Kiana Khansmith
let's talk about Bridgerton tea, my ask is open
d e v o n
tumblr dot com
almost home
occasionally subtle
Cosmic Funnies
KIROKAZE
Misplaced Lens Cap
styofa doing anything
Show & Tell
No title available
Cosimo Galluzzi
Stranger Things
cherry valley forever
if i look back, i am lost
noise dept.
Lint Roller? I Barely Know Her
titsay
ojovivo
seen from United Kingdom
seen from Iraq
seen from United States
seen from United States
seen from Malaysia
seen from Italy
seen from United States
seen from T1
seen from France
seen from Türkiye
seen from T1
seen from United States
seen from Singapore
seen from Canada
seen from Malaysia
seen from T1
seen from Netherlands
seen from United States
seen from United Kingdom
seen from T1
@rubinovitz
AI researchers from 30 organizations propose a toolkit for turning ethics principles into practice, including implementing "bias bounties" for AI.
I'm featured in a writeup about our cross-institutional paper on trust mechanisms in AI.
Bias Bounty Programs as a Method of Combatting Bias in AI
This policy comes about as a response to continuous deployment of biased Artificial Intelligence systems into production, only to quickly be found biased with the only consequences being unfavorable news coverage. Bias Bounty Programs could provide scalable oversight to harmful discrimination by AI.
This policy assumes that this is unfavorable for both parties:
Those affected by bias - they will rarely receive enough news coverage of the bias to maybe get an apology and maybe a fix
Those deploying biased systems - usually a homogenous small group deploying a biased system, with no clear guidelines and imperative to debias it (in fact they could be penalized for taking the time to do so), but will maybe get a retroactive slap on the wrist if the media picks up on their bias.
Proposal
A similar problem exists in information security and one solution gaining traction are "bug bounty programs". Bug bounty programs seek to allow security researchers and laymen to submit their exploits directly to the affected parties in exchange for compensation.
The market rate for security bounties for the average company on HackerOne range from \$100-\$1000. Bigger companies can pay more. In 2017, Facebook has disclosed paying \$880,000 in bug bounties, with a minimum of $500 a bounty. Google pays from \$100 to \$31,337 for exploits and Google paid \$3,000,000 in security bounties in 2016.
It seems reasonable to suggest at least big companies with large market caps who already have bounty reporting infrastructure, attempt to reward and collaborate with those who find bias in their software, rather than have them take it to the press in frustration and with no compensation for their efforts.
Determining what bias is
It is assumed here that the company will determine what bias is in accordance with "their company's values" as they want to market to perceive them.
Potential Problems
Spam
Possibly the most cited issue with security bounties is the amount of false reports that come in, and the amount of people it takes to triage them. However, this does not seem like too much to ask from companies who professionally make content prioritization software.
Reluctance to Hire Triage Staff
These companies are controversial already for not hiring staff to even interact with paying customers, so this could be a hard sell. However, press pressure has led to hiring of more moderators at both Youtube and Facebook recently.
Adoption
Option A. Voluntary Enrollment
Companies decide this is a great idea (or better than eventual government intervention) and budget and implement them themselves.
Option B. Regulation
Bounty programs can be mandated by the government, most easily in any software government themselves use.
UX Practices
Where should the bias bounty program live?
In the application? (e.g. under "help")
On a company run separate webpage?
An independent bias bounty marketplace where companies can work together to share biased models?
I think a combination of one and two are the most likely, with one and two being mobile and web versions of submission forms, respectively.
Conclusion
This is a first attempt at solving a hard problem. Feel free to send jb@rubinovitz dot com feedback. I would love to figure out a way to hasten the iterations on debiasing in production AI models while compensating those affected by them who have to expend labor reporting them.
Acknowledgements: Thank you to Omar Bohsali for sharing his expertise in information security bounties.
Bias Bounty Programs could provide scalable oversight to harmful discrimination by AI.
