sumodds

I agree and disagree. I am a candidate and prefer coding challenges provided they are reasonable. Meaning, if you are wanting to check the person's ability in x, check the person ability in x, not y. So for example give a starter code with part that would showcase person talent in x is left out. Yes, this involves prior effort on the part of the company to come up with this exercise and on the person. I don't quite vouch for code samples, because it is very difficult to understand where they came up with the ideas and who wrote the code, and if it was a multi-person project. In the former case, it is a very controlled experiment and one you designed, so one can judge much better.

That is a weak argument :). Not really, there is a certain and rather high possibility that common man can buy Tesla cars. Sure, there is a premium price tag now, since we are still chipping the first blocks in making that technology a n economic reality. And that specific case is all the more important, since as you too argue helps to reduce pollution.

"Anything that is profitable and legal is of benefit to society" is a little misleading. The definition of what is "legal" is an evolving thing. We know it is dangerous, in that can cause huge fluctuation in the financial control system without real benefit for the society as a whole except a very small percentage. I am not speaking of a central authority that decides it, but our common collective consciousness that should decide it. And again there is not one thing, which has no obvious negative externalities. But the question is whether the positives overcome the negatives. In this case, for the society as whole there is little benefit.

Exactly, when the very society is built on that idea that everyone can pursue whatever they want, the access to opportunities should be equal. HFTs are tilting the balance in one way (helping the rich getting richer) , can cause economic crashes, which hit the rich and poor alike. Now the aftereffects of crashes on the rich (except in case of exceptions) are not as tragic, as in case of the poor. For the poorer part of the society, it pulls the rug under them.

I am not as much against automated trading, as I am against High Frequency Trading.

Automated trading itself is still dangerous, in that many of these algorithms are black-boxes. They induce complexity and turbulence into the system. For example the minor bubble just before 2008 crash. May be automated trading/HFT was lucky that, the crash wasn't majorly their part. But it is a ticking time bomb.

Now with HFT it is all the more dangerous, because of the speed of interactions. If I make a bet in 50micro seconds, the space of possibilities explodes on you.

And how is it of enormous benefit to society, when it is concentrated in fewer hands. Again, these micro variations hardly say anything about the quality of the goods being produced than what other traders think about the goods. You are optimizing on a parameter, which does not model the problem you are solving. You might find a minima, but one that has nothing to do with making/distribution of these goods.

I disagree with this assumption. The problems with military research, law firms, advertisements are of degree not of its existence. Sure, they have their own problems. But we need military research: which is arisen out of insecurities of nation states, to given protection to their states. Law firms, too have a place so too medical clinics and advertisement firms.

In case of HFT, the argument is not of degree but existence. Military solves security problems of nation states, law firms communal issues of individual citizens, advertisements to help the product reach the intended audience. But what problem does HFT solve other than making a few people very rich.

I do Computer Vision/Machine Learning for a living, but would caution a little against super-focusing on a narrow area. AI in general is known for its AI winters, where jobs dry up and opportunities are far fewer and you could have a lot of people with undergraduate in ML (or whatever that era's AI is called). Now that said, who knows may be you might find it interesting enough that you may decide to go for graduate school. If I was giving advice to my younger self, it would be to learn Algorithms, Linear Algebra, and Probability really well. Get exposure to Machine Learning, and a little bit of Linear Programming. But if you really want to be able to apply machine learning or math heavy subjects to work, it might be a reasonable idea to do a Masters.

No problem here in Mountain View :)

Wallstreet ;)

I loved the first part of the argument, that many a time causality is not from one source but multiple (often overlapping and even contradictory sources). But towards the end, you sort of do the exact opposite :). The assertion that "But dropping the bombs on Japan saved millions of both Japanese and allied lives". Allied life, I can understand that, but Japanese life. Also the assertion that the bomb somehow magically safeguarded the Japanese culture.

Not that only, I had a high variance on my bias.. ;)

Lol.. my bad. I did not pay attention. I thought the error was in percentages. (I was comparing with MNIST and somehow assumed this too was percentages). Come to think of it, that is really dumb (what that would mean) !!

Am not sure if you can apply winner takes all for such marginal difference in error. Give a slightly different database and things go awry.

Check out : "Unbiased Look at Dataset Bias", A. Torralba, A. Efros,CVPR 2011.