I'm surprised people don't take the time to figure out what the options actually mean and negotiate some of the terms. When I have negotiated with startups I have paid for legal advice to help me figure out what language I needed in the option agreement (for example, a pinterest-style clause that prevents me from having to exercise within 3 months of leaving) and negotiated for the terms I felt were important. Just because the agreement is written to give you ISOs doesn't mean they will qualify for ISO tax treatment when you eventually exercise, so it is better to let you give up ISO treatment and convert to NSOs in order to avoid having to effectively forfeit vested options upon departure from the firm or face massive tax consequences. The pinterest-style structure should be standard.
It is really sad that the people funding these excellent researchers have no fucking clue about AI (they know plenty about other things and have done plenty of good work, but the level of nonsense here is striking). Thank God they are giving the money to competent people instead of doing what they think they are doing with it.
Even if you don't take what he says seriously, I think what he wrote in this particular post makes sense. When I negotiated with startups I demanded the 10 year exercise provision. I can't believe that isn't standard. I don't want to be, in effect, forced to forfeit vested equity. It is almost like leaving a company and being asked to return your wages or face a huge tax penalty!
Stay the course if the investment time horizon is 20+ years. Make sure you have an asset allocation you are comfortable with and aren't taking on too much or too little risk. Do nothing and ignore the noise.
Many people would prefer the NSO since they don't lose out on vested shares upon departure (or get saddled with a massive tax bill). If you have a good chunk of equity, the outcome can be pretty binary: worth something substantial or basically nothing, in which case the slightly higher tax isn't nearly as important as actually getting a piece.
Although that isn't the exact same statistic. Basically in NY black kids go to school with black kids and white kids go to school with white kids, although there might be both a white public school and a black public school.
In the USA, outside of California and possibly a few other states, it isn't clear this is legal. Employers may be able to prohibit employees from publishing their salaries, although they can't prohibit employees from discussing compensation with each other.
Because it doesn't meet academic standards for publishing in NIPS and ICML, the most prestigious machine learning conferences.
Edit: To clarify, research papers generally cite other peer-reviewed research papers in similar venues preferentially. ML papers should mostly be citing ML papers in high-quality, peer-reviewed venues. HTM doesn't have papers like this to cite.
He won't do things for the non-open journals without getting something in return. In this case, he gets a Nature paper in return. He presumably would be willing to edit one for sufficient pay as well, he just won't do it for free.
The author clearly doesn't understand the Szegedy et al. result and isn't really saying anything interesting. Pretty much ALL machine learning image classifiers suffer from these pathologies from Szegedy et al. so calling them "deep learning pathologies" is absurdly misleading. Logistic regression has them too as do almost all algorithms that learn to classify images. Human visual systems can also get confused by sensory input that (too us) looks like some object, we just are much better at integrating evidence from other sources to suppress the mistaken neurons.
The point of a professor at a research university is to interact with graduate students. At a teaching college they should be interacting with undergraduates.
I completely agree that there is very little chance AI research will succeed any time soon, but not because AI researchers haven't listened enough to Jeff Hawkins. The way to understand human intelligence is to build computational models that reproduce some of its properties.
Using your analogy I would guess Geoff Hinton's goal is to understand aerodynamics and which parts of bird flight are actually necessary for flight and which are simply accidents of evolution.
That is not accurate at all. These people can and did do lots of research in academia and had plenty of data. They obviously get paid a lot more in industry, however.
Hawkins has no credibility in machine learning so I don't think people should really care about what he says. To the extent that anything he does say about the field makes sense, it isn't novel. Unfortunately our society reveres the super wealthy and treats them as experts on everything.