nav search
Data Centre Software Security DevOps Business Personal Tech Science Emergent Tech Bootnotes
BOFH
Lectures

* Posts by Michael Wojcik

5499 posts • joined 21 Dec 2007

You: 'Alexa, open Cortana.' Alexa: 'Who?'

Michael Wojcik
Silver badge

Re: Magnum, P.I.

I'd watch the hell out of that!

I'm not sure I would, but it does sound better than the actual reboot, coming 24 September to US televisions whose owners aren't paying enough attention to change the channel.

0
0

Microsoft's Chinese chatbot inspired by images to write poetry

Michael Wojcik
Silver badge

Re: Waste of time

Why do you think you know what the poet meant?

Why do you think that has some special relationship to interpretation?

0
0
Michael Wojcik
Silver badge

Re: not sure which is worse...

Leave us not forget J. Gordon Coogler.

(And special mention to John Lillison, unfortunately disqualified due to ontological deficiency. Oh, pointy birds.)

0
0
Michael Wojcik
Silver badge

Poetry often seems to leave the reader trying to work out what the poet meant.

You're welcome to do that, but that is very likely not how human communication actually works in general (it's the Intentional Fallacy; see the later Wittgenstein, Davidson, Rorty, etc for more satisfactory explanations of language).

And many poets, when discussing their own work or poetry in general, have disclaimed the notion of presenting intent-puzzles for the reader, or trying to convey some specific set of ideas intact. Take Auden, for example: "There are many definitions of what art is, but what I am convinced art is not

is self-expression. If I have an experience, it is not important because it is mine. It is important because it's worth writing about for other people,worth sharing with other people." Shared experiences do not have the same meaning for all participants.

0
0

Support for ageing key exchange crypto leaves VPNs open to attack

Michael Wojcik
Silver badge

Hanno Böck and co. just got a Pwnie for Bleichenbaching a bunch of TLS implementations. Maybe that's the last we'll see of that implementation mistake, but my guess is if people are still making it 20 years after publication, they'll continue to make it.

Oh, well. More and more are moving to ECC, which isn't vulnerable to Bleichenbach. And if current trends continue, in a decade or so we may be commonly using some form of PQC, possible RLWE variants. So there will be new and exciting implementation errors to exploit.

0
0

Foreshadow and Intel SGX software attestation: 'The whole trust model collapses'

Michael Wojcik
Silver badge

Re: And by "Trusted computing" MS meant "Trusted by corporate media providers"

What in the world does this have to do with Foreshadow / L1TF?

Incidentally, neither Microsoft nor the Trusted Computing Group invented the term "trusted computing".

It is documented. It just doesn't do what the document says it does at all.

What blatant lie in the Intel CPU documentation are you referring to? Care to provide a citation?

1
0
Michael Wojcik
Silver badge

Re: Intel only?

You do realize that the researchers interviewed here - those how discovered the 'Foreshadow' flow in *Intel* CPUs - are Israelis, right?

Some are. The University of Michigan is still in the US. I mean, I haven't checked recently, but it's just down the road and I think I'd've heard if it had gone missing.

1
0
Michael Wojcik
Silver badge

and this is why security by obscurity doesn't work

Security by obscurity does work, for some value of "work". It just falls under Kerckhoffs's principle: the obscured elements are part of the key, and as key material they're difficult to re-key and liable to exposure, so they make a poor key component. It's more accurate to say that security by obscurity is uneconomical.

That said, there's no "security by obscurity" at work in SGX attestation (or other SGX aspects), Foreshadow, or L1TF. The security components are all documented, as are most or all critical aspects of the hardware in question. The problem is that the security mechanisms do not cover a branch of the attack tree, and that branch is exploitable, undermining the security guarantees.

This is not security by obscurity. It's another common security problem: complex systems are very difficult to secure. They're difficult to reason about (as Yarom says in the article); they suffer from combinatorial explosion in their state spaces and attack surfaces; they have unexpected interactions and revenge effects; and they have more side channels.

1
0

Google bod wants cookies to crumble and be remade into something more secure

Michael Wojcik
Silver badge

Re: Store cookies on a blockchain

Needs more AI.

