, it's unfortunate that they appear to be using the same data to test the pattern matching as was used to create it.
Unsurprisingly, when you use the *same* EEG data as input to make the sounds identified with activity in each identified area of the brain to create your pattern matching algorythmn as you use to test it, you end up with something that sounds vaguely like the input.
Had they recorded new EEGs of (say) Pink Floyd live in Pompei, and not used it in the original dataset and then used that EEG recording to generate audio using the model created from the first dataset, the study would be much more interesting. (and weird...)
-
Saturday 19th August 2023 03:26 GMT M.V. Lipvig
I don't think you thought this all the way through. They will use the same data to see if they can reliably repeat the experiment. Once they are sure, they will move on to other songs. Let them test 2-3 songs when they know the song and repeat it, then they will start trying blind - let the person listen to a random song, see if they can retrieve the correct song.
This is the beginning of a very, very dangerous branch of technology.
-
Thursday 17th August 2023 17:44 GMT Mike 137
Richness and complexity a matter of opinion?
'what the boffins described as a "rich and complex auditory stimulus."'
In terms of actual complexity (which can be numerically defined in terms of rhythmic, tonal and harmonic variation), I wonder (rhetorically) how 'another brick' compares with, say, a flamenco guitar solo or a choral motet like Tallis's 40 part Spem in Alium. It seems that only commercial music from the last half century or so out of the 600 years and vast array of genres on record registers with these folks as existing.
-
Thursday 17th August 2023 20:22 GMT steelpillow
How far can this go?
All our brains are similar at the coarsest scale. But even in the slightly less gross details we differ. Musicians, artists, mathematicians, dyslexics, epileptics and so on have some surprisingly big differences. Down in the nitty-gritty, every brain is different: no two tangles of neurones are exactly alike and some are quite radically different. So there has to be a bottom level of detail that these telepathic chatbots must bottom out at, thank God. Below that, they become increasingly unreliable and the details soon become drowned in the randomness. Reconstructing Pink Floyd riffs through the bottom of a bucket is probably not too far from the theoretical limit. Wonder if they'll try and drag out my memories of the Animals tour London gig, back ca. 1977...
-
Thursday 17th August 2023 20:25 GMT DS999
Re: How far can this go?
Well all this is doing is essentially tapping into the line between the ear and the rest of the brain, since the music is being played for people while their brainwaves are monitored. Being able to reconstruct audio that's recalled from a memory would be orders of magnitude more difficult - quite likely impossible.
-
-
Friday 18th August 2023 23:19 GMT PRR
Another Brick... has many variants. Just casually I have accumulated a Rap version and a banjo pickin'-on. And a Brick from Pink Turtle (cool French swing).
Hotel California may have more parodies.
YouTube is good for discovering odd mash-ups.
Biting the hand that feeds IT
