Take as much probability and linear algebra as you can conveniently do – as much for the intuition as for the symbol-manipulation mechanics – and don't underrate the importance of domain expertise in any problem you get interested in!
The Shipping Forecast is a very significant cultural reference in the UK. It crops up all over the place. Britain is _fundamentally_ an island and a seafaring nation, and that's something Americans miss; you're never more than seventy miles from the sea. It's as iconic as, I don't know, Thanksgiving football in the US; it's a thing everyone knows about without explanation.
And the music that runs up against it at night, Sailing Away, equally important reference. The nights I lay there, imagining the storms out there, particularly when depressed and alone, listening to the post-midnight "ceremonies" including the national anthem and lulled off to sleep before the World Service crept onto the airwaves... so many nights. It really is a part of me in some ways.
It's interesting to me just how important BBC Radio 4 (formerly the Home Service), on which it is broadcast, is and has been to our collective culture.
It's now seen in a more middle-class "sniffy" light - Radio 4 listeners are a certain "type", but think what it's given us:
- It's where The Goons became famous: Peter Sellers, Spike Milligan and Harry Seacombe defined a certain age of comedy and inspired Monty Python and others.
- Mornington Crescent is ironically one of the most iconic stations on the London Underground thanks to the game from the R4 show
- All of Churchill's war-time speeches were broadcast there first
- In the event of the death of the Queen, it will be announced first on the 8am Radio 4 news broadcast the following day (I'm not sure that can work in the modern era, but there we are)
- Royal Navy Nuclear (weapon) submarines are to open a safe with hand-written letters of instruction from the Prime Minister if they can't pick up Radio 4 long-wave
>> Royal Navy Nuclear (weapon) submarines are to open a safe with hand-written letters of instruction from the Prime Minister if they can't pick up Radio 4 long-wave
It is one of many factors. Shutting down Radio 4 will not send the subs into attack mode. I do laugh every time I read about the "hand written" part. The assumption is that PMs are unable to use typewriters. Anything typed would therefore have been handled by their secretaries/PAs, making them a target. I have this image in my head of a sub captain not being able to read the PM's handwriting. These are embellished protections generated by novelists.
NB Hennessy is obviously well in with the Navy - he has attended during the Perisher course - maybe an advantage to be being a historian who is also a Peer!
Be wary of believing any account when it comes to nuclear weapons procedures. There is much misinformation. Changes happen all the time. Some of the public stories have been adopted strait from fiction. The letters have been the subject of so many spy thrillers they are now more myth than functional.
The very concept of the letters presupposes that sub crews have full control over their missiles, that they can fire them based on instructions in a hand-written letter. That does not mesh with the safeguards surrounding other weapons that require external command/codes prior to launch.
It's also fairly well documented that UK Trident sub warheads do not have Permissive Action Links - the crews of the subs do indeed have the ability to launch without receiving any codes.
The justification for this is pretty simple - timing. In the event of a launch from a likely enemy (the Soviet Union in the bad old days) there simply wasn't enough time to guarantee that a code be transmitted before weapons bursting.
> The very concept of the letters presupposes that sub crews have full control over their missiles, that they can fire them based on instructions in a hand-written letter. That does not mesh with the safeguards surrounding other weapons that require external command/codes prior to launch.
The whole point of the submarine component of the strategic triad is to be the ultimate guarantor of MAD by presenting the capacity for launching a retaliatory strike in the event of sudden destruction of the highest command authorities (and, potentially, the land-based components of the triad), so it is absolutely plausible that submarines have a looser set of controls than bombers and land-based missiles because otherwise there would be no point in having them.
It is there to pad out the schedule and help people to find Radio 4 on a crowded night-time dial. Same with the National Anthem that follows. Even the 'pips' are there for this type of calibration, a tuning in to a British person's 'Britishness'.
The way that the BBC is deeply woven into the British establishment and armed forces is obvious yet not so obvious. The BBC is a spin-off of British Forces Broadcasting Service, as per 'Better Call Saul'/'Breaking Bad' some content overlaps but the 'producers' are one and the same.
You should never believe that the BBC is independently funded and free from some capitalist proprietor controlling what you think, it really is an organ of the British military industrial complex. Anyone 'socialist' gets weeded out by the very real 'Room 101'.
