With the help of Derek Robertson
In the second issue of our regular Friday column The future in five questionsSeturaman Panchanathan, a computer scientist who is head of the National Science Foundation, weighs the power of science to change the world.
Answers have been edited for length and clarity.
What’s one underrated big idea?
Revealing all the talents of this nation. I don’t think we took full advantage of the talent – every day I wake up thinking this is a missed opportunity. We must make full use of all local talent and then welcome global talent like there is no tomorrow.
I want people in schools to have what I call the “spark of STEM” – that moment when science lights up. I had a STEM spark when Apollo 11 landed on the moon. Back in India, the US consulates received the moon rocks and put them on display. Seeing this rock, knowing that it came from the moon and what made it possible, I, an 8 year old boy, was really excited.
What technology do you think is overhyped?
I don’t think any technology is exaggerated.
Take artificial intelligence for example: over the past few decades, artificial intelligence has been criticized as a hype. But it’s like any scientific process: you take risks, you fail, you get up and do better. You may not get the perfect solution the first time, but that doesn’t make it too hyped.
What book has most influenced your vision of the future?
I’m looking for nuggets of incredible ideas. At the moment, a book that I find very valuable, Fast Start America Jonathan Gruber and Simon Johnson. This reinforces my view that, yes, America has plenty of room for innovation and potential.
What surprised you the most this year?
Our resilience in this moment of COVID. I don’t have to say I’m surprised, but I’m grateful and bewildered. I came in as director of NSF during the pandemic and the incredible human endurance and tenacity that people have shown has done so much more than in previous years.
What can government do about technology that it is not?
Our economic and national security depends on our ability to invest heavily in today’s and tomorrow’s technologies, innovate everywhere, creating ecosystems of innovation in every region of the country, and quickly develop our local talent in every geographic and demographic context. Investments in future technologies such as artificial intelligence, quantum technologies, semiconductors and alternative energy are of great importance for national security and climate change. Therefore, it is imperative that the federal government redouble their efforts.
Google trained AI on a massive dataset of Reddit comments last year. for the better emotion detection in typewritten language.
It sounds conceptual, but like everything else in computing, it obeys one of the oldest laws: garbage in, garbage out. Researchers from Surge AI recently digging into the dataset and found that about a third of them were flagrantly mislabeled by people unfamiliar with American idioms. A few notable (or, frankly, just entertaining) examples:
- “girl yeaaaaaamn!” – tagged as “anger”
- “Hi, I’m dying, I’m dad!” – labeled “neutral”, according to Serge, most likely “because the labelers don’t get daddy’s jokes”
- “Hooray, cold McDonald’s. My favorite.” – tagged as “love”
- “Those two are disgusting little kids” – tagged as “approval” (??)
- “But a lot of storytelling! orange man caused it!!!!!” – marked as “neutral”
The author of the post, Edwin Chen, uses these examples to emphasize that in AI research amount or elements included in the data set are often erroneously given priority over them qualitative, resulting in poor-quality moderation when the tool is applied to actual content. — Derek Robertson
And now almost a literal afternoon snack: Florida researchers are using AI to help protect watermelon, the state’s $200 million marketable crop, from blight.
As Hannah Farrow reports for Pro subscribers, a study backed by the USDA, is designed to protect watermelons from so-called “downy mildew,” which can destroy an entire crop in a few days. BUT Scientific research published by the researchers describes how they used machine learning models to better identify the different stages of mold development and therefore make it easier for farmers to quickly treat or remove diseased crops.
The National Institute of Food and Agriculture of the United States Department of Agriculture has $300 million program which partly directs funds for research AI technology in agriculture, in everything from robotics to early detection efforts like this one to data evaluation to make farms more efficient and sustainable. (There is also purely AI-driven altruism in agriculture: Last month, researchers presented an AI that can detect distress cries of chickens on agro-industrial farms.) Derek Robertson
Stay in touch with the entire team: Ben Schrekinger ([email protected]); Derek Robertson[email protected]); Konstantin Kakaes (ur.[email protected]); and Heidi Vogt ([email protected]). Follow us on Twitter @DigitalFuture.