0
0
Michael Wojcik
Silver badge

Re: Makes no sense

Given the vast array of other potential tracking mechanisms, I'm not sure how much this matters in practice, except for honest sites that want to use it as a standard way for users to discard their prior relationship with the site.

0
0
Michael Wojcik
Silver badge

Re: Zero understanding of cookies

What happens on the server entirely depends on the application

Well, no. That depends on the server, and other components of the application's execution environment, such as frameworks, language runtimes, and application engines. And there could be middleboxes that record cookies, etc. There are many possible server-side components with access to cookie values.

0
0
Michael Wojcik
Silver badge

Re: Zero understanding of cookies

So a session identifying cookie is useless if the server doesn't store it, it's just a key into a server side data store.

If only. Perhaps you haven't seen the vast numbers of web applications that serialize server-side objects and send the serial representation as the cookie value? Then the server doesn't have to store anything, and you get fun "Marshalling1 Pickles" vulnerabilities in the bargain.

Since this class of vulnerabilities was popularized, some applications have moved away from using serialized objects as cookie values, and others have introduced mitigations like HMACs, signatures, or encryption into the cookies. But there are still plenty of offenders.

1 sic, from the title of the famous AppSecCali 2015 presentation.

0
0

Drama as boffins claim to reach the Holy Grail of superconductivity

Michael Wojcik
Silver badge

Re: Would have got away with it too

Yeah, if I wanted to mount a scam based on some imaginary energy technology, I'd go the E-Cat route. It seems to be easier to separate fools and money using the "trade secret" whiz than the "bogus science" one.

I'm hoping there might be some less-damning explanation - an honest mistake (an error in data handling, perhaps), compounded by an email forgery from someone who wants to believe in the supposed discovery but isn't actually connected to Thapa and Pandey. (I really doubt the results are genuine.) Even that would be bad; you really don't want to put something on arXiv without double-checking, and a history of sloppy work, particularly this prominent, could be enough to end a career. But that's probably just wishful thinking.

1
0

CADs and boffins get some ThinkPad love

Michael Wojcik
Silver badge

Re: The laptop of my future

1250g? I'd consider that "barely noticeable". But then I often walk around with two laptops in my shoulder bag or backpack. I really don't understand this reluctance to carry a few pounds of equipment and material. When I'm working on the house I have more than that in my toolbelt.

0
0

May the May update be with you: OpenSSL key sniffed from radio signal

Michael Wojcik
Silver badge

Re: Sidechannel attacks

most of the security critical code is not tested for constant time execution

Constant-time arithmetic for cryptographic code is standard practice these days, and has been for some years. Amateur implementations may lack it, but the grownups all take it into consideration. When a timing side channel is found in a major crypto implementation these days, it's a bug, not a failure to consider the problem.

Constant-time algorithms outside of crypto arithmetic are obviously far less common (and in many cases infeasible), but timing attacks outside crypto arithmetic are also rare. Obviously there are the various Spectre-class attacks, but they're not timing the victim application. There was the TENEX password-validation attack, but that was a very unusual case.

If it could be done from across the street then I would be more worried.

Sure, because no one ever gets their phone anywhere near an antenna, and we have no idea how to create directional antennas, or amplify signals.

1
0

Boffins blame meteorites for creating Earth's oldest rocks

Michael Wojcik
Silver badge

Re: OK, come on folks ...

Less than half of Lake Superior is Canadian territory, so it's the largest lake partly in Canada.

That said, Superior has more than twice the surface area of Great Bear, and I don't know precisely how much is Canadian (didn't find it in a quick search and I'm too lazy to do more). So there may yet be more Superior in Canada than there is Great Bear.

0
0

It's official: TLS 1.3 approved as standard while spies weep

Michael Wojcik
Silver badge

Re: That's right...

The protocol is different, but the cipher suites and certs are still the same.

The suites for TLSv1.3 are all different. There's no overlap in cipher suites between 1.2 and 1.3.

Now, many of the same algorithms are present in suites on both sides. But the suites themselves, including what algorithms and modes they specify, have changed. 1.3 also standardizes some algorithms that are not available in standard 1.2 suites, and removes a great many deprecated ones.

