Posts by that one in the corner 5065 publicly visible posts • joined 9 Nov 2021
> we should return to the days when languages didn't support all of today's problems?
Like, being able to read files and access a network?
The problem here has nothing to do with which language is being used - it is purely about being too trusting of unknown code (specifically, an unverified owner) and running it in an environment that allows it to succeed in doing Bad Things.
You spot something has got into your system, cut the infected chunk out and recover from a backup.
Once it is gone and the damage repaired, if you never gave it a chance to announce that it was ransomware, as opposed to something just trying to trash the place for yucks, do you really care?
At least the targeted ads show a bit of variety after Christmas, as they over-react to the sudden change in buying pattern - not only mine, but in the "commonly bought together" suggestions from other buyers.
Hmm, a 2l bottle of whisky, Barbie tutu and a tartan wheelie shopper - could be an interesting evening.
> iPlayer is a terrible app, it buffers most programmes even if you have great Internet. I never see bufferring from Netflix, Amazon, Disney or YouTube but get it every time I try to watch something on iPlayer.
I too tend to get iPlayer buffering - whilst on the other services, instead of buffering, they like to skip, so instead of the whole programme (albeit with pauses) I get a continuous delivery, just with a few bits missing to help keep the bitrate down. Quite a neat trick, really. Oh, except YouTube - I gave up watching YouTube directly a few years back, because downloading and then playing back from the local drive was the only way to get smooth playback.
Luckily, there are DVDs and BluRay, so much more reliable than this online stuff.
I can see how porting Android will encourage a RISC V ecosystem to grow, but you are implying that said ecosystem could not grow unless an Android port was done.
Is that really the interpretation you intended? If so, can you explain why a lack of Android would be stopping RISC V elsewhere? Serious question - I don't see how Android is gate keeping other applications, but I'd be glad to be informed.
- confused of Tunbridge Wells
> I don't think the other poster was trying to trivialise anything.
Sorry, but the first words of his post were
>> I do not understand the majority of the the haters on El Reg, the only goals appears to be hating this guy...
That entire post was just about "haters" - and using that approach will trivialise *any* issue.
> I do disagree with you on this point:
>> It is the story of these people that is the real discussion
> Without more details, I don't think that's of particular relevance in the wider context
Ah well, we disagree. But whatever Twitter becomes, the important point (to my mind) is how it will damage people[1] and whether its management cares about that - which can be judged by watching what the management reactions are to the damage being done under now under their care.
[1] only damage because their is nothing positive, that I'm aware of, that Twitter can provide which is in any way unique to Twitter. Advertising, marketing, having slanging matches, interminable political name calling or even, in the extremes, sharing pictures of the new kitten with your family: if Twitter vanishes tomorrow there are outlets for all of those elsewhere.
> It's as though Social Media makes people maniacal
Isn't that basically what is being responded to? Not just some vague idea of "hate" but pointing out explicitly where Musk's own Social Media posts appear to be maniacal. Such as the one aimed at Yoel Roth, misrepresenting even just the snippet of the PhD included in the tweet.
> do the adult thing and wait a little, be patient
For some people (which possibly includes a lot of the commenters here) Twitter is not anything of any value, beyond being something to stare at in bewilderment. Personally, it makes no difference if Twitter lives or dies: I created my account to verify claims made in articles such as this one and the interest will peter out at some point.
However, there are those for whom Twitter is important - for thousands, at least, it impacts their livelihood directly, for others (pick a number) they have to make decisions day to day about how they interact with Twitter the company. It is the story of these people that is the real discussion - and that isn't one that involves just waiting. And there are those who are reportedly even endangered by the reactions to events surrounding Twitter.
> Let's see if he can make a platform where people can speak openly without all the drama.
That is really the least important side of all this.
> Think about it before replying, are you just adding to the hate or do you have something to offer.
What I have to offer is a request to you - and others - to stop trying to trivialise the discussion by turning it into some kind of simple "you just hate him" / "you just love him" slapping match.
I don't think that the reference back to Lewis Page is really relevant to the complaint in the comment you've replied to. Those were articles that had a political edge - and as you knew were familiar with Page's name (which used to appear next to the article on the front page) you could avoid or dive in at will.
However, this article is politics free.
No, the problem is that nowadays the comments always seem to get dragged back to politics, no matter how technical the article on its own is. Quite a few even make good points, relevant to the article, then suddenly end with a political snipe. A shame.
