if you go back and actually read the original 'GMO gives rats cancer' paper, you find that the whole papers statistics rests on two female rats of a cancer-prone strain of rat getting cancer, when the control group did not.

Two rats is not significant.

There is a difference between no identifying information, and no information.
Rips of DVDs, for example, would be information - but they would not contain any identification other than the program used to make them, and the DVDs in question.

'This is where I think we get in trouble with forensic science. Certain things, like finger prints and DNA, can exonerate a suspect but we have seen enough analysis around here to know that it is a fallacy to think that these things prove guilt. it only proves guilt if we assume the probability of guilt is 100% initially. When comparing the sample to a database, random error can create a match under certain common circumstances.'

However, in this case, they were comparing not against a database of millions, but one of several possibilities.

one failure in 13 launches.

This relies on the GPL 'you can't distribute at all if you can't comply with the licence' - if you distribute with a binary blob (effectively) that you have no rights to in your allegedly GPL codebase - then you're in violation of the GPL.

You can then come into compliance with the GPL in principle by supplying source for that blob. It is possible that the only way that compliance could be done is to release the whole of minecraft under the GPL - but the impossibilty of that does not factor into the legality of distributing.

The only way you can be forced to release source code is for a judge to compel you to release source code, after a copyright suit.
They are probably more likely to assess financial damages - rather than compelling release of source code.
I'm not aware of any compulsory release of source code ever happening.

The cop cameras will presumably have evidential and tamper resistant design.
if you're concerned about off-camera and camera broken incidents - it would be interesting if you could also purchase one of them.

Every second of video GPS timestamped, and the GPS logs extracted and used to index the video.

Privacy is a hard topic.
To a degree - I find the fact a police officer, who could have arrested someone was on the scene - makes the case rather different from that of a random surveillance camera.

My starting point would be that all video from cameras while the cop is in a public place have a much, much, much lower threshold for access.

18 petabytes a year isn't much.
Taking the assumption above that there are 5000 cameras working at once.
They are paid around $35/hour. This would make the wage bill 1.5 billion. Budget is $5B - so this seems order of magnitude right.
18 petabytes, on amazon redshifts '$1000/tbyear' is only $18M.
It seems quite plausible to get that to $5m without trying really hard.

Perhaps more important than storage, is access.
It should be possible to say 'show me a list of officers and car cameras within 1000 yards of 1 WTC between 8Am and 9am last friday'.

And yes - this implies the cameras must have GPS too.

'Patents are supposed to be for inventions, not just "useful things".'

Quite.

Patents were originally meant to be a trade - you got protection in exchange for showing how your device worked that may take months or years to reinvent.

You should never, ever be able to get a patent for being the first to come up with a problem, and doing the obvious solution.

This is especially the case if that solution took you less time than a full and proper patent search to see if that solution was already patented.

If that is the case, the patent system is _completely_broken_.

I can see the argument for patents that have some extreme and brilliant novelty to them.
But patents that are basically obvious restatements of the problem that the designer was facing - followed by the obvious solution - should result in the applicant getting set on fire.

They are _a_ bottleneck.
lithium-ion batteries are quite adequate for 10 or 20 minute flights in a small drone.
3 minutes flight at 60km/h takes you 3km.
Taking for example - https://en.wikipedia.org/wiki/... - this needs around 60 drone stations to cover the entirety of London.

Each drone can do 10 deliveries an hour - 180 a day, 65000/year.
The network can do 4 million deliveries per year.
The cost per drone is perhaps pessimistically, 1000 pounds.
They need to survive a week or so to easily recoup their losses.

The shuttle had several very near accidents in its run.
NASAs current criteria for man-rating would - for astronauts that fly six times a year be safer at work than:
A) Deep sea fishermen
B) Lumberjacks
C) Librarians.

Hint: It's not the first two.
(actual figures from US statistics).

https://en.wikipedia.org/wiki/...
Eben Christopher Upton is a Technical Director and ASIC architect for Broadcom. He is also a founder and former trustee of the Raspberry Pi Foundation, and now CEO of the Raspberry PI trading company.[4] He is also responsible for the overall software and hardware architecture of the Raspberry Pi device.[5][6]

'One spectacular explosion' - One explosion would be 92% reliability. (one failure in 12 launches)

By a not too unreasonable extension of the theory that allows the judge to compel microsoft to deliver things they control on a computer in another country - I see no reason exactly the same would not apply to compelling them to deliver a personalised update to one particular computer and deliver access to that computer - wherever in the world it was, and whoever owned it.

SLS is not expensive because it's so damn big.
SLS is expensive because it's so damn expensive.

It has been a goal for many in the space community to hit $1000/lb for space launchers.
SLS will beat that.
Unfortunately - in the wrong way - by exceeding it for the cost of the actual fuelled rocket on the ground.
(At the flight rates that NASA is projecting - on the high end of likely for the first several flights).

For the cost of the SLS program up to first launch, you can lift around 5500 tons to LED - using the published per-flight cost of Falcon Heavy.