You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
ML researchers work with fixed benchmark datasets, and spend all of their time searching over the knobs they do control: architecture & optimization. In applied ML, you're likely to spend most of your time on data collection and annotation -- where your investment will pay off.
In general, there is very little research done on best practices for data curation / cleaning / annotation, even though these steps have more impact on applications than incremental architecture improvements. Preparing the data is an exercise left to the reader
So true... there’s really a disparity of information on generalized data collection and processing (at least when comparing to the amount of resources for core ML/DL topics).
What do you recommend as best practices for data preparation?
Decades of work have happened here in the chemometrics community. It’s not deep learning but working with semi-designed experimental data with ML methods have been discussed in that literature.
@fchollet do you know of a good “cook book “ of common practices (excluding random notebooks written for Kaggle challenges ) ?
And what does this tell us about the ethical issues, bias and racism in particular, in AI? Is AI research oriented towards better and better reaffirmations of systemic biases? Who's job is it to study the ethical failures of the field?
Yann Lecun says the AI researchers are off the hook on that, and the responsibility falls on Engineers. Ethical issues with AI are when Google fires their ethicists & countries laws and data produce mostly class imbalanced models overly targeting a particular minority or subset.
Data Governance and Master Data Management is widely neglected. It's not sexy, and has to deal with all the nasty technical debt that accumulates exponentially. It also has to keep pace with the enterprise roadmap. So fundamental, but so underpowered, like much of EA.
a gift that keeps giving @MikeHypercube. Add data quality management for more fun.
Speaking as someone who does this as a job, basically, with major penalties for getting anything wrong, it's a brutal place to be missing best practices. I miss applied ML - especially NSF R&D grants where you need to research and productionize. Good times.
Exactly. There is a definite need for AI experts with a cultural history/social science/library background only AI companies and research orgs don't know it yet.
Above all, multidisciplinary teams and awareness are needed so that publications that address concrete cases without beating the state of the art by 0.001% can be more widely accepted.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