As for certificates, the bit Kieren quoted from Eric Rescorla isn't accurate. DSA keys are no longer supported, and so neither are DSA certificates. That affects pretty much only some US Federal users - few or no other people were using DSA. (ECDSA is still supported.) SHA1 is allowed as a certificate hash in TLSv1.2 (though of course many applications now reject such certificates, at least under some conditions); it's not in TLSv1.3.

It's impossible to break into. We haven't found a way in so we gave up.

No one is claiming that, least of all the NSA. Security researchers of all imaginable hat colors are trying, and will continue to try, to break TLSv1.3 and the protocols, algorithms, and primitives it uses. I have no idea why anyone would think otherwise.

0
0

US voting systems: Full of holes, loaded with pop music, and 'hacked' by an 11-year-old

Michael Wojcik
Silver badge

The problem isn't the voting machines it's the voters.

The voting machines cause problems with the voters.

Weaknesses in the voting equipment and process discourage voters, and reduce participation in the electorate. And that means most of the people voting are ideologues of one stripe or another, who will tend to favor single-issue candidates that focus on their particular fetish. That does not improve the quality of representation.

Democracies do better when more people who aren't certain of their convictions come to believe that their participation is important, spend at least a little time researching the candidates and issues, and then vote for what they believe is the best (or least bad) choice. That tends to reduce relative support for fanatics and demagogues, and encourage the election of politicians who make an effort to compromise.

2
0
Michael Wojcik
Silver badge

Re: how about

As Claptrap314 and Big John wrote, mark-sense ballots (which are typically "fill in the oval" or "connect the arrow" or the like), while not perfect, seem to be the best compromise. They can be machine-tabulated and also manually verified; they are easy for voters to use, provided the ballots are well designed.

Hand-counted paper ballots, lever machines, and mark-sense systems significantly outperformed the alternatives on fraction of residual votes (undervotes and misvotes) in the well-known MIT / Caltech Voting Technology Project study. Mark-sense has no statistically significant difference from lever in this respect. Mark-sense provides immediate visual confirmation of the vote for the voter, and unlike lever doesn't need a separate paper trail, since the voter is marking up a paper ballot.

There's some thought that voters are more comfortable with mark-sense ballots than with lever machines, thanks to school training, as Big John suggested. The MIT/Caltech paper mentions this.

It also notes that mark-sense optical scanning machines were somewhat more mechanically reliable. That may just be a fluke.

2
0
Michael Wojcik
Silver badge

easy to point fingers at the Republicans but Dems do not have clean hands

True. Like gerrymandering and tampering with the franchise, this is a game that is enjoyed by whatever party is in power.

2
0
Michael Wojcik
Silver badge

Re: Old joke!

They are in at least some states.

Yes. And this is an important factor in the problem domain - the US is far from a voting-system monoculture. It varies from state to state, and often from district to district within a state.

So on the one hand, it's difficult to hack enough votes to alter a Presidential election, or in many cases even one for the Senate or other office with a statewide electorate. The real payoffs are in House elections and others with smaller electorates, which are more likely to present a single target; and in general from sowing uncertainty and cynicism about the validity of the process.

On the other hand, that makes it that much harder to get all of the vulnerable machines replaced, because you have to persuade many sets of election officials and the legislatures that control their budgets.

2
0
Michael Wojcik
Silver badge

Re: Obligatory

planes that have someone trying to make them crash don't typically fair very well

"fare", not "fair". And actually they do generally fare pretty well, unless that person has access to the flight deck, which is much less common after 11 September 2001.

Nor, I think, would elevators that had a person knowledgeable about them trying to make them fall.

And that's even more difficult, even if you're knowledgeable.1 It's certainly more difficult than hacking one of these direly unsafe voting systems.

1It's quite straightforward to make an elevator that can't fall without massive compromise of the car or shaft. For example, put an asymmetric weight on the car, so that if it's not under tension from the lift mechanism, it will tilt to one side. Then, if that doesn't provide sufficient braking force on its own, add a ratchet to the side of the shaft that the bottom of the car will press against when it's tilted. There are other purely-mechanical safety mechanisms that achieve the same result.