Radio 4 is the true voice of establishment, and this differs slightly from the clowns of the day that happen to be in parliament. If you listen carefully then you can glean facts on BBC Radio 4 that are not presented on the TV news or written to paper, an off-hand comment here or there on an odd-hour Radio 4 program sneaks through when the rest of the world are self-censoring themselves during a crisis.
When BBC Radio 4 rolls over and defaults to Number 10 propaganda without being honest, e.g. asserting that Russians are poisoning people in the UK without any 'alleged' or 'suspected' style words to distance fiction from fact, then that is when the folks in the submarine should go for the envelope...
I grew up in Miami. I can't imagine living more than a few hours drive from the ocean. So I sort of understand. FWIW, Hawaiians, Puerto Ricans, and other island-living Americans probably get it too. :-)
I grew up very near the Puget Sound in Washington state, and I feel the same way!
I just did a bit of research and discovered that in 2010, 39% of Americans lived in a coastal county. Given that many counties are rather small, especially on the East Coast, I wouldn't be surprised if as many as 50 or 60% of Americans lived within 70 miles of the coast.
Density functional theory, for example, is taught at undergraduate level as part of both chemistry and physics triposes. Probably materials science too, and it certainly used to be an option within earth sciences (as part of the mineral physics path).
There is nothing intelligent about machine learning. It’s cool but far from intelligent. AI buzzword to describe machin learning is equivalent to people calling computers cpus. Let’s me know really quickly just how clueless they are.
Not at the kind of resolution you'd want to be using on, e.g., Twitch. In that setting, you could just use chromakey, though? That's '70s technology, cheap and very reliable.
It's not quite as simple as "this one has highest mAP, let's use it"; the tradeoffs are complex. In particular, as you can see in the image here, one thing DeepLab doesn't do is segment instances – so you get a mask of "people", not a mask per person. Mask R-CNN does a better job on that by design, because it predicts both bounding boxes and a mask per bounding box.
Overall I'm really happy to work in a domain where people share their code and models in such an open way.
I take issue with detectron in particular though, because a company the size of facebook in the year of 2018 has no excuse to publish a major software package in python 2.
The oldest models they implement are from 2015 (excluding VGG16 which is so prolific it's available in literally every library as python 3) and caffe2 is quite a bit more recent than that. Like I said. No excuse...
The team behind Detectron have published an enormous amount of really good research, but the Detectron codebase struck me as "good research code" rather than something you'd ideally want in production.
Of course, I'm not criticising the fact that they publish those models, nor the models themselves.
But even publishing arguably polished python2 code in 2018 is something I take issue with if it's not a legacy code base
You hit the nail on the head. They also reveal this strategy through their data science "thought leadership" posts, which are seemingly meant to appeal to data folks (own your code! add value through modeling! don't worry about collaborating with engineers!) but do not reflect the realities of industrial data science.
This means that codec development makes no sense for anyone who doesn't own either a large distribution platform or a large playback platform. (Much of the research has been done by middleware companies attempting to tax the two; their business goes from royalties to work-for-hire, at best, which is way less attractive for them).
This is fine – in a macroeconomic sense – but of course it sucks if you're one of the companies being disrupted.
By that logic Free Software wouldn't exist, not to mention codecs like Vorbis.
The logic of software patents systematically underestimates the existing incentive structures. That's because of the simplifications economists usually make. (See https://en.wikipedia.org/wiki/Assume_a_can_opener)
The assumptions underlying patents become less justifiable as
1) the pace of innovation in an industry increases, compounding first-mover advantage;
2) complexity increases, because integration of an invention into a viable product becomes much more difficult--often such that copyright and other IP protections alone are more than sufficient to provide a moat;
3) as financing becomes cheaper and potential markets larger the time horizon for and amount of ROI demanded by investors diminishes;
Those are especially applicable to many modern high-technology industries, software (Silicon Valley) and industrial (SpaceX).
Other industries have proven to be able to thrive without strong patent protections or even copyright protections for myriad other reasons. The fashion industry has thrived without strong patent or copyright protections, largely based on trademark and marketing. Likewise for the beverage industry and consumer foodstuffs more generally. Major automakers had poor copyright protections, and while they had stronger patent protections, as safety regulations increased they shifted to a market segmentation model such that the big auto manufacturers often freely license patents among each other long before the patent expires.
