Conan's Newsletter: No. 4
AI Talent Tracker, economics and job listing plummet; Games for remote teams; The biggest lake in California was made by an engineering mistake; How Moai on Easter Island "walked";
Happy labor day! 🎉 Here are recommendations for this long weekend and hope you will enjoy it.
The Global AI Talent Tracker. This great analysis from macropolo.org shows how AI talents migrate among different countries. The United States has a large lead over all other countries in top-tier AI research, with nearly 60% of top-tier researchers working for American universities and companies. The result also clearly shows the US lead is built on attracting international talents as more than ⅔ of the top-tier AI researchers received undergraduates in other countries.
Taming the Tail: Adventures in Improving AI Economics. This impressive essay by Martin Casado and Matt Bornstein, two analysts of Andreessen Horowitz, discusses the economics of AI.
AI development often feels “closer to molecule discovery in pharma” than software engineering. This is because AI development is a process of experimenting and the job of an AI developer is to fit a statistical model to a dataset, test how well the model performs on new data, and repeat.
One of the complexities in the real-world data is that a lot of them have long-tail distribution. Intuitively, what this means is that that there are a lot of edge cases that appear infrequently yet important. This essay presents a guideline to tackle the issue and other similar ones.
Another interesting thought that I have after reading this essay is that the common metaphor of “data is oil” is not quite accurate. The energy of one barrel of oil is always the same regardless of the amount of oil you have used. In contrast, the benefits of extra data diminish as the size of data increases. The reason is that although training with more data helps the handling of new edge cases, the chance of seeing new cases decreases as you move further down to the tail. As a result, it is an art to find a balance of the predictive power and the complexity when we decide how much data we need.
AI job listings plummet as COVID-19 recession appears imminent. François Chollet, the creator of Keras and a researcher at Google Brain, posted a tweet that shows that AI jobs plummeted since the start of the COVID-19 recession. One of the reasons is for the correction of the hype in the past few years. However, this doesn't mean that we should be bearish for the future of AI. Instead, COVID19 has been a catalyst for technology adoption for many industries (e.g., Zoom for video conferencing). I am optimistic that the pandemic would actually make more traditional companies start to embrace AI technologies.
A lot of teams have moved to be fully remote during the pandemic. This blog lists a lot of virtual games that may be helpful to increase the bonding of the team and. In addition to those listed in the blog, I also highly recommend the drawphone game, I have tried it many times and enjoy it a lot.
🍭 Interesting Facts:
The Salton Sea is the largest lake within California and it is fascinating to see that it was created only in 1905 purely because of an engineering mistake. Unfortunately, it is dying because of the reduction of the input water recently.