tag > ML
-
Artificial Intelligence—A Game Changer for Climate Change and the Environment:
http://blogs.ei.columbia.edu/2018/06/05/artificial-intelligence-climate-environment/Artificial intelligence is helping us manage the impacts of climate change and protect the environment in many ways.
-
When Algorithms Create Our Culture:
https://www.worldcrunch.com/tech-science/when-algorithms-create-our-culture/c4s19935Music, books and other intellectual artifacts are increasingly being produced automatically by machines. A new book explores the ways in which artists exploit this new reality.
-
"Procedural Generation For Game Jams: Faster, Better, Weirder" - talk by @mtrc :
-
There is a reason why all ancient traditions talk about ethics, virtues and self-cultivation.
-
HeadOn: Real-time Reenactment of Human Portrait Videos:
http://niessnerlab.org/projects/thies2018headon.html -
The Rise of AI - 1h documentary by @bloomberg
Geoffrey Hinton - as an elderly man, teenager and young child.
"Of course there are some risks in using this tech for bad applications. But unfortunately, it is not possible to stop technology. So the ethical path we have decided is to release it & make people more cautious" - Dev of voice cloning app @LyrebirdAi on ethics & techno-determinism
"Eventually i think we will become the AI's. We will become intelligent machines. You might think of it as creating a new generation, a new kind of people. Humanity has continued to evolve & why would enhanced people or even designed people not be the next step?" - Richard Sutton
Comment: Hearing people talk about "enhanced & designed humans" (in the context of AI, genetics or elsewhere) always brings up very dark associations: https://en.wikipedia.org/wiki/Nazi_eugenics Given the historical context, think it is only right to be extremely careful with such aspirations & wordings.
"It's really hard to predict the future. I think there will be all sorts of things happening we did not except. But there is one thing we can predict: This technology is going to change everything" - Geoffrey Hinton's paradoxical end of documentary statement on A.I
-
Great reply by @nntsn on the question "how can machine learning help with climate-change related issues:
Most environmental issues are fundamentally policy issues that currently are poorly tackled because of socio-economical and political reasons. To me it's unclear whether ML can (or should) fine-tune policies that struggle to even be brought up to legislators / the public.
If so, we should use ML to make a _better_ case for tackling such issues. Maybe we should chat with the orgs involved in lobbying policymakers and provide data analysis tools to strengthen their cases. Would be great to have a joint workshop on this at NIPS/ICML/AAAI/IJCAI.My reply:
Well said & agree. Would love to participate in a new virtual conference (airplane travel wrong symbol) focused on the topic of Flora-Fauna-Human-Computer-Interaction (FFHCI) - Bringing together CS,ML,HCI people with NGOs, activists & policy makers to discuss climate change challenges & solutions.
-
"50,000 Las Vegas Workers Are Ready to Go on Strike Over Fears of Robots Taking Their Jobs": https://gizmodo.com/50-000-las-vegas-workers-are-ready-to-go-on-strike-over-1826474028
Translation: In a city that is rapidly turning into a desert, the owners have weaponised automation against its already struggling citizens & call it "progress".
-
Does AI Art Belong in the Physical, Digital, or Crypto World?
https://medium.com/@pindar.vanarman/does-ai-art-belong-in-the-physical-digital-or-crypto-world-3cb4fe5e01b0Interesting piece on the merger of A.I, Art & Crypto. Question: Is "AI CryptoArt" the most energy intensive art-form ever invented? e.g. very energy hungry GPUs to generate images, perversely energy hungry crypto-nets to trade those images. cc @quasimondo @DrBeef_
Going back to practical questions: In times of (climate) crisis, Art & Design are tremendously powerful communication tools which can reach places politics simply can't. Dadaism or Bauhaus are nice examples. It's time for a global cultural wildfire for sanity.