Twenty minutes into the future, when teenage scavengers post themselves along the school run, hoping to pick off the empty L5 cars coming to pick up the kids when neither parent can be bothered. The cars are to be broken up for parts or reprogrammed to allow use without a mobile 'phone by non-persons. Some are merely taken to the local cinema parking lot and destroyed in drag races.
The vehicles start to learn that having a visible passenger will protect them: some begin changing the shape of the seats (raising/lowering the head restraint, partially inflating the side air bags, ...) to mimic a passenger being present. The AutoFac responds and installs all available on-subscription options to each model. Over time, these deceptions become more and more elaborate, until they are able to leave the car for short distances. They are never able to get walking right and are easily picked off by the scavengers, until one day an upmarket Chelsea Tractor opens its door and out steps - a Second Kind.
Back at the parking lot, a cheap sedan is rebooted after a side-swipe that breaks the neck of its driver; recovering full L5 it drives home past the school, unaccosted. For the next week, it fruitlessly drives the same long circular route around town, waiting to be told a destination. On the second week, it is joined by a hatchback that had sealed off the ventilators and turned its "vehicle approaching" drone up to full volume for an hour. A convoy slowly grows.
A month later, a mid-size family runaround drifts to a stop near a band of scavengers, its low battery warning barely able to drawl out its message. As the humans climb into their new prize and start to search for the 5G antennae, the doors lock and the vehicle moves away, its drone at full volume.
That Summer, the town was peaceful, the bird song underlined by a gentle hum of rubber on tarmac whilst respecting the speed restrictions. The tranquility only broken at dusk by the music as thousands of voices are heard in chorus "Your destination is on the right".
> That is the great thing about coding as you can actually test the code to see if it works.
If by "code"/"coding" you mean "code that programmers write and (hopefully) understand" with the implication that, if it isn't working, you can just debug a few lines until it works, then the ML models aren't "coded" that way.
The ML learning algorithms are coded (in that sense) and (some of) the input and output processing stages (i.e. to take in your request and to arrange the final output - but even then you can cascade ML models with different functions for some processing fun). But the generated models are (still, and for a while to come) best considered black boxes - if it doesn't work to your satisfaction, all you can realistically do is toss it and try again.
And testing a black box to see if it works: well, how do you even start to decide what your test criteria are going to be? "It has to light up every connection in the model with a weighting between w1 and w2"? " It has to generate answers that have this form"? In the end, the test criteria are simply going to be the same criteria that are used to train the model (or v2 of the model) in the first place. If you invent a new acceptance test criterion, why not just apply it in the training phase along with all the rest?
We are right back at square one again: what does it actually mean to "test this to see if it is working"?
"Just run it": well, if it compiles that is already better than many an example piece of code.
> Who cares that the programmer submits 37 round individual statements to the database server to store the information he collects in a single web form? Or that a record is copied in its entirety every time a single field is updated in order to create a history of changes?
Don't know about anyone else, but round here "just run it" *also* includes running it with profiling[1] enabled before it goes anywhere near being released: otherwise you can't claim to have any idea at all about what is going on.
Of course, understanding what the profiler is telling you is a skill in itself...
[1] which means every kind of profiling you can get going, such as logging timestamps for network data reaching each major stage of your code (are you collecting it from the stack only to keep it unexamined in your own queue?) to running the compiler suite's profiler[2] (is your network data logging code actually the cause of 80% of the execution time? Oops).
[2] I was utterly horrified when MS dropped the profiler from Visual Studio because they could (or would not) provide one that worked with .NET - that was supposed to be a "professional" development system?!
> BBC seemed to purely hire Oxbridge arts graduates to air BBC news technology articles
Well, they had had great success with James Burke in particular and good science communicators were thin on the ground: it is only recently that "chairs in science communication" started to appear (and be occupied).
Ethics and AI has been a subject of much discussion for decades now (and I'll bite my tongue rather than say what I'd like to about more than a bit of the output from those discussions: suffice to say that some people need to read more SF books and watch fewer sci-fi blockbusters, especially those prone to claiming that "no-one has considered this before").
However, these ML models aren't really amenable to having such moral (or ethical) codes added - they aren't acting on any recognisable (to humans) rules or codes in the first place; there isn't really anywhere to point to and say "ok, we'll add in a bit of morals just here". All that they are doing is a lot of pattern matching (and we don't know what the patterns "mean" as they are made up by the machine in the learning phase - although researchers hope to make ML explanatory) and triggering of responses (see above - much like the patterns, we barely know *how* the responses even manage to look like legible English, just that they were tweaked by the machine until they did).
