Posts by JetSetJim 2156 publicly visible posts • joined 4 Jul 2009
It's not automation, is just curve fitting. All an AI does is fit a curve to data points, and you kinda hope that the surface you end up with interpolates well for any other data you want to throw at it.
It suffers from the same problems as other curve fitting techniques with the added bonus that plotting and interpreting the surface you've ended up with is fiendishly tricky.
I doubt the aim of this is to actually improve anything, merely to add to the layers of control.
AI based polygraphs are going to be no better than a regular polygraph, and just adds to the theatre of security being put in place because we have to "think of the (illegal) immigrants", now that we're pulling up the European drawbridge and going all isolationist.
We can't be British (aka English) if we let Johnny & Jane Foreigner in to dilute our National Integrity(TM)
It's aimed at commercial developers, so suspect single sites will either not benefit as much, if at all.
Rather oddly, I built my house and attempted to get BTOR to install. Nearest pole in the neighbours front garden, although terminator on the next pole down the track. Had the order all prepped 3 months in advance (it was for copper), appointment made, and naturally it was a no-show. Rang to complain and they cancelled the order and said "we're only installing fibre to that postcode", even though the fibre area was on a different exchange, and the one I was to be connected to wasn't even "superfast".
At times, BTOR can't even wipe its own arse
IIRC, BTOR were installing fibre-cored copper on new builds - although I can't find the link ot the product now.
All developments *should* install decent amounts of ducting as art of the groundworks as that is obviously the cheapest time to do it. Unfortunately there is no value add to the price of the house in doing so, so developers don't bother.
Until the govmt regulate new-builds to support fibre easier, this will continue to be the case.
> To be fair to EE, they don't actually make the thing
To be fair to EE, they (should have) wrote the requirements for the thing and sent it out to manufacturers RFx teams, who wrote compliance statements against each line item and then a £ amount at the bottom of the report. One of the requirements should have been to do this properly, but I guess they either didn't think this was important to be mandatory, or they didn't think of it.
I suspect agile methodologies on a completely new hadware design don't quite cut the mustard - it's quite difficult and expensive to iterate forwards.
In which case, the next step once "development" has been completed is testing. I suspect they're going to be quite careful of putting this into the field until rigorous testing has been completed - they don't want to get burnt (literally and metaphorically) as they did before with the Note 7, particularly as this is likely to be part of a flagship product
> Potentially very useful, if the computer is able to record all the same relevant factors that a doctor considers, and it might be able to identify other factors that the doctors weren't aware of; but it isn't in any way intelligent.
Well, the article says there are 48 variables they think influence the result, and they have a data set of 17k samples. This data set needs to get partitioned into training and validation sets (80:20 is usually good), so you've actually got 13,600 training points. No description as to the coverage of the 48-dimensional space (although perhaps some of these dimensions are correlated, which can lead to a feature reduction), but I'd contend that this is not enough training data to get a decent result anyway.
In other words, more study/funding is needed. The best conclusion a research paper can reach :)
> Some companies regard the knowledge and experience you gain whilst working for that company as the property of that company. Say you know nothing about radio physics and a company hires you and trains you to develop radio devices, everything you learn about radio physics is that companies property and you cannot use it at any other company in the future.
Hmm - learning laws of nature and other public pieces of information? Probably not, although you should not be able to take course materials with you when you leave.
Learning proprietary methods/designs/algorithms developed for your employers products - well, they might be in your head, but it's protected information that you shouldn't have a right to disseminate.
After that, it's a tricky minefield. You may be asked to develop a competing product, in which case you may make design decisions that are influenced by the design work you did at a previous employer - perhaps without doing the work that justifies that decision (e.g. here are 2 ways of doing something, but I know from previous work that the first one is not very accurate, or costs a boat load). That seems like a murky area.
Saying that, the case as it stands should be simple to resolve by examining the legality of the clauses in the contract, and the dates on the patent filings.
But Huawei have lots of form for nicking IPR (not that that gives a free pass for nicking their IPR)
> how much do you earn a month for it to be "prohibitively expensive"? Clearly, if it were "prohibitely" expensive, it would prohibit enough "consumers" to stop consuming and this pricing wouldn't be sustainable? And you really NEED three streaming services to live a happy life?