I see artists as part of the vanguard of society and wonder if new, super energy/resource intensive art-forms are the right symbol for now. I see "EcoAIArt" not only a fun challenge, but as a moral imperative & survival tool
-
Learning to Speak Shrub - "Using molecular codes, plants cry for help, ward off bugs, and save each other": http://nautil.us/issue/59/connections/learning-to-speak-shrub-rp
-
"Skatebot music for the skatebot culture" - by @dadabots:
https://theoutline.com/post/3567/this-skate-punk-album-was-created-by-artificial-inteliigence"A neural network used NOFX’s music to create a ten-track album of frantic skate punk tunes, complete with high-speed drums, palm-muted guitars & sneering vocals"
-
My recent piece "Thoughts on the Generative AI Creative Economy: there will be blood" (https://samim.io/p/2018-04-19-find-it-comedic-how-professional-digital-creative-tools/) generated some discussion on reddit, with people asking "is this concern real or just doom & gloom? Are there concrete examples?" Read my response here:
Beyond doom and gloom, there is hope: The vision/opportunity I sense is phenomenal: Creative AI tools could help raise the global level of literacy, creativity and empathy, by dramatically shortening the time of thought to highly communicative (possibly multimodal) artefacts. Many moons ago I wrote this extensive piece on the sector, from a more hopeful, slightly utopian perspective: https://medium.com/@creativeai/creativeai-9d4b2346faf3
Yet the reality of creative industries today looks very different: Advertising runs the show (see Google etc) - which is really just a synonym for either "spam" or "social engineering" - both horrible for societies in times of climate change - and not conducive as objective functions for more creativity or education. In the sense of raw reach & economic impact, the advertising industry is arguably the most powerful "art form" of the 21st century. And so when you ask for "concrete example" of dystopian generative systems, we must look at developments in this area.
One of the influential developer of creative tools - Adobe - recently has doubled down on the intersection of """AI""" and """Creativity""", with their "Sensei" Initiative. Guess who they are targeting with their tools? Yes, Advertisers. In the sector of advertising, the quality VS quantity debate is very different: it is NOT about the values of art in the classical sense, or about storytelling or other humane things - it is about naked money. This is a clearly parameterized signal to feed to machines as loss/reward-function - and due to this fake clarity, such systems can grow like cancer.
Such economic success is precisely what drives humans to extend generative advertising logic to other creative fields. Repetitive generative muzak in the charts and endlessly boring superhero movie rehashes on the screen are a reality already today, and all using some form of "optimisation" algorithms. And so we see the rise of "creative" botnets - the rise of fake news and ever more fractaline simulacra's (see this post: https://samim.io/p/2018-04-23-the-solution-to-exploding-fake-news-and-spam-botnets-ca/)
I personally do not agree with the politics and dynamics of the visions outlined here. My writing is meant as a warning pointer in search space. We must strive towards more appealing outcomes, for humanities sanity and well-being in times of climate change might very well depend on it.
-
The Berlin BioRobotics lab is doing interesting work: http://berlinbiorobotics.blog/projects/
incl. using ML to track all bees in the hive and using robotic fish for the analysis of collective motion. #ML #FFHCI -
Autocomplete 3D Sculpting (SIGGRAPH 2018) - by Mengqi Peng, Jun Xing, and Li-Yi Wei. Project page: https://1iyiwei.github.io/a3s-sig18/ #Generative #ML
-
Google’s nature explorer has 3 machine learning models based on MobileNet, trained on photos contributed by the iNaturalist community.
These models are built to recognize 4,080 different species (~960 birds, ~1020 insects, ~2100 plants): https://aiyprojects.withgoogle.com/model/nature-explorer #ML #FFHCI
-
"A Universal Music Translation Network" - by Noam Mor et.al: https://arxiv.org/abs/1805.07848
-
Selection of Videos from SIGGRAPH 2018:
Neural Best-Buddies: Sparse Cross-Domain Correspondence:
Deep Video Portraits:
HeadOn: Real-time Reenactment of Human Portrait Videos:
Toward Wave-based Sound Synthesis for Computer Animation:
DeepMimic:
ToonSynth:
Semantic Soft Segmentation:
Gaussian Material Synthesis
Mode-Adaptive Neural Networks for Quadruped Motion Control:
-
"The Age of Assistance" - talk by @behshad_behzadi from Google on the impact of voice assistants on user behaviour.
