> But what are the real trade-offs of edge vs. cloud?
Maintenance - having to fix a networking issue when the kids want to watch their cartoons and you're halfway making dinner. Or when you're away for the weekend and your partner can't connect to the photo server.
When you pay for an online service you're also paying for someone to fix things for you.
True, when it breaks. It's annoying. Netflix has its own issues as well.
My small setup with 1gbps internet tunneled via cloudflare is super handy to be honest. Jellyfin, Frigate works with over 2 months of uptime without issue so far.
There are too many glaring holes in this article for it to be taken seriously. It's simply not a complete argument if you discuss the UK's current economy and don't mention
1) Brexit
2) Green energy policies
Not so much green energy policies per se as the incoherency of banning onshore wind, due to .. NIMBY planning objections. And objections to the construction of pylons. And objections to the construction of new nuclear. And objections to the construction of batteries. And offshore electricity links.
The UK's energy crisis and its BANANA (build absolutely nothing anywhere near anyone) planning system are very linked.
(credit to the Economist some years ago for that acronym)
I didn't vote for Brexit, but the thing is it's not really helpful to re-rake the coals on that one - it knocked our GDP, some kind of common market with the EU would clearly be good for trade.. but to what end, we can't change this.. surely better to discuss things we can actually change?
May of 1943 was called Black May because so many U-Boats were sunk, 25% of the force.
The reason was a bunch of technologies coming together. New weapons like Hedgehog and new tactics. More escorts and better ones. The Mid-Atlantic gap was closed by B-24s, there were more escort carriers.
I thought the German subs used more sophisticated encryption and were more disciplined in operational procedure such that the subs were largely not decrypted by the Allies.
Supposedly the better question is why it took so long for an allied victory. Britain’s economic output was absolutely vast, and its colonies made it very powerful. Yet it still took the US industrial capacity coming online and a significant portion of the USSR’s population dying before victory.
It's a silly thing to say, because ship losses are largely driven by German submarines in the Atlantic and American submarines in the Pacific, which are about as disconnected as things get in WW2.
You can see clear trends in the data in each of those theatres, both of which turn drastically in the Allies favor later in the war. The main driver for that is American war materiel production, but there's also plenty of decisions in strategy, tactics, and weapons systems that are entirely different.
If you look at the graphs there is nothing like an 'inflection point' or qualitative change in the graph behaviour.
The Allied line goes slowly down and the Axis line goes slowly up. At one point they cross, but there's nothing particularly significant about that crossing point. Nothing happened in the month that they crossed other than the two numbers were equal momentarily.
Patent appeal for artificial neural network, the appellant argued that the ANN was self-trained and not directly programmed with detailed logical steps, thus, it shouldn't be considered a computer program under the exclusion.
I am not a patent lawyer but my understanding is that a patent requires that the application show that the invention is reproducible. And if it's reproducible then the logical steps must therefore be sufficiently detailed :/
Hmm, I read the article as explicitly calling out "clinical trials" (as referenced in the title and abstract) and it makes no reference research studies. I don't understand the distinction between "research studies" and "clinical trials", surely all research studies where an RCT is performed with real patients and real drugs is a clinical trial?
I meant “trials for research studies” as opposed to “trials for drug or device approval.”
The amount of record keeping and oversight of a drug approval trial is enormous (and as a consequence insanely expensive) — data handling, having disjoint groups at each stage handling and analyzing data, etc and detailed records of every manufacturing step — think ISO9000 on steroids.
Nobody would bother to go to that effort for a scientific exploration, nor should they. So the bar is much lower.
I am making no excuse for shoddy science! But it is quite unlikely for a licensed drug.
It feels as if there should be a PaaS alternative to hosting your own PyPI server (and I would gladly pay for one) but alternatives like AWS CodeArtifact require a lot of manual configuration. JFrog have the closest service that I can find, but you have to subscribe to their whole CI.
I was hoping that GitHub Packages would solve this issue (as they do for NPM, Docker etc.) but sadly this isn't going to happen.
Maintenance - having to fix a networking issue when the kids want to watch their cartoons and you're halfway making dinner. Or when you're away for the weekend and your partner can't connect to the photo server.
When you pay for an online service you're also paying for someone to fix things for you.