It *is* prohibitively expensive for a lot of people - the price point the offerings are at clearly need to be more reasonable, as the expectation for digital media is that it is inherently not that valuable - perhaps in part because it's trivially easy to copy and distribute, but also perhaps in part because it's a new-ish distribution medium where in yester-year everything was broadcast for free.
In terms of "how many streaming services do you need", well, in a family household, there will typically be 3-5 people, each with their own interest. Let's go with stereotypes (YMMV): dad will want some sports channel plus The Walking Dead, mum might want lifestyle channels, a young kid will want the cartoon channels, an older kid might want Dr Who / Twilight / whatever's trendy right now. It's unusual that a single streaming package will offer all of them - add to that the ankle-biters clamours for either/both of Netflix and Amazon Prime to watch movies and other TV series. How many services of any type do you need to cover all the bases? In the UK at least you can get Dr Who for free via Autnie Beeb, but sports coverage is laughable without shilling to Mr Murdoch (although perhaps alternative providers may be available). And all this is on top of the ISP costs and TV license fee (which I am in favour of, btw). It all adds up.
The cheapest Sky package is £20/mo (no sports tho, that's £40/mo). Amazon Prime is £80/yr, and Netflix is £6/mo for 1xSD, but more likely to want at least the £8/mo package for HD. Add to that ISP charges of around £20/mo minimum (assuming you can find one with unlimited data without the *), plus the costs of phone contracts for all in the house. You can easily get up to around £150/month on telecoms bills
Hear hear - I'd be delighted if a govmt organisation could come up with a secure and connected syetem that allowed medical professionals to get the relevant information for you when they needed it. Unfortunately, a prerequisite for that is the govmt minister & associated flunkies having a scooby as to what they were dealing with - otherwise we're back in the hole dug for the previous incarnation of an NHS IT system.
We got trenching in our village as presumably BT wouldn't let them run overheads on their poles - they're mostly in verges, so probably cheap to install. Ironically I have a BT pole at the edge of my boundary, but OpenReach couldn't pull enough thumbs out of their collective arses to run a cable with over 3 months notice. Fortunately the amount of trenching required for fibre was within scope of the standard installation cost.
I seem to recall reading that OR had an overhead "cable" that had a fibre core that they could use on new deployments to make it easier to swap folks across - but I don't have any BT/OR stuff in my house.
With OpenReach saying it's "available" in Coventry, does that mean fibre runs past each home with a breakout pot? Or just that the exchange has a bank of fibre terminators (presumably with a juicy backhaul fibre link) and that they would then start trenching on receipt of an order? And what is the cost to install it from Openreach considering it's not yet considered part of the mandatory provision requirements for a house?
The smaller players like Gigaclear charge a similar amount to BT for a new connection, possibly a bit more for trenching the fibre to the house from the roadside (unless you DIY), and the cost of a decent fibre service with Gigaclear is a bit more than your BT DSL service (admitedly at a much higher line rate).
It still looks like Openreach are dragging their heels...
His employment contract would definitely assign all rights to Apple, but yes if he has any meaningful contribution to the patent his name should be on it, and he would have received a one-off bonus payment as compensation (typically done as a small payment for the idea, another (larger) payment for the filing, and then finally one for the grant). I can't see it getting invalidated, though, as all his IPR rights were assigned to Apple.
Seems like they were documenting ideas in that bug tracking tool - different companies have different methods. One I've worked at had a dedicated tool that would track the progress of an idea through comittee, drafting & filing. Another just used a few spreadsheets. The one I'm at has a very informal process, so I take care to keep records of stuff.
> That only works if the other side is paying a lawyer and will run out of cash to pay the lawyer with eventually. He's representing himself, so it shouldn't happen.
Well, until he runs out of cash to pay for food, then. Although if there is a modicum of truth in his allegations then I'd like to see him win.
Either way, he may be on to a loser. Depends on what he documented in the ticket he made, and what's written in the patent applications. If the methods used are different enough to what he documented then he won't get his name on them.
As to "unlawful termination of employment" and associated gubbins, US employment law seems to be more on the employers side mostly - not sure at what point on the spectrum California sits, but I'd say it doesn't bode all that well for him.
> I kind of expect organisations that are sensitive about this sort of thing start are going to start mandating in contracts that their data only be processed on PCs without Microsoft Office installed.