Economists, especially conservative economists, often argue against anti-trust, reasoning that as markets become larger (finance markets, consumer markets, etc), the typical lifetime of a monopoly become shorter--often shorter than the typical lifetime of anti-trust litigation.
I would argue that as markets grow larger and more fluid they're far more capable of discovering and leveraging sufficient incentives to support innovation in the absence of government-imposed monopolies. Especially in the realm of software, the circumstances where there's a legitimate market failure requiring patents are so uncommon that there's no reason to even entertain the notion of a software patent system.
In the absence of _manifest_, _systemic_ market failures, patent advocates make the same mistake Communists make--an inability to imagine how a market can support an endeavor is not evidence that a market is incapable of supporting that endeavor; it's merely evidence of one's own lack of information and imagination. And the surest way to introduce market failure is by dangling the prospect of government-imposed monopoly rents, artificially increasing the opportunity costs of discovering and leveraging all the non-obvious incentives that naturally exist or could evolve.
Free software addresses a fairly limited universe of software, and often is bankrolled by some other monopoly or quasi monopoly. Where would web standards be without the trio of Google (advertising monopoly), Microsoft (operating system monopoly), and Apple (brand-marketing driven fashion company)?
The value of the patent system is that it permits decoupling. Compare the Wintel monopoly to the decoupled ARM ecosystem. It wouldn’t be better if mobile chip R&D had been bankrolled by Google’s say search business or Apple’s cell phone business instead.
> The value of the patent system is that it permits
> decoupling. It wouldn’t be better if mobile chip R&D had
> been bankrolled by Google’s say search business or Apple’s
> cell phone business instead.
I think your examples prove precisely the opposite. The development and emergence of smartphones were effectively bankrolled by Google and Apple. The mobile phone sector was stagnant while every vendor was busily pursuing IP monopolies (Wintel, Sun Java, etc). And in any event, ARM's intellectual property moat largely derives from copyright-like protections on its IC masks, not from patents. MIPS was also a contender early on, but importantly ARM won because it had strong SoC designs ready-to-go.
Your argument is a variant of the argument that by creating a transferable property right, we reduce the transaction costs of negotiating access to inventions. My point is that those transaction costs are dramatically overestimated (don't confuse complex or opaque for costly), and in any event far exceeded by the unintended costs of a patent system.
Importantly, when it comes to complex technology a patent no longer embodies the requisite know-how. People and companies don't lose much by giving away their secret sauce because very often the only people capable of applying and productizing it in a timely enough fashion to capture the lion's share of profits are the inventors themselves. Moreover, the best way for a market to discover who is best at leveraging an invention--if not the inventors--is by making it more freely accessible.
Also, the shift to SaaS and cloud computing means that complex systems never leave the control of their creators, anyhow. This is happening despite the fact that the patent system is supposed to incentivize disclosure.[1] Worse, the patent system makes it easier to stop competitors without substantial disclosure of the end-to-end systems. You can patent all the small sub-inventions, permitting you to block competitors, without every having to disclose the larger, integrated whole that makes it functionally useful in the market.
Even if the patent regime ensures that we get an MPEG codec that is 10% more efficient (or arrives 10% sooner) than an equivalent codec that would arrive in the absence of patent protections, is that worth allowing MPEG rights holders to effectively extort money from smartphone makers?
Because that's what we're talking about here. In this day and age, patents don't make innovation commercially viable; at best patents making them viable a little sooner and a little better. But the _costs_ are enormous, and I would argue far outweigh any meager benefit.
[1] Similarly, film and TV producers pursue DRM despite copyright protections, and despite the fact that DRM seemingly provides little marginal extra profits. The transactional costs to Hollywood of negotiating DRM would seem huge. But either the costs are less than they seem on the outside, or they're already making such huge profits that they can afford to try to lock down the market even more. But copyright was created to reduce the transactional costs of negotiating a locked-down distribution system. If companies are pursuing DRM anyhow, we really should question our presumptions about the necessity fo government intervention. For copyright we probably need some minimal protections (weaker than we have), but we can probably forgo patents altogether, especially in the realm of software.
Again, just because it seems that a patent regime would make a market more efficient doesn't mean that it would. Our measurements of efficiency might bear no relation to what a free market needs to maximize benefits. Communists thought they could devise a more efficient market in grain production, but obviously they missed or misjudged a few necessary incentive structures that allowed for a self-sustaining market (including secondary markets, etc) that maximized output.