2
0

Second-hand connected car data drama could be a GDPR minefield

Michael Wojcik
Silver badge

Re: Spot on sir.

Drove my rear wheel drive through the hills in the snow and only had to dig myself out once.

My parents lived on top of a mountain in rural Vermont for nine years, with a FWD Toyota Tercel and a RWD Toyota Van (the Toyota R20 or R30, sold in the US in the 1980s). There were a couple of miles of gravel roads, which were occasionally graded but nothing more, before you reached pavement; and while Vermont is pretty good about clearing snow, there are a lot of road miles relative to population, so when it snows (as it does most days of the winter) it may take a while.

They drove pretty much daily and rarely got stuck with either vehicle. And that was generally with all-season (M+S) radial tires, not snow tires.

It's mostly a matter of knowing what you're doing.

When I taught my stepdaughter to drive, in my Honda Civic coupe (manual transmission, naturally), we had a nice snowy day so I drove around to the alley in back, ran it into a snow drift, then let her get it back out and back to the street. Took a while, but since then she's had no problems with 2WD vehicles in the snow.

When I was in high school, living on the New England coast, almost no one I knew had 4WD or AWD. We kids mostly drove dreadful old American RWD cars with open differentials, primitive suspensions, vague steering, and ridiculous moments of inertia. And we flung 'em around the roads regardless of the weather, generally pretty successfully.

These days everyone in the family has AWD vehicles because they're so common, why not? And it does make things easier. But necessary? I don't think so.

0
0

Reckon you deserve a Wikipedia entry? Try getting this bot's notice

Michael Wojcik
Silver badge

Re: Someday, machine spewed cr9p is all there will be.

easy to spot

Only when it's poorly done. Look at some of Bing Liu's work on machine-generated product reviews, for example; with a number of the systems in use, no one's yet identified any features that distinguish them from the reviews produced by humans with significantly better than random accuracy.

And this arms race has been going on for about as long as we've had online product reviews. It's an established area of R&D.

0
0
Michael Wojcik
Silver badge

I've a cousin (non-IT) who keeps bugging me on this, he seems to think computers should be able to write articles for him so he can put together a wildly successful webpage and retire to play videogames.

Phillip Parker beat him to it years ago.

Machine generation of specialist monographs is a solved problem. The results are quite dry, of course, but the audience for these things isn't looking for a page-turner.

Machine generation of readable fiction is still pretty primitive, as far as I've seen, but it's coming along slowly.

(Machine-generated prose literature would actually be a fascinating area to work in. Some other arts, notably music, have long had machine-generated work that fools expert judges. Literature has been tougher, but that just makes it more interesting. I think a really successful system would be heterogeneous and use a consensus mechanism, possibly a GAN-style architecture, on top of competing contribution algorithms, drawing on research areas and non-formal models such as Rhetorical Structure Theory and narrative theory.)

0
0

'Can you just pop in to the office and hit the power button?' 'Not really... the G8 is on'

Michael Wojcik
Silver badge

Re: Just a beer?

The man who invented the word was not a kid.

Your source?

It has a nearly twenty year pedigree, while emoticon as a word only precedes it by a few years.

A trivial check shows emoji has been in use for at least 50 years. There's no substantive difference between emoji in the sense of "pictogram in a pictographic written language" and "Unicode character that's an icon of some sort", particularly since emoji has also been used to refer to other sorts of icons in computing contexts.

0
1
Michael Wojcik
Silver badge

Re: Just a beer?

This is your regularly scheduled reminder that some of us millennials will be forty in a couple of years, so using the word as a short-hand for 'bloody kids' is increasingly inaccurate.

Bloody kids are turning forty now. Always repeating history.

2
0
Michael Wojcik
Silver badge

Re: Just a beer?

Nah, emoji is an nipponism (or a japanese neologism if you want to be picky about it).

There are kids in Japan, and Japanese kids are certainly capable of being ignorant of history. And we can also still assign blame for the spread of the term to other languages.

While complaining about one neologism replacing another is an exercise in futility, the Japanese origin of emoji is not evidence against jake's claim.