I suspect they're just going to get sold SharePoint, too, for a lot of cases (or OneDrive). Even if it's a "Special" SharePoint on locally sourced servers, not replicated back to Redmond.
> Maybe the article could get some more details?
OpenSignal don't publish much more in the way of detail. They collect "stats" from their app that is installed on myriad devices around the world. The app presumably collects RF and performance metrics much like other network performance monitoring apps (e.g. Root Metrics), perhaps even running their own in-app performance tests to test network speed, jitter and whatnot while collecting GPS. These are then rolled up into the metrics they publish.
Knowing a little about programming for Android, and that iOS is more restrictive, it seems unlikely that they are collecting information about YouTube & iPlayer performance, but are instead spoofing a video stream from a server to their app, collecting packet timings if at all possible. OpenSignal then infer some sort of score based on how much jitter there is and how the bandwidth varies during the test, possibly assuming a minimum required bitrate to support a video stream, perhaps adding in some magic to cover application buffering.
Not forgetting a metric without defined units beyond a percentage and a wishy-washy phrase about video quality & buffering.
At the end of the day, if the network can offer more than sufficient speed to serve a video, the buffering shouldn't be an issue as the s/w should be able to work far enough ahead. At the application layer, there's usually some rate adaptation stuff built into the s/w to optimise what the user sees based on the end-to-end connection characteristics.
I suspect they're fighting to form the beach-head in the domestic "software helper" market. Whether you need an AI to do it though, hmm...
I could see a market for something that does the following:
a) user does meal plan, software produces shopping list from it's defined recipes (whether user defined, or server defined + user tweaks is moot) and the known remaining items in storage
b) s/w brokers deal with supermarkets/other grocery vendors, arranging deliveries and shelf-life expectations for the items being delivered - ordering from multiple stores if necessary/desired.
c) meat-sack receives delivery and does the manual labour of putting it on shelves/into cold storage. Probably need some mechanism to update s/w with shelf lives - e.g. embed use by date in an RFID/barcode that is easy to scan as you unload
d) s/w can then (re-)organise what you're cooking on a daily basis to minimise what you throw away, as well as perhaps suggest additional recipes based on what's in your cupboards if you're running out of inspiration (perhaps where the AI comes in)
e) when cooking, app can then send simple commands to devices such as the oven to set the temperature correctly, and do the timing
Barriers to entry:
a lot of this is simple to do in your head, or on a piece of paper - personally I might find it useful for something to warn me that stuff is going off in the fridge, as I don't remember the dates very well, but the market for this might be small
Scanning stuff in so that the s/w knows everything you have is not going to be easy as it probably requires food suppliers to create and adhere to a single standard of labelling that is easy for the consumer to use. RFID seems the best option here as you can read it from a phone, but some items such as fruit/veg don't have any packaging, so a solution to this would be needed. Barcodes can already be done (e.g. various apps that monitor your food intake by scanning the barcodes), but I think that's far too clunky a solution at the moment.
Overall, it can probably be achieved. Not sure it needs to be integrated into your fridge/cooker though - it should be a standalone app that can interrogate/control *all* the different brands.
I'm sure someone will try and do it with Blockchain Technology (TM) next...
Oddly there's no detail on the various space sites that I can find, although there's an earlier article on RELL that is an external scanner that has been used to isolate leak locations before.
Also, there's this article on using sound to find leaks.
My guess:
Step 1: seal each compartment, monitor each compartment for pressure. This identifies the compartment the leak is in.
Step 2: Option 1: Float something really light and non-damaging to anything in the compartment around the walls of the target compartment. A leak should induce a slight current towards the hole. Follow this to the hole. I vaguely remember some S-F story that had balloons filled with gunk deployed in the event of a breach - they'd get sucked to the hole and then burst over it - the balloon material providing some of the seal, the gunk being the glue. Good enough for a temporary fix and to identify the area in need of attention.
Step 2: Option 2: If some sort of remote camera is present outside, look for the air venting out. Manual inspection to then find it on the inside.
> So either the magic money tree really does exist, or our government is hiding away in its shelter
Perhaps the tree does exist, and at a later point in time we'll be told of the massive amounts of money printed by the Bank of England to get us through the "initial rough patch*" that is the Brexit fallout.
(*) - May last several decades...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
