754 karma · joined January 27, 2012
That being said, an easy-to-understand explanation of why it happens (to the extent that is possible), like in some of the other comments, would be super interesting as well.
Would large farm lobbying groups fight against Deere's practices, when harming small family farms drives business to larger agri-businesses?
You mentioned that Kaggle competitions can be solved "without requiring deep learning". This benefits Google because Google doesn't need a world where every developer is a deep learning expert. Instead, Google needs a world where developers have a little experience with AI, enough that they can look at Kaggle as a friendly, non-intimidating way to get started solving problems with AI. And fundamentally, more people are interested in "solving problems with AI" than "building AI to solve problems from scratch", which will lead to more customers for Google's AI/ML based APIs.
In my limited past experience, I've liked the fact that on Kaggle, I can understand the marginal benefits of different approaches. I can try a naive approach with simple ML techniques, and achieve a score of say ~60. I can see (after the competition is complete) that an optimized version of my same ML solution can score ~70 from some clever tricks. And perhaps I can see that the winning solution added a lot of complexity to achieve a score of ~75.
Far more people have intentionally built their way to $9B than have intentionally reduced their fortune to $9B.
Definitely shady, but did the donation still make the world a better place?
Anecdotally, it's been interesting to see GoPro take off over the years, spreading from "extreme sports pros" at the beginning to include more casual/recreational uses now. A lot of friends who wouldn't have considered buying a GoPro 5 years ago (because "they don't do anything exciting/dangerous enough to need it") are now considering GoPros to document the slightly more exciting parts of life. These same friends are even more strongly considering Spectacles, because Snap made a product that fits into their lifestyle (lowering the barrier of what is "interesting enough" to capture)
But that's a great point that any kind of active trading strategy needs to be compared against the right index. I think this holds true for pension/sovereign funds which invest in multiple hedge funds as well - comparing to the overall market performance may be less relevant than comparing to sector-weighted market performance, for example.
A friend who works in quant finance said something along the lines of "if a trader can identify a mom-and-pop investor, it's like taking candy from a baby" in terms of making an easy profit.
That's why many 'hot' areas (like machine learning, for example) have a large gap between industry and academia. Researchers focus on "innovation" in terms of new approaches, aiming for marginally higher performance, and requiring increasingly complex models/infrastructure to unlock the remaining performance. On the other hand, combining an existing technology with industry knowledge of how to apply it can have a huge impact.
That being said, we have some beautiful state parks in Florida (great springs, rivers, camping etc) which charge entrance fees. Some parks have a once-a-year "free day", and it's great to see so many people out and enjoying the outdoors. Not at all like the "beach version of public housing", as one commentor rudely claimed above.
More restrictive set of courses for the AA means you're taking classes you don't enjoy for a requirement you don't need. The "name value" of a community college AA also isn't as great, particularly for kids who want to go to a higher-ranked public or top private university for undergrad.
That being said, if you know you want to go to a local/state school where the credits will transfer easily, getting the AA early can be great. I know people who graduated early but still received all 4 years of scholarship money (direct deposited in their bank).
I agree that defining "advanced" is very challenging, and goes beyond just the academic as well. There's also something to be said for being academically engaged with people similar to you (ie - some (most?) high school kids may learn more from peers in a self-selected AP Calculus class than they would from a Calc 1 class at the local college taken by kids looking to fill a requirement).
A few machine learning-based classifiers (we've used Bayesian and SVM approaches). Word embeddings and topic modeling (similar to word2vec) which are based on shallow neural networks.
Those are a few of what I would consider the "real" machine learning tools we use. Most of the application, though, is statistics/pattern recognition/visualizations on top of the data calculated by the ML approaches.
The interesting thing is (in my opinion/experience) that a 10% improvement in some of the ML performance (a 10% increase in accuracy, for example) will translate to a 1-3% improvement in end user experience (they see slightly better insights and patterns, but it is a marginal improvement). On the other hand, layering a new visualization or statistical heuristic on top of the data can lead to a significant boost in user experience.
Again, this is just for our specific application/domain, but we focus on making the ML results more accessible to users instead of focusing on the marginal accuracy of the ML results themselves.
We use "real" ML for sentiment classification, as well as some of our natural language processing and opinion mining tools. However, most of the value comes from simple statistical analysis/probabilities/ratios, as other commenters mentioned. The ML is really important for determining that a certain customer was angry in a feedback comment, but less important in highlighting trending topics over time, for example.
I thought this was an interesting perspective - Alex Roy is famous in car enthusiast circles for his driving records over the past 30 years. He's set speed records with drives across the US and around Manhattan, and has set electric/semi-autonomous records in a Tesla Model S in August 2016.
The point being that Alex Roy has probably spent more time thinking about driving, planning trips, and understanding traffic rules than most other people alive today. His perspective is just a single perspective, but given his massive experience with car-driver systems I think it's an interesting perspective.
However, I think that in this case, the author of the tweet has a lot of credibility (https://twitter.com/AlexRoy144/status/791996855114694657) as the holder of multiple performance driving records. You're right that it is only one anecdotal perspective, but it's kind of like a prominent Apple blogger tweeting that a new Microsoft product has phenomenal UI. It's just one perspective, but (to me) it's interesting enough to merit further consideration.
Journalists can be manipulated, but Alex Roy's tweet (https://twitter.com/AlexRoy144/status/791996855114694657) is more an endorsement by a car enthusiast with 30+ years of high performance driving experience than a journalist, IMO.
To me, a 996 sounds like a great deal today because prices are driven down by people who don't like the IMS issue and the egg-shaped headlights.