1
0

Imagine Python fan fiction written in C, read with a Lisp: Code lingo Nim gets cash injection

Michael Wojcik
Silver badge

Re: Interesting but ugly

https://github.com/nim-lang/Nim/wiki/Unofficial-FAQ#why-is-it-caseunderscore-insensitive

Well, this gets my vote for Unconvincing Argument of the Day. "Case insensitivity is widely considered to be more user friendly"? Wave any harder and you'll break your wrist.

6
0
Michael Wojcik
Silver badge

Re: Indentation-significant syntax

I have mixed feelings about indentation for structure, but I can live with it.

I find other things to dislike about Nim's syntax. In particular, the developers are in love with obscure and non-intuitive symbols and constructs. Why use "*" as a suffix to mean "exported" (which is apparently equivalent to "public")? The dire character shortage has ended; we can now use actual words in our code. And the embedded linker directives in one of their examples:

{.passL: "-lsfml-graphics -lsfml-system".}

Besides being obviously non-portable, that's ... well, simply vile.

It's nice Nim can compile to C or Javascript. It's nice that it offers a variety of GCs with different performance characteristics. It's kind of nice that it has macros, though that's a double-edged sword too; you really want to rein in your developers on that one. But this is not a language I find appealing.

1
0

Oi, clickbait cop bot, jam this in your neural net: Hot new AI threatens to DESTROY web journos

Michael Wojcik
Silver badge

Heh. In all seriousness, though, it's generally quite difficult to explain what features have been developed by unsupervised training of a convolutional neural network or LSTM network (which is a type of recurrent neural network that has a relatively complex state).

It's quite likely that there are various "combinations of words" which result in a high clickbait score in some headlines and a low score in others.

1
0

Facebook insists it has 'no plans' to exploit your personal banking info for ads – just as we have 'no plans' to trust it

Michael Wojcik
Silver badge

Jeez, guys, we just want to chat

it just wants to connect bank customers to their bank's chat accounts and give useful financial updates

And there's nothing that could go wrong with that. Great to see Facebook inventing exciting new phishing vectors.

In other news, Facebook has been contacting Nigerian princes with large sums of cash to relocate...

0
0
Michael Wojcik
Silver badge

Re: No wiggleroom to be given

United airlines and Hertz car rental are two international companies that no longer even bother to reply to emailed or web-from submitted complaints or concerns.

To be fair, United's standard way of handling complaints is to have some of their goons rough you up, so it's not clear this is any worse.

(Killing your pet is a premium service.)

4
0

Hey, you know what a popular medical record system doesn't need? 23 security vulnerabilities

Michael Wojcik
Silver badge

They claimed it's "the most popular" system of its type.

I'm thinking those organizations using it are now on the line for some nifty HIPPA violations if they don't patch mighty quickly.

1
1
Michael Wojcik
Silver badge

Re: Why are we still seeing...

Yes, you know this means they're constructing ad hoc queries using string concatenation and interpolation. We need to start assigning liability for antipatterns like that. There is really no excuse.

4
1

Funnily enough, no, infosec bods aren't mad keen on W. Virginia's vote-by-phone-app plan

Michael Wojcik
Silver badge

Ah, yes, more barriers

If, for any reason, the voter falls off the voter registration rolls, the jurisdiction will no longer send a mobile ballot and the voter must restart the process of registration and authentication.

I believe the Voatz shill meant: "If, in order to promote the party of their choice, election officials remove the voter from the voter registration rolls, our system will have imposed additional hoops to be jumped through in order to be restored. And there won't be any provisional ballot, so if the voter only finds out on election day, tough luck."

Everyone involved in this firm needs court-ordered remedial training in civics. (Where's Julie Robinson when we need her?)

1
0
Michael Wojcik
Silver badge

Re: There is so much to be wary of here . . .

Once I got to the point where the name of the app was given (Voatz)...

Agreed. I'd like to see them banned everywhere merely for the crime of horrible branding.

7
0
Michael Wojcik
Silver badge

Re: Could I vote?

I'll bet these lazy asses will vote for whichever candidate pays them to stay on that couch.

