HN Theater @HNTheaterMonth

The best talks and videos of Hacker News.

Hacker News Comments on
Artificial intelligence, video games and the mysteries of the mind | Raia Hadsell | TEDxExeterSalon

TEDx Talks · Youtube · 10 HN points · 2 HN comments
HN Theater has aggregated all Hacker News stories and comments that mention TEDx Talks's video "Artificial intelligence, video games and the mysteries of the mind | Raia Hadsell | TEDxExeterSalon".
Youtube Summary
Artificial intelligence could be the powerful tool we need to solve some of the biggest problems facing our world, argues Raia Hadsell. In this talk, she offers an insight into how she and her colleagues are developing robots with the capacity to learn. Their superhuman ability to play video games is just the start.

Raia is a senior research scientist on the Deep Learning team at DeepMind, with a particular focus on solving robotics and navigation using deep neural networks.

---
TEDxExeterSalon: From driverless cars to diagnosis of medical imaging, artificial intelligence is being heralded as the next industrial revolution. But how does AI relate to us in all our glorious complex humanity? Our first TEDxExeterSalon explored the ways in which we’ll interact with algorithmic intelligence in the near future.

TEDxExeter: Now in our 7th year, TEDxExeter is Exeter’s Ideas Festival with global reach, licensed by TED and organised by volunteers who are passionate about spreading great ideas in our community.

More information: http://www.tedxexeter.com
Production: http://www.youtube.com/familygamertv
Filming: http://firstsightmedia.co.uk/ Raia Hadsell is a research scientist on the Deep Learning team at DeepMind. She moved to London to join DeepMind in early 2014, feeling that her fundamental research interests in robotics, neural networks, and real world learning systems were well-aligned with the agenda of Demis, Shane, Koray, and other members of the original team. Raia’s research at DeepMind focuses on a number of fundamental challenges in AGI, including continual and transfer learning, deep reinforcement learning, and neural models of navigation. Raia came to AI research obliquely. After an undergraduate degree in religion and philosophy from Reed College, she veered off-course (on-course?) and became a computer scientist. Raia’s PhD with Yann LeCun, at NYU, focused on machine learning using Siamese neural nets (often called a ‘triplet loss’ today) and on deep learning for mobile robots in the wild. Her thesis, ‘Learning Long-range vision for offroad robots’, was awarded the Outstanding Dissertation award in 2009. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx
HN Theater Rankings

Hacker News Stories and Comments

All the comments and stories posted to Hacker News that reference this video.
# First Topic

The idea would be to be able to write something like this:

    function yt --pure
        youtube-dl $argv
    end
The parser should stop interpreting arguments after the yt function call and pass the args as strings to the function as before.

Then such a call would be possible: yt https://www.youtube.com/watch?v=mqma6GpM7vM

# Second Topic

I wrote a little utility [1] to quicker navigate on the file system. Right now I have two commands:

    m  [folder]  // goto highest ranked folder
    mm [folder]  // list results for folder-keyword
I would like to be able to write 'm foobar' and then press tab to get a vertical completion-list. When the user selects a completion-list-item I would like to pass the result-path to the cd command. Some time ago I tried to write such a completion, but failed – that's why I still use a separate mm-function to list search results.

[1]: https://github.com/thibran/maybe

I think JSON might be a good choice here because it would simplify the interaction with the completion program for advanced completions. Passing all those flags – mixed with some other shell function calls – to the 'complete' program is a bit cumbersome. My dream would be to have something like 'complete -c maybe --json [maybe-flag-to-generate-json-output]' and handle everything inside the application.

    {
      "fish_completion_version": 1,
      "header": "optional, completion header",
      "orientation": "vertical",
      "on_single_result": "optional, choose-item or display-item",
      "arguments": [
        {
          "item",
          "description",
          "optional return value, if empty return item"
        }],
      "item_handler": "optional, name of fish function to call on-item-chosen"
    }
I know JSON is not something you want to assemble in a script, but all programming languages have support for it and it's easy to use.
faho
>Then such a call would be possible: yt https://www.youtube.com/watch?v=mqma6GpM7vM

Okay, so I did understand what you are saying.

Like I said, that's a rather large layering violation, since the function has nothing to do with the expansions happening on its arguments. They are expanded before they ever reach the function. It's essentially allowing functions to modify the syntax around them.

How exactly would this treat e.g.

    yt $URL/somefile
Would the variable still be expanded? Would an asterisk glob be? If I _did_ want it to expand something even though the function is declared "pure", how would that work?

Now I couldn't be sure how my arguments would be interpreted, or I'd have to go checking for everything I invoke.

I use youtube-dl myself quite often, and what I do is just type

   youtube-dl '
then press ctrl-v (which is a keybinding that uses xsel or pbpaste to get the clipboard contents and insert them into the commandline) or ctrl-shift-v (which is my terminal's paste binding). Then I press `'` again and can download the video.

Additionally, it's looking like 3.0 will remove the `?` glob, which is the thing that is most annoying with URLs (though "&" is also a thing).

> I would like to be able to write 'm foobar' and then press tab to get a vertical completion-list. When the user selects a completion-list-item I would like to pass the result-path to the cd command.

I would assume the reason you failed is that the argument you want doesn't match the argument the user typed. Because fish's completions undergo filtering (with some fuzzy-matching, so e.g. `f_c_i` matches `fish_config_interactive`). Otherwise every completion script would have to implement that by itself, and consistently to not confuse people.

For completions, that's the right thing to do.

What you want, I'd argue, isn's a completion. What you want is a selection menu. Which could be implemented with fish's completion pager (we have an open issue about making it accessible to the outside), but isn't the same thing.

JSON here is a red herring, since you don't just want to add items (with descriptions), but also control the behavior of the pager ("orientation" and "on_single_result"). That you could do just with arguments to this `select` command (`list_of_items | select --orientation=vertical --on-single=choose`).

thibran
> How exactly would this treat e.g.

    yt $URL/somefile
> Would the variable still be expanded? Would an asterisk glob be?

No. Pure should mean don't do anything from here on, just pass the arguments to the function.

> If I _did_ want it to expand something even though the function is declared "pure", how would that work?

If expanding should be possible in a pure context, then maybe by using a known syntax like:

    yt ($URL)/somefile
Eshell expands Lisp code when using (lisp-code-here) which is kind of elegant. So when in a pure context, you would have to switch to expansion-mode by using curly braces.
Nov 30, 2017 · 10 points, 0 comments · submitted by giacaglia
HN Theater is an independent project and is not operated by Y Combinator or any of the video hosting platforms linked to on this site.
~ yaj@
;laksdfhjdhksalkfj more things
yahnd.com ~ Privacy Policy ~
Lorem ipsum dolor sit amet, consectetur adipisicing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.