Bias Bounty Programs could provide scalable oversight to harmful discrimination by AI.
Amid a series of scandals and sins, a few righteous tech innovators actually brought positive change this year.
I have not spent the time it deserves to do a writeup on Bail Bloc, but working on it as a co-creator last year was one of the best things I have ever done and I will definitely keep holding my work to this standard. Thanks NYT for mentioning us in "Some Things About Tech Were Good in 2017." That's what I at least was going for :)
NIPS 2017
Here are notes from two talks I particularly enjoyed
Talk Notes: Slow Judgement
Talk Notes: Surpassing Human Intelligence
Note: This post contains explicit and offensive language in the form of "toxic" Wikipedia comments used as training and testing data in this analysis. It's no longer controversial to say the internet needs better moderation tools. One attempt at automated moderation was Conversation.ai's Perspective.ai. Unfortunately, the
New blog post on the blog I need to migrate this blog to: a preview of my research in interpretability of toxicity models with attention!
Fast.ai vs Deeplearning.ai: which deep learning courses should you take?
As deep learning has become more popular, two courses stood out to me as having really useful teaching styles and reputable staff behind them: Fast.ai and Deeplearning.ai. As someone who has a theoretical background in deep learning and picked up Tensorflow on an as needed for a project basis, I was really interested in learning how to build complex deep learning architectures from scratch, and I’m always found I can sharpen my skills by hearing different experts describe known concepts.
Along with my goals of sharpening skills, I have been working alongside folks at the Recurse Center who are learning deep learning from scratch, and am using some of their feedback for evaluating the courses from that basis as well.
So, I delved into both courses and will share here what I found.
Theory
Theory in Fast.ai
The way Fast.ai handles theory is through visual examples and take home readings. I do not think this would be sufficient for a strong grasp of theory unless you can form a great study group to go over the readings and concepts. Both courses do have forums though, which could help with this.
Theory in Deep Learning.ai
The theory teaching in deeplearning.ai is really strong, with optional videos to watch that are of use if you have a calculus background to go into detail.
Winner: Deeplearning.ai . It's hard to beat Dr. Ng at explaining theory.
Applications
Applications in Fast.ai
I think it helps a lot that Jeremy Howard, a fast.ai professor, started a deep learning startup after seeing early on that modifying an out of the box ImageNet model could provide better results on classifying some medical imagery than top physicians. The hacker mentality is strong here as shown by show by having you able to submit to a Kaggle contest for what was at the release of the course a placement in the top 50% of the leaderboard, after lesson 1, and I’ve already reused several code snippets I’ve developed while going through the Fast.ai course, in projects.
Applications in Deeplearning.ai
While I could see myself using some code snippets from Deeplearning.ai (provided I downloaded them), I have not been able to easily translate the Deeplearning.ai code I’ve written into projects I’m working on. Deeplearning.ai also spends a lot more time teaching theory, especially early, and has yet to release their computer vision and sequence model components, so this will probably change soon.
Winner: Fast.ai since I am using code I’ve written there in projects already, whereas Deeplearning.ai code is mainly about learning the theory until their more application driven courses are released.
Portability
Portability of Fast.ai
Assuming you have access to a machine with GPU, this code is super portable, as you will be developing it all locally or on your own cloud box, and you will be developing modular solutions for real deep learning problems that you will continue to build on through out the course, within jupyter notebooks.
Portability of Deeplearning.ai
Deeplearning.ai really pales in comparison to Fast.ai here, for several reasons:
They wiped my homework notebook clean without prior notice when I stopped paying the monthly fee.
Since all your work is in their cloud, you need to explicitly download each Jupyter notebook of your homework before it is wiped.
I don’t find the code that applicable/extendable to the projects I’ve been doing in industry/academia.
Winner: Fast.ai hands down is here with reusable, extendable code provided you have a machine to run it on to begin with.
