91–97 of 97 posts
This is impressive, and certainly technology like this will probably find lots of very good uses, but I can't help imagining(!) dystopian societies where "thoughtcrime" can actually be monitored and "made actionable". A world in which "you have nothing to hide" turns into "you have nothing you can hide" is immensely disturbing.
I bet they will soon replace the search bar by just a button.
Earlier quoted context omitted.
Harvard updated their Buzzfeedesque post on this topic [ https://www.health.harvard.edu/mind-and-mood/12-ways-to-keep... ] which spans from crossword puzzles to avoiding alcohol. But to keep it simple: read, do puzzles, stay active, do not drink or do drugs heavily. —— Everybody already has a genetic makeup that limits their health—especially the brain—to a preset level. Some people suffer from infantile amnesia—basi…
I don't really think that you can form memories only from when you are 2 years old. I have memories of when they removed the cast from my broken arm and I was probably 1 year and some months.
Have you spoke about it at great lengths when older? Is it often brought up about how they took the cast off? If so, you’re looking at the idea of explicit (declarative) memory—the storing of facts and events. If over the years you are constantly reminded of this event, it is more than likely that you’re trying to generate a (possible semi-accurate) image or memory of that particular event.
As far as your second comment goes, mother tongue and languages are different all together. When you learn as a yongue age, your cerebral cortex is still developing and your mind is absorbing everything like a sponge. As you learn these words, continue to use them, that language and those procedures are going to get stored in your long-term memory and then you’ll be able to recite all of these words and form sentences without thinking twice—something we would call implicit memory which helps us with procedure memories (riding a bike, writing, etc.).
I guess it makes sense that HN is skeptical about results more closely related to the community's expertise but it's a pretty crazy leap from being able to reconstruct something vaguely similar to what a person is currently seeing vs. trying to read someone's mind. I still have seen very little evidence that such a thing could at all be possible with FMRI. The spatial and temporal resolution are just far too low. The more everyone is impressed with results like this the more we delay the hard work of developing tools that actually have a shot at doing something like that.
Earlier quoted context omitted.
Do people actually see images of things in their head? I can close my eyes and think of something, but I won't perceive anything more than a linear description of it. At best, I can use these visual descriptions to draw an outline in my head, but it instantly fades out if I don't continuously re-draw it. Kind of like this: https://i.imgur.com/0zuPIPV.gifv
Interesting. What are your dreams like? Do they look like abstract pictures (lanes? shapes?) or have no picture at all?
Earlier quoted context omitted.
- Can you imagine the color of the floor, walls and ceiling at work? - Can you imagine the color of the eyes and hairs of your coworkers? - Can you imagine the home screen of your phone and see how many application icons can fit in a row and column? - Can you imagine Google's logo and see the color of each letter? - Can you imagine your car manufacturer's logo and draw it?
1) I can do that easily, and even can imagine the paint texture for the walls. 2) Hair for all of them, but I can only picture the eyes of the people who had unusual colors (blue, gray, green). Maybe I never paid much attention if they had normal brown eyes. 3) Row yes, but I wasn't sure about the column (4 or 5). 4) Yes. 5) easiest task on the list.
Earlier quoted context omitted.
1) I can do that easily, and even can imagine the paint texture for the walls. 2) Hair for all of them, but I can only picture the eyes of the people who had unusual colors (blue, gray, green). Maybe I never paid much attention if they had normal brown eyes. 3) Row yes, but I wasn't sure about the column (4 or 5). 4) Yes. 5) easiest task on the list.
I'm pretty sure less than 1% of people know the colors of Google's letters.
Earlier quoted context omitted.
Very interesting. For my curiosity I did some searching on VGG vs ResNet. * ResNets were better at feature extraction for image clustering [0] * One of the trade-offs of ResNets seems to be their relative complexity to VGG [1] * It surprises me that ResNets aren't significantly faster to train based on the large reduction of FLOPs (from the ResNet paper VGG-19 had 19.6 billion FLOPs vs ResNet-34 with 3.6 billion FLOP…
You might enjoy this recent article on Distill: https://distill.pub/2018/differentiable-parameterizations/#s... They discuss VGG vs non-VGG architectures in the context of style transfer in Section 2, which was interesting to me.