gjstein

DC local. Total outage this afternoon. SOS on my phone for a few hours. Unable to log into my account.

What sort of laptop/desktop do you have? I regularly use my AirPods with my Steam Deck and they work great!

Ursula Le Guin in her preface to The Left Hand of Darkness [1], describes Science Fiction as "descriptive." She invents "elaborately circumstantial lies" as a means of describing what she sees as some truth in our being. The full quote:

I’m merely observing, in the peculiar, devious, and thought-experimental manner proper to science fiction, that if you look at us at certain odd times of day in certain weathers, we already are. I am not predicting, or prescribing. I am describing. I am describing certain aspects of psychological reality in the novelist’s way, which is by inventing elaborately circumstantial lies.

This is how I think about both science fiction and fantasy. Elements of world building are different, even within each sub-genre, but this element of incorporating elements that are inconsistent with our world to tell stories is common to both. It's also why the term "speculative fiction" persists: a category that subsumes sci-fi and fantasy.

[1] Read that full preface here: https://www.penguinrandomhouse.ca/books/342990/the-left-hand...

Feynman is no random celebrity. In addition to be a renowned physicist, his famous "Feynman Lectures" and his thoughts on pedagogy are similarly legendary.

So far, no one has mentioned "Bug Algorithms", which have a similar structure of (1) walk in the direction of the goal, (2) walk around obstacles as they are encountered, (3) leave the obstacle to proceed when some condition is met. They are very simple to implement (though not optimal) and there are a number of variants to play around with. Howie Choset has some good lecture slides that describe them [1]. However, as some others have mentioned, something like Jump Point Search [2] is likely a better option given the described scenario.

[1] https://www.cs.cmu.edu/~motionplanning/lecture/Chap2-Bug-Alg... [2] https://en.wikipedia.org/wiki/Jump_point_search

The idea that this is a drop in replacement for numpy (e.g., `import cupy as np`) is quite nice, though I've gotten similar benefit out of using `pytorch` for this purpose. It's a very popular and well-supported library with a syntax that's similar to numpy.

However, the AMD-GPU compatibility for CuPy is quite an attractive feature.

Similar: XKCD conducted a color survey back in 2010 [1]. The results are detailed and the writeup quite interesting.

[1] https://blog.xkcd.com/2010/05/03/color-survey-results/

I'd also like to mention the classic book "Reinforcement Learning" by Sutton & Barto, which goes into some relevant mathematical aspects for choosing the "best" among a set of options. They have a full link of the PDF for free on their website [1]. Chapter 2 on "Multi-Armed Bandits" is where to start.

[1] http://incompleteideas.net/book/the-book-2nd.html

Doesn't Slack violate that constraint? Slack allows (or at least seems to) using iOS system emoji to respond to messages.

This is true enough, though remember that material properties change dramatically when you start moving through wavelengths by orders of magnitude. Silicon is transparent in the mid-infrared, which is what makes silicon photonics possible [1]

[1] https://en.wikipedia.org/wiki/Silicon_photonics

This is a cynical take on what I think is a meaningful benefit that the above commenter is talking about. If you use ChatGPT to learn concepts and approach it with the caveat that you should make sure you understand what you are trying to learn (something that helps to verify that what is being presented to you by the system is "accurate"), it's a fantastic resource. That isn't to say it's true for learning all things, but I'm generally impressed and optimistic for ChatGPT to be as valuable a teaching/learning resource as any other I might find on the web.

I have some mixed feelings about this. I used to do this ~10 years ago when I was a distance runner, but now exercising so hard will throw my body off in other ways and I don't feel good having pushed so hard. Perhaps this is exactly the point you're making, but I don't target zone 5 for long these days. Deep sleep, sure, but at what cost to the rest of my body?

Agreed! I've never seen something like this before, but it seems that all articles in this (nascent) journal have the same sort of information. Does anyone know if this is standard across journals for research in this space?

It seems like using np.random.choice is indeed a slow way to get a grid in which 5% of the values are 1. I would recommend using np.random.rand(size) >= mutations:

  > python3 -m timeit 'import numpy as np; mutations=0.05; rows=10; columns=10; np.random.choice([0,1], p=[(1-mutations), mutations],size=(rows,columns))'
  50000 loops, best of 5: 8.88 usec per loop`

  > python3 -m timeit 'import numpy as np; mutations=0.05; rows=10; columns=10; np.random.rand(rows,columns) <= mutations'
  200000 loops, best of 5: 1.06 usec per loop

Excited for this, but I think with all this conversation about the role an AI assistant should play in work and development, this line feels incomplete to me:

Think of Claude as a friendly, enthusiastic colleague or personal assistant who can be instructed in natural language to help you with many tasks.

It omits that the colleague may have outdated knowledge or not understand whatever problem you give it. The colleague's "enthusiasm" should be tempered with oversight so that the outputs they produce are not directly used without scrutiny. It seems that most people using these tools increasingly understand this, but to leave it off the website seems ... sloppy at this point.

