Over the past two months, the White House ran a series of workshops geared to “prepare for the future of artificial intelligence.” At each session, technologists, academics, policy makers, and social scientists discussed social and economic issues stemming from data technologies, which will feed into the development of a public report later this year (likely similar to the FTC’s January report on the social impact of big data).
The last session took place yesterday in NYC. Mike Williams, Friederike Schuur, and Miriam Shiffman represented Fast Forward Labs at the event, and reported these highlights:
- Roy Austin of the White House Policy Council underscored that the government has a long way to go to improve data collection practices. To illustrate his point, he mentioned that the 2014 data on hate crimes only contained one hate crime incident in Alabama for the entire year. Before reaping the benefits of analytics and algorithms, therefore, the government has to update systems to collect data, being mindful not to bias minority and racial groups.
- Deep learning icon Yann LeCun said unsupervised learning is the next frontier for the research community. Yoshua Bengio (and many others) expressed the same in his interview for a recent O’Reilly publication. Indeed, supervised learning, where systems take in labeled training data to learn to perform tasks like identify objects in pictures, is currently the state of the art in many machine learning techniques. Getting useful labeled data can take time, requiring lots of manpower behind the curtain to get systems working.
- Cynthia Braezeal from the MIT Media Lab commented that machine intelligence is different from human intelligence. As such, use should be focused on how to pair to complementary but different skills to improve outputs: we should accept the oddness of machine intelligence instead of aiming to mimic human intelligence. Eric Colson from Stitch Fix thinks of the human-machine partnerships in their recommendation algorithms similarly.
- Harvard Professor Latanya Sweeney, who’s done extensive work on data privacy, discussed how easy it can be to identify individuals from fragments on health data. She commented that “computers aren’t evil” in themselves, and called for efforts to “scientifically harmonize computers with society.”
As always, participants mentioned the syncopation between innovation and regulation, the fact that Silicon Valley and Washington operate at different frequencies. Knowing policies and regulations can’t keep up, what ethics should we espouse as we build products? If we anticipate potential negative social impacts, does that stymy creativity or pollute research? What is the most productive way for us to address potential risks while still fostering and promoting creativity?
The debate is open. It’s ours to shape.
More from the Blog
Jun 10 2016
with — A probabilistic latent component analysis of a pitch class sequence for The Beatles’ Good Day Sunshine. The top layer shows the original representation (time vs pitch class). Subsequent layers show latent components. What is music? Or rather, what differentiates music from noise? If you ask John Cage, “everything we do is music.” Forced to sit silently for 4’33”, we masters of apophenia end ...
with Juan Pablo Bello
Jul 11 2016
If you’re a New Yorker looking for something interesting to do Saturday afternoon, Fast Forward Labs will be speaking at two events. PyGotham | July 16 | 2:15 pm Mike Williams on Text Summarization PyGotham is a conference for the Python developer community. Mike’s talk focuses on text summarization: taking some text in and returning a shorter document that contains the same information. ...
Oct 4 2017
by — We’ve started work on our next prototype. While the design is still evolving, we’re pretty sure one element of it will be a visualization of tens of thousands of data points, clustered through a dimensional reduction algorithm (most likely using T-SNE). For the past week I’ve been exploring how to render that many points in the browser and I wanted to document some early lessons in this post, s...