Accessibility
Accessibility of Fast.ai
I would say Fast.ai is super accessible as far as teaching style and language choice (Keras and now Pytorch which they will use in the future, are a lot more accessible than Tensorflow), so as long as you can afford to rent an AWS GPU box and figure out how to run their environment installation script on it or have a deep learning rig already (I do), it is a very accessible introduction.
Accessibility of Deeplearning.ai
One way Deeplearning.ai makes itself super accessible is by hosting cloud Jupyter notebooks for you to do all your work in. This made the class pretty frictionless to start on, and is why I started it first.
By pretty frictionless, I do imply there is still friction, which is true. One of the big turnoffs of mine towards this class is that they ghost auditing it. By ghost I mean that when I and several other of my peers first tried to take this class, they did not show auditing as an option, but when we came back to the site through a search engine, the auditing option appeared on the site. I believe this will inhibit many people who want to take the class, but can’t afford the $50 a month cost, from taking it.
Also, at this point I am not convinced learning Tensorflow is the best way to learn deep learning, the syntax is highly nuanced and takes time to grasp that one could be spending learning more about the concepts.
Winner: Fast.ai if you can access a GPU machine, Deeplearning.ai if you cannot.
Conclusion I ultimately think this is a trick question, even though Fast.ai did win on the personal evaluation scale I chose to evaluate the courses and I’ve spoken to several people who regret not starting with Fast.ai. I think some combination of both courses, fast.ai for applications and deeplearning.ai for theory, is the optimal use case.
Jennifer Rubinovitz and Amelia Winger-Bearskin offer an overview of how artificial intelligence researchers and artists at the DBRS Innovation Lab have collaborated on five different projects (and ...
I spoke on leveraging AI in creative technology with Amelia this week at O'Reilly AI. I will update this post with the slides when they were up. Thanks O'Reilly for a great conference!
“Creative Jobs May Not Be Safe From the Robot Takeover—Here’s Why”
Aaron Arntz, at piano, plays with Recurrent Neural Network, at Le Poisson Rouge. (Photo: Brady Dale for Observer)
Observer covered a generative music piece I collaborated on, Deep Piano. I am REALLY looking forward to the recording of the performance of the piece by the band Aaron Arntz assembled, Recurrent Neural Network, to share with you. The MIDI file released in the Observer article does not do this piece justice.
“You guys may be surprised to hear this, but I didn’t really get into tech until college.”
Here's a write up about an in-progress virtual reality convolutional neural network exploration coming out of the lab!
Recurrent Neural Network, a group of amazing musicians playing work from my generational music project in the post below, will be performing at le poisson rouge March 30th!
The DBRS Innovation Lab teaches a recurrent neural network to compose piano music.
I recently had the opportunity to collaborate with an amazing keyboardist and friend, Aaron Arntz, as well as the rest of the DBRS Innovation Lab, on a recurrent neural network to generate piano music. You can read more at the Medium post linked above.
Open Hackathon Data Talk
I gave a talk on Open Hackathon Data at the last Hackcon. Excited to update you on the progress soon.
We've seen how hackathons change lives. And students tell us: they blog, they post, they tweet, and they scream at the tops of their lungs about it from dorm rooms. Unfortunately, if we only sung the successes of hackathons, we wouldn’t be telling the...
I'm working with the Major League Hacking on an official inclusivity partnership for hackathons so we can put more resources towards training organizers to foster events that are welcoming to everyone. Pretty excited!