Edit: upon logging in, I'm greeted by a warning "It may occasionally generate incorrect or misleading information, or produce offensive or biased content."

Rewatching Person of Interest now. Amazing how relevant it is; aged incredibly well.

What sort of Docker support will this have? I have a very shaky setup right now to run Unity headless that requires OpenGL and VirtualGL. It feels like that pipeline won’t work forever and am looking for alternatives.

That really is a somewhat incredible statistic to me, not because it's surprising given my own empirical evidence, but because of all the things one does in/with a car for years at a time. My car has been driven across the country multiple times, survived fairly extreme cold and heat multiple times a year, and still it shows no signs of stopping. It's always been amazing to me how long cars last considering what most people put them through.

This gets shared on here from time to time and while I love it (and Tufte), it really falls short on mobile devices. When the page is sufficiently narrow, revealing sidebar items is done via small clickable elements: if the user has never seen those before, I expect they will miss out on them. Note that this includes both footnotes and also figure captions, which are often not "optional content" as are many of the sidenotes. My own blog / lab website [1] has a similar style, yet instead makes all sidebar elements visible even when the page is collapsed; fell free to emulate!

[1] https://cs.gmu.edu/~gjstein/2018/12/toward-real-world-alphaz...

Just as ChatGPT seems pretty capable at summarizing text, an AI with "unlimited memory" could potentially answer analytical questions about larger datasets and non-linear data (in the sense that prose is read from start-to-finish).

The OP is most excited about this ability to remember to create more structured longford outputs with internal consistency (e.g., asking questions about a fantasy universe that respects the characters that exist elsewhere in the story or universe).

I have a similar background in Electrical Engineering and, while I enjoyed the book, it did not change my perspective. I think the book does a good job of opening up what many programmers (and, of course, others) may see as a "black box" that they interact with on a daily basis. I recommend it to those who do not already have a fairly comprehensive of how a computer works, but if you think you pretty much already understand how transistors become ALUs and have touched Assembly, you may find it a bit boring (as I did).

Would love an answer on this too. It would be even better not just to try using this, but also be able to run it locally, something that has been impossible for GPT-3.

Philips Hue lights. I had wanted them for a little while for the novelty of it, but upon buying a couple my wife and I converted nearly all of our small apartment (except the bathrooms) to use their color-changing bulbs. Being able to change the light color and temperature has been surprisingly nice and has made our home much "cozier" under quarantine. We change the lights to dim reds and blues during movie nights or when we exercise or do yoga. It was a somewhat expensive shift, though I'd happily do it again. Being able to use Siri via our phones to control the lights by voice is also convenient as we head off to bed (as the light switches are across the room).

[Others have also said AirPods Pro. Spectacular little earbuds.]

I'd actually like to pose a similar question: is there a "Bob Ross"-like series for drawing with a tablet? My wife and I have had some fun in quarantine following a few Bob Ross videos using our iPads but the sort of techniques you can get with a physical brush aren't really available in Photoshop / Affinity Designer (or if they are, we don't understand them yet).

The one killer feature I'm still missing from Blender is real-time simulation. I absolutely adore using Blender for hobby and art projects, but as a full-time robotics and computer vision researcher, Blender just still isn't quite there when it comes to being able to generate images in "real time" (even with Eevee) for robotics sim applications, and I'm still using hacked-together tools in game engines like Unity to achieve this and render images. Would really be great if the Blender Game Engine effort could continue as a part of the main branch, but I understand the need of a software to limit it's scope.

Never hurts to dream...

The "search" here refers to the idea that, in principle, you could search the entire space of possible hands of cards and exhaustively predict the optimal action by imagining every hand. However, like in the game of Go, this is computationally intractable, so instead they use machine learning to "guide search" towards more promising "moves". In AlphaGo (and here) this learning happened as part of a reinforcement learning pipeline.

There was some discussion of this in the AlphaGo Zero blog post from a while back: https://deepmind.com/blog/article/alphago-zero-starting-scra...

Though I suppose it's possible to switch to Computer Modern via the CSS, the equations seems to be included as PNG's. The font of those would not change simply by modifying the CSS.

Can second both Ann Leckie (her new book "The Raven Tower" is also a great page-turner) and N.K. Jemisin (who did totally blow my mind with her Broken Earth Trilogy).

Ursula K. Le Guin also appears down below; The Left Hand of Darkness was a profound read.

I object to the use of the word "optimal" for a task like object detection; it feels counterproductive to claim that this is the "optimal" way of solving such a broad and complex problem. Great results, but their language needs some tempering.

I was a physicist for a time and I learned physics via numerical simulation: I would find problems I could solve by hand and code them up---solving integrals, derivatives, systems of equations all numerically and comparing the results. Only a handful of physics problems have closed-form solutions, and being able to turn an interesting problem into code and "play around with it" was enormous fun for me and helped me build intuition as well. This advice strongly depends on your mathematics background, but with some basic calculus you can already start playing around.