And the great thing about voting from home is that a third party can confirm how you voted, and reward you appropriately.

3
0
Michael Wojcik
Silver badge

Re: You'd think that by now anyone building an "online voting system"...

No one is anymore stupid enough anymore (I hope) to trust anything else.

Apparently many election officials are. Or, equally likely, they're mendacious enough to pretend they trust a closed system, for which they're getting kickbacks or some other quae pro quibus.

3
0

NSA's crummy crypto crop Suite B binned, and other network nuggets

Michael Wojcik
Silver badge

CNSA did make some changes, notably reclassifying RSA and other algorithms so they were still allowed, and retreating from PFS. It's significantly different from Suite B.

But, yes, the article is rather misleading. Most of the algorithms endorsed by Suite B and CNSA are not "NSA specs". And the NSA's "run away from ECC!" panic that led to the replacement of Suite B was largely ignored (for practical purposes, even if it excited a lot of discussion) outside the Federal government.

It's also hard to see how "[m]oving the RFCs to historical status formalises the death of the suite". Suite B is a NIST specification. All the IEFT have done is updated their endorsement to bring it into line with NIST.

1
0

AI on Raspberry Pi, Waymo touts robo-rides to Arizonians, and more

Michael Wojcik
Silver badge

Re: If it comes from any mainstream media

Israeli intervention in U.S. elections vastly overwhelms anything the Russians may have done

I'll note that Chomsky here speaks of "intervention", not "hacking", as DCFusor put it. That's a very different matter. The bulk of Israeli influence on US politics, as far as I can see, employs traditional means: money, favors, and so forth. There may well be IT dirty tricks, because what nation state doesn't employ every means at its disposal in pursuit of its foreign-policy objectives? But whether that particular aspect saw a greater investment from Israel than from Russia is dubious, and certainly not what Chomsky claimed above.

0
0
Michael Wojcik
Silver badge

Re: If it comes from any mainstream media

You can make six figures now and not be able to afford a home, much less kids and college, in any town you can make that.

My household brings in six figures, and affords two homes, in two towns, in two states, where, indeed, we make that.

Also put a kid through college - in fact paid for an assortment of degrees (we each have three, as that's a nice round number).

It can be done. It helps to live in a place where the cost of living is relatively low, certainly; and having skills that are in high demand helps enable that. In the region around my northern home, good houses can be had for under $100K (ten years ago, they could be had for under half that). In the area where I have my southern house, prices for something decent are steeper, but then energy costs are lower, which balances things out a bit.

0
0

Uptight robots that suddenly beg to stay alive are less likely to be switched off by humans

Michael Wojcik
Silver badge

Re: IMHO obvious why

Appliances don't need to be constrained

Counterpoint: IoT

2
0

Click this link and you can get The Register banned in China

Michael Wojcik
Silver badge

Re: Travesty...

I wouldn't have expected otherwise.

Yeah, their Little Mermaid isn't exactly faithful to the source material either. Ditto anything that isn't Disney ab initio. Disney is not in the business of maintaining any tradition other than their own.

0
0
Michael Wojcik
Silver badge

Re: whatever’s wrong ...

I truly wonder if Xi doesn’t realize how ridiculous and Barbara Streisandy the Winnie censorship makes him look.

That may be the point. If you implement plenty of petty tyrannies, many people will get distracted from the real ones. If some of the clever kids spend their efforts bootlegging foreign films, they won't be putting them into real resistance.

I have a grudging respect for Xi. As autocrats go, he seems to have a very nuanced and careful grasp of power. His cult of personality seems to be extremely managed, not the sort of wild-eyed populism we see with Trump, for example, or even the still rather visceral "hard man is good to find" thing Putin promotes. His repressions also strike me as calculated, and often deliberately obtuse, to present a sort of circus of ineffectual resistance. Let the people enjoy inconsequential defiance as entertainment.

(Of course, this is by no means a novel technique, but as far as I'm aware it was mostly practiced systematically in pre-capitalist economies, as shown in e.g. Stallybrass & White's The Politics and Poetics of Transgression. There was a bit of it with the US Hippie movement, which was notable for achieving nothing but the neutering of a potential political opposition,1 but I suspect most of the US establishment genuinely loathed the hippies and it didn't occur to them to use the movement this way.)

