China’s Hottest Grandpa:
Wang Deshun is China’s latest and oldest model at 80. He started lifting weights at 50. Fifteen years later he started riding horses. Seeking something a bit faster, he shifted to motorcycles at 78. “How [should] we live our lives as we get older?” he quips. “However you want.” As we approach exit velocity on aging, Wang’s words of wisdom can inspire us all. (link)
The Neural Mirror
Italian design studio Ultravioletto has created a mirror that lets you see yourself the way AI sees you: as a collection of data points. Get an inside look at how self-driving cars, security cameras, spacecraft and AI toys perceive the humans that control and operate them. (link)
AI’s role at work? Stress relief.
A study by Verint of 18,000 workers shows that AI’s role in the workplace is basically stress relief. Taking over menial tasks and elevating the level of work eliminates toil, one of the top drivers for adoption of tools in society. Workers state that AI tends to improve the quality of work while at the same time reducing levels of stress, positively and directly impacting customer experience. (link)
GOFAI loves Deep Learning
Ben Dickson writes about recent advances in (1) combining Good Old Fashioned AI (GOFAI) with deep learning, (2) transfer learning, and (3) using GANs to generate synthetic data as techniques to combat the data problem in deep learning. The first one caught my eye. Here, MIT researchers first use a neural network to map an image to a collection of symbols / labels at different parts of an image. Next, they applied GOFAI rule based systems to answer questions about an image. (link)
Notebooks are on fire
Data Robot has raised another $200 million to fund expansion of their notebook-centric approach for data analysis. I look forward to a Software 2.0 world where your entire IDE sits within a browser, with all the rich interaction we enjoy today in laptop editors, with complete access to near infinite compute and storage. Notebooks are a key part of this. We’re probably 10% of the way into creating the new toolchain that lets us combine data-centric software of deep learning with traditional coding. (link)
The Larry Bird of eGames
e-Games has its Larry Bird moment as @bugha earns $3m. Larry was the first basketball player to earn $5m as an annual salary, which seemed outlandish back in the day. I keep wondering when we’ll see a National College E-Games Association (NCEA) as a peer to NCAA. (link)
Last of the Innocents
Those born in the early 1970s are the last generation to grow up without the Internet. The article notes we could regain some of that innocence by going for a long walk without our phone, spending an afternoon writing in longhand, or reading 150 pages in one sitting. These activities are “simple in theory, but strangely terrifying in practice.” (link)
Three problems in ML
A leader of Google’s Applied Science notes that problems often arise in machine learning when we (1) split data in awkward ways between development and testing, (2) artificially isolate a problem by leaving out critical variabilities that occur in real life, and (3) optimize our models for the wrong outcome. (link)
The Market Potential of AI
When challenged by critics of his $100B AI Vision Fund, Softbank’s CEO said “when the Internet revolution began more than 20 years ago, people asked me similar questions. But what happened in the end? The internet spread far and deep into all corners of life. The same goes for artificial intelligence. The world will not grow tired of AI.” (link)