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
Machine learning has the potential to make a big difference in solving some of humanity's biggest problems -- making renewables more efficient, optimizing our transportation networks, recycling our trash, making medical care more broadly accessible, accelerating science.
Most of it will take the form of 10-50% improvements to existing processes. It won't always be very sexy, but in aggregate, it will be transformational. It seems to me we aren't investing nearly enough researcher brain power on this type of impactful, "boring" applied problems.
Meanwhile, what is our community focusing on? What are the media & public focusing on? Misplaced hype. Fantasies about lofty AGI research that is sure to generate unlimited profits. Fantasies of AI apocalypse peddled by those who need you to be afraid. It's a bit disheartening.
Perhaps billions of $ will be wasted on those who make the craziest claims or shout the loudest -- or those whose tales of the future most accurately match the fantasies of tech billionaires. If that happens, the lack of ROI will eventually lead to a new AI research winter.
This may slow us down at a critical moment. Let's counter that. Let's pay more attention to those who are working on important applied problems. Let's fund them. Let's tell their stories, so others will be inspired to follow them. Let's work on the research & tools they'll need
If you know of any team or individual working under the radar on applying ML to solve an ambitious problem that benefits the public good, please reply and share their work. I'll amplify. Very curious to see what will show up :)
It is a real focus of @jeremyhoward and fastai to bring people from diverse fields with their domain expertise into the ML community
That’d be @jeremyphoward
Btw, is ML going to be the driver of the savings, or broad application of approximate dynamic programming? Often hard to disentangle effects (see Deepmind & Datacenter cooling - what part of the savings were due to "deep" function approximation?)
Wouldn't a search for 10 x 4% AI helped improvement projects on the course of 10 years be more credible/reproducible by others, so more appealing ?
One reason might be because it is a more complex and lengthy process. In our work in the critical care and hospital setting (e.g. recognizing patient functional activities) a project takes a few years.
It takes a long time to properly design the experiments, obtain IRB approval, build relationships with the physicians, consenting patients one by one to get a good sample size, properly cleaning data, and going over and over interpretations many times to ensure clinical validity.
If you compare that to developing and testing a (novel) model on ImageNet or any other public datasets, they have a much shorter cycle (and thus more rewarding, possibly).
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
