6 ms·

Using Artificial Intelligence to Augment Human Intelligence

254 points by runesoerensen 9y ago

partycoder 9y ago
You are already being augmented each time you use a search engine, a calculator, an encyclopedia, a GPS, etc... an abacus even. e.g: The pathfinding algorithm on a GPS is technically narrow AI. And if we stress this even further, a conditional statement could be AI. Your washing machine runs on AI (fuzzy logic). Same with the term "technology". A spoon is technology, a chair is technology, a roof is technology. The term you would be usually looking for is high technology, which is state-of-the-art technology. Having said this, AI is an ambiguous term and usually leads to clickbait, large disambiguations and waste of time.
olalonde 9y ago
The quality of this journal is mind boggling.
- hmexx 9y ago
  I opened it up after reading your comment expecting to be disappointed due to raised expectations. I wasn’t!!
- bpicolo 9y ago
  Programmed frontend demos here and in places like news sites can be used to great effect. I how one could increase accessibility of these sorts of widgets to non-programmers? Deeplearning is a bit out of scope for that, but interesting graphics have a bunch of other uses.
  - 52-6F-62 9y ago
    Processing.js[0] might be a good start? Or maybe Paper.js[1]. Possibly including Tween.js, or something else[2]. It wouldn't be a memory-conscious thing at the outset (without investing much more work), but if there was a more convenient layer overtop of one of those, or a combination, as a drop-in widget that can be customized by some simple UI or much more rudimentary scripting as something possibly embeddable or native to a blog or news platform might cut it. Sounds like a good idea. IIRC D3 came out of the NYT? Of course that was rather programmer-centric, but similar root need seems to have inspired it. [0] http://processingjs.org/ http://processingjs.org/ [1] http://paperjs.org/ http://paperjs.org/ [2] https://www.createjs.com https://www.createjs.com
    - bpicolo 9y ago
      I imagine some sort of excel-like tabular data -> charts with sliders and filters could work pretty well as a starting point. Take a BI tool and trim all of the fat. All the BI tools in existence right now go hard on enterprise targeting, and they have all of the complexities involved in that.
- icc97 9y ago
  Brett Victor has done some lovely stuff allowing people to interact whilst telling a story [0]. Doesn't work so well in mobile, but the Google page suffers in this aspect too. [0]: http://worrydream.com/LadderOfAbstraction/ http://worrydream.com/LadderOfAbstraction/
strawcomb 9y ago
So far the generative models I've seen (VAEs, GANs) really just seem to morph between images.
rememberlenny 9y ago
My work is very relevant here. I'm a graffiti artist and a programmer. I started using machine learning techniques to augment street art with various nature landscapes (drone images, rolling hills, etc). By using simple techniques, such as Style Transfer, I have been able to create some interesting images that are aesthetically pleasing, but unusual. I took the digital results and painted the contents on large canvases. I recently invested in some GPUs and a new deep learning rig. Im planning on exploring what unique results I can make around graffiti LETTERS, which are interesting in themselves. Take a look! https://becominghuman.ai/digital-processes-inspiring-analog-paintings-a358eb7801a0 https://becominghuman.ai/digital-processes-inspiring-analog-...
- kowdermeister 9y ago
  They look great, but I suggest you to showcase your work on a platform designed for artists, on Medium is barely enjoyable: https://imgur.com/a/48YYr https://imgur.com/a/48YYr Sites like cargocollective are a good start.
- abcd_f 9y ago
  Dude. You can't be serious about bastardizing stunningly detailed 9th Wave by Aivazovsky and calling it "Process for turning classic paintings into beautiful color muses." That's like making a confetti from the Mona Lisa with a hole puncher. https://cdn-images-1.medium.com/max/2000/1*sph_3TactpBHo_mZ-65g2w.png https://cdn-images-1.medium.com/max/2000/1*sph_3TactpBHo_mZ-... Original for reference - https://upload.wikimedia.org/wikipedia/commons/4/4a/Hovhannes_Aivazovsky_-_The_Ninth_Wave_-_Google_Art_Project.jpg https://upload.wikimedia.org/wikipedia/commons/4/4a/Hovhanne...
  - sctb 9y ago
    This is an uncivil, low-effort dismissal. Hacker News doesn't need any of those, so could you please not post them? https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
    - abcd_f 9y ago
h3ctic 9y ago
Isn't that Elon Musks approach with Neuralink? I'm referring to the article: https://waitbutwhy.com/2017/04/neuralink.html https://waitbutwhy.com/2017/04/neuralink.html
windmaple 9y ago
Great article but some many typos ... sigh
euyyn 9y ago
It's super interesting that by removing serif from a sans-serif font you get Comic Sans!