The Future of Hackathons
This is a post I have been afraid to write for a while, because I could not find a happy ending for it. I went to my first hackathon, PennApps, in Fall 2011. I did not have any friends in the hacker community yet, so I mainly hung out with dev evangelists and eventually ended up asking a Tumblr evangelist how to apply for a Tumblr internship, which led me to applying to hackNY, which led to me being a hackNY fellow for a summer and a hackNY mentor for two and meeting many of my closest friends and collaborators there. So when the other day my father looked at me and voiced "hackathons changed your life..." like it was some epiphany, I said "yeah..." as if it was obvious. I have been afraid hackathons will no longer change lives for the better, like they did for me. The idea that anyone can come to a hackathon and build is incredibly powerful. Hackathons have grown so quickly, but at the expense of not putting the work into keeping them inclusive. After my first hackathon, I ended up interning at a startup in the same incubator and office as Tess Rinearson, who ended up writing this piece earlier this year. Tess was one of the first female hackers I became friends with, and reading that piece devastated me. It was the last straw after having so many female friends telling me they no longer would go to hackathons, other female friends telling me about the harassment they endured, reading about others enduring harassment, and enduring occasional harassment myself that made me realize I was not happy being passive anymore. I love the ideals of hackathons. I love empowering people to build things and giving people the tools to build the solution to their own problems. However, keeping any activity at this scale inclusive takes work, and while putting the resources into having the latest and greatest hardware, speakers, programming (I am sad I missed MHacks laser tag), and more into hackathons is tempting, I ask you to think about putting resources towards making sure everyone is welcomed into to the community. I think establishing a culture of inclusion is more important than some of us getting to enjoy playing with new technology $X, while on a moon bounce, listening to a speaker who sold their company for $Y million dollars. Please do not lose sight of the root of what makes hackathons great, powerful, and positive: they can (and should) be a place where anyone can learn, collaborate and build the solutions to problems. So, what of the future of hackathons? I think it relies on the members of this community stepping up and taking action to make sure hackathons accommodate everyone. I spent a lot of time being overwhelmed by the inclusivity issues in technology, but have recently realized there are ways I can make a difference and contribute. I have been able to do things at the scale of hackNY that I feel have made a difference in inclusiveness. Being able to see the incredibly diverse hackNY fellows class hang out as a single unit and collaborate despite their differences this summer are some of the more rewarding experiences of late I can remember. I have faith in the hackNY community because I know the community will fight for inclusivity. Example: hackNY was one of the first hackathons with a code of conduct, and said code of conduct was put into place by an ally. Not because he was the only one with the ability to do it, but because he was the first to have the foresight and drive to make it happen.
At the time of writing this I am the only female representation in Major League Hacking team and this past weekend I became the first female admin of the 6500+ members and growing Facebook group Hackathon Hackers (which has thus far given minorities in technology a lot of grief). I feel like I have a lot of responsibility for making the hackathon community the better place, and while I cannot do it alone, I do not have to.
There has been discussion about hosting a git repository for hackathon organizers in general that would, among other things, have resources for fostering inclusivity at hackathons. Unfortunately, the "other things" have not been published yet, but I have started a git repo with some standards to start iterating on. It includes the code of conduct suggested by Cassidy Williams during MHacks, and some resources myself and others have made during our time in hackathons. I also expect great inclusion efforts to come out of MLH.
Two good things that have come out of hackathons growing this big are:
1. Major League Hacking is emerging as a unifier of hackathons, and this can allow us to set standards across hackathons. 2. People are starting to step forward and say they want to help making things better. However, this is largely being done in a 6.5 thousand person Facebook group, which is not very productive.
How can we do this?
Check out the git repository where I started brainstorming hackathon standards. Contribute and/or show it to hackathons you are involved in or that are at your university so they understand what can be done and how. Get someone to own tasks that would make your hackathon environment better.
If you and/or your company is interested in sponsoring hackathon inclusivity, we are in talks of figuring out how to do this and what it would mean. Shoot me an email at [email protected] to join this conversation.
I have received some messages from people losing faith in the hackathon community. I would ask you to have some hope that things can change in the community this year, but that you put yourself and your needs first. I personally believe things can get better given the great, smart, people that do exist in this community.
The future of hackathons is up to all of us. I hope you will join me in trying to push for the standards we deserve.
A presentation from the 2014 NYCRIN Networking event on startup metrics and Lean Workbench.
A project I worked on in a Computer Vision and Machine Learning for Mobile Devices class at Columbia for the Spring 2014 term.