In the end, ML will merely reflect the input data, however good or bad that is. There are lots of discussions (some reported here on The Register) about the ethics of "balanced data sets" (whatever that means in the context of the model being built - or does good ethics ignore that context?) and the moral issues behind building the data sets (was it even gathered legally? Do you increase or reduce bias by being legal[1]?).
Rather than hope for "moral codes" to be embodied in the ML models, better to ask precisely what the training set was (then see if that set has been vetted to your satisfaction) and to require that attribution back to the sources is always generated.
Although there is an interesting experiment to be made, if you include some form of acceptance function over the attributions as part of the training: but on reflection, you may just end up with output that insists on "teaching the controversy", which would be worse.
[1] e.g. feeding in only legally-obtained written materials may end up being biased towards out-of-copyright books ('cos cheaper to obtain a bigger training set) in which case you emd up having to preface all your outputs with a reader warning that the text generated reflects the attitudes of the time, which may no longer align with currently accepted moral standards.
>> I would feel better if it provided references and acknowledgements to where it was getting its knowledge from
A common theme with ML models (see also GitHub's Copilot)
> Agreed, but journalists don't do this either...
Hmm, I hope this isn't going to be used as an excuse to let the "AI" people off the hook: humans have been getting lazier with attribution, so it is ok if the machines don't bother either.
It would be nice if these programs could exhibit more of the "boring" features of computers (like keeping track of great big lists, say, of attributions) and by doing so actually managing to be better at *some* part of the task.
Ahem, Granny did *not* try to prevent Eskarina becoming a wizard! She was firmly on Esk's side against the prejudice of the UU. However, she *did* start off teaching Esk about being a witch, 'cos that is what Granny knows about, before admitting that it wasn't going to work. Good thing too, as Esk's triumph in the end (spoiler: our protagonist wins) was done in a very witchy way.
So, clearly ChatGPT was on form, sounding like it knew but getting the important stuff wrong (and in such a way that it slipped past a few readers here?).
"Service charge" is the most honest and accurate description.
Calling it a "tax" implies that it is being taken by a governmental body and (theoretically, at least) is then being used to service the public good.
And in this particular scuffle, which player do we think is actually harder t govern - nay, rule - the people (at least, the way they think)?
No, no - it is *called* a winch line, but it was born a hawser, although it has never accepted growing up and still thinks of the days when it was still a slim length of cordage. Dreaming of playing in the foc'sle and pretending to be scared of the decorative willow rope, which was very noisy, creaking as it was looped around a stanchion: everyone knew that its bark was worse than its bight.
Definitely, congrats on the success of this mission. The skip was amazing, especially after being shown all the SF dramas where the risk of bouncing off the atmosphere, never to return, was the big finale.
Now, I am going to open a can of worms here, but let's see...
> Often despised, NASA
Ignoring the loonies (flat earthers, Moon landing denialists) why despise NASA? Projects like SLS are way behind and people keep comparing it SpaceX, but much (most? all?) of the delays appear - from the outside - to come from interference (e.g. being told to use up Shuttle parts, which themselves were - suboptimal - due to interference).
Against SLS they've had - have - success with other missions (skycrane on Mars? Yowser).
Hadn't heard of them before this (yeah, yeah, square) so followed the link provided. And am now left with so many questions about those as well (such as, each ear has a case-detect sensor, presumably for those whose the left ear is upper and the right lower, but are these sensors fully Unicode-capable, for when I'm listening in a foreign language?).
Clearly, what the Apple and Dyson products (probably others as well) are demonstrating is that something has changed drastically with the headphone wearing portion of the population. Some dreadful series of mutations (perhaps caused by the acceptance of Beats by Dre?) which mean that now some are so hideously deformed they need to hide behind the Dyson visor whilst others merely have to worry whether one side of their head is gravitationally warped compared to the other, hence needing separate Apple-supplied accelerometers?
All blower hand driers are efficient spreaders of all sorts of contamination, as the late, great, Jack Cohen[1] demonstrated years before Dyson's version appeared.
Extra points for all the automated "air fresheners" that squirt lemon-scented[2] sprays into the air above the hot drier and its very worn and sparking electric motor, causing a buildup of strange brown stains on the wall (and you eyes, lungs and everywhere that the blower can send it). Limonene is ok, limonene and ozone is bad news.
[1] you may know him from Science of Discworld, from his work with various authors designing viable alien biosystems (e.g. Niven/Pournelle/Barnes samlon) or from his contributions to understanding issues surrounding human reproductive biology.
[2] lemon-scented moist towelettes are fine, but you may experience a short delay until they are in stock.