254 points by runesoerensen 9y ago

You are already being augmented each time you use a search engine, a calculator, an encyclopedia, a GPS, etc... an abacus even. e.g: The pathfinding algorithm on a GPS is technically narrow AI. And if we stress this even further, a conditional statement could be AI. Your washing machine runs on AI (fuzzy logic). Same with the term "technology". A spoon is technology, a chair is technology, a roof is technology. The term you would be usually looking for is high technology, which is state-of-the-art technology. Having said this, AI is an ambiguous term and usually leads to clickbait, large disambiguations and waste of time.

olalonde 9y ago

The quality of this journal is mind boggling.

hmexx 9y ago
I opened it up after reading your comment expecting to be disappointed due to raised expectations. I wasn’t!!
bpicolo 9y ago
Programmed frontend demos here and in places like news sites can be used to great effect. I how one could increase accessibility of these sorts of widgets to non-programmers? Deeplearning is a bit out of scope for that, but interesting graphics have a bunch of other uses.
- 52-6F-62 9y ago
  Processing.js[0] might be a good start? Or maybe Paper.js[1]. Possibly including Tween.js, or something else[2]. It wouldn't be a memory-conscious thing at the outset (without investing much more work), but if there was a more convenient layer overtop of one of those, or a combination, as a drop-in widget that can be customized by some simple UI or much more rudimentary scripting as something possibly embeddable or native to a blog or news platform might cut it. Sounds like a good idea. IIRC D3 came out of the NYT? Of course that was rather programmer-centric, but similar root need seems to have inspired it. [0] http://processingjs.org/ http://processingjs.org/ [1] http://paperjs.org/ http://paperjs.org/ [2] https://www.createjs.com https://www.createjs.com
  - bpicolo 9y ago
    I imagine some sort of excel-like tabular data -> charts with sliders and filters could work pretty well as a starting point. Take a BI tool and trim all of the fat. All the BI tools in existence right now go hard on enterprise targeting, and they have all of the complexities involved in that.
icc97 9y ago
Brett Victor has done some lovely stuff allowing people to interact whilst telling a story [0]. Doesn't work so well in mobile, but the Google page suffers in this aspect too. [0]: http://worrydream.com/LadderOfAbstraction/ http://worrydream.com/LadderOfAbstraction/

strawcomb 9y ago

So far the generative models I've seen (VAEs, GANs) really just seem to morph between images.

rememberlenny 9y ago

My work is very relevant here. I'm a graffiti artist and a programmer. I started using machine learning techniques to augment street art with various nature landscapes (drone images, rolling hills, etc). By using simple techniques, such as Style Transfer, I have been able to create some interesting images that are aesthetically pleasing, but unusual. I took the digital results and painted the contents on large canvases. I recently invested in some GPUs and a new deep learning rig. Im planning on exploring what unique results I can make around graffiti LETTERS, which are interesting in themselves. Take a look! https://becominghuman.ai/digital-processes-inspiring-analog-paintings-a358eb7801a0 https://becominghuman.ai/digital-processes-inspiring-analog-...

kowdermeister 9y ago
They look great, but I suggest you to showcase your work on a platform designed for artists, on Medium is barely enjoyable: https://imgur.com/a/48YYr https://imgur.com/a/48YYr Sites like cargocollective are a good start.
abcd_f 9y ago
Dude. You can't be serious about bastardizing stunningly detailed 9th Wave by Aivazovsky and calling it "Process for turning classic paintings into beautiful color muses." That's like making a confetti from the Mona Lisa with a hole puncher. https://cdn-images-1.medium.com/max/2000/1*sph_3TactpBHo_mZ-65g2w.png https://cdn-images-1.medium.com/max/2000/1*sph_3TactpBHo_mZ-... Original for reference - https://upload.wikimedia.org/wikipedia/commons/4/4a/Hovhannes_Aivazovsky_-_The_Ninth_Wave_-_Google_Art_Project.jpg https://upload.wikimedia.org/wikipedia/commons/4/4a/Hovhanne...
- sctb 9y ago
  This is an uncivil, low-effort dismissal. Hacker News doesn't need any of those, so could you please not post them? https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html
  - abcd_f 9y ago

h3ctic 9y ago

Isn't that Elon Musks approach with Neuralink? I'm referring to the article: https://waitbutwhy.com/2017/04/neuralink.html https://waitbutwhy.com/2017/04/neuralink.html

windmaple 9y ago

Great article but some many typos ... sigh

euyyn 9y ago

It's super interesting that by removing serif from a sans-serif font you get Comic Sans!