1Unlike the actual civil-rights movements in the US, which did achieve significant change.

1
1

Dear alt-right morons and other miscreants: Disrupt DEF CON, and the goons will 'ave you

Michael Wojcik
Silver badge

Re: Excellent

When I see a certain race or religion being hearded into camps and murdered like Christians in the middle East the word nazi springs to mind

Even then, the Nazis hardly had a monopoly on genocide. I'm certainly no fan of the contemporary hard-right political movements, but I agree that the "Nazi" label is overused and distorting.

The more general "fascist" is also often incorrect, since many of these people don't appear to be supporting a particularly fascist system; for example the ones I've spoken with aren't in favor of mixed economies - they're laissez-faire liberals, or imagine themselves to be, in economic matters. While they may advocate some of the attributes associated with fascism, such as economic autarky, and may explicitly or implicitly endorse autocratic leaders and the like, there are innumerable variations on political systems, and a right-wing ultranationalism isn't automatically fascism.

While there's sometimes some short-term rhetorical advantage to throwing around those historical labels, in the long run I think we do better to try for a bit more precision.

3
1
Michael Wojcik
Silver badge

Re: False flag attack

Why would anyone want to be such an idiot, annoying everyone, making asses of themselves

... by trumpeting "false flag!" at every news story that paints alt-right idiots in a bad light? A severe lack of critical thinking skills, I suspect.

4
0
Michael Wojcik
Silver badge

Re: Irony

Don't conflate anarchists (wanting less centralised- /more self- government) with chaosists (wanting no rules anywhere)

You can invent your own definitions all day long, but anarchy literally means "without leader". It's a straight transcription of the Greek ἀναρχία. That's not "more self-government"; it's "no government at all". By definition.

(And "chaosists" is impressively cacophonic; no one with any sense of prose style would employ it. Seriously, ugh. For the love of English, "chaotists" at the very least.)

4
0

Boffins build a NAZI AI – wait, let's check that... OK, it's a grammar nazi

Michael Wojcik
Silver badge

Re: Commas (missing or extraneous) can change meaning

So how would the "AI" be able to determine the original intent?

Thus, I call (partial) "AI BS" on this one.

Argh. Did you read the article?

The point of punctuation-replacement systems, such as this one, is to take text streams that lack punctuation (such as from ASR) and attempt to inject appropriate punctuation to improve parsing.

"Determin[ing] the original intent" isn't the goal. Yes, there are ambiguous phrases in natural languages. Every single person who works on natural language processing is aware of that, and the many people who commented on this article to point it out may congratulate themselves on having made what might be the most obvious point conceivable.

A punctuation-replacement system is a model-based transformation. That model could, in principle, be extremely sophisticated. It could build competing parse trees and select among them based on sentiment, metadata, rhetorical structure analysis, or other secondary features. It could keep whole-document context. It could use a world model to determine probable meaning of text segments. Or it could just be something like an LSTM network (or even something simpler) trained on a large corpus.

But it won't "determine the original intent", any more than human readers do. That's the intentional fallacy. Writers (or speakers) don't transmit their intent through language to readers (or listeners). Readers construct interpretations, which will correspond to some degree with the writer's interpretation.

And there's no "BS" here. Shan claims the system injects punctuation with an F1 of around 0.7, if my memory of the article serves (I'm not going back to check because the details don't really matter). That's the claim: it's a specific one, about the measured output of the system run against a particular set of inputs, compared to the ideal output.

1
0
Michael Wojcik
Silver badge

Re: Am I missing something ?

We already have grammar checkers built in to stuff things like Word up totally into American English.

While we're at it, they're not "grammar checkers". They're systems that apply a bunch of mostly-inappropriate heuristics, nearly all concerned with usage, mechanics, diction, and other things which are not grammar, to prose which has been mechanically chunked but not actually parsed.

Those things may help marginal writers massage prose into something closer to someone's idea of preferred form, but they are by no means a substitute for learning to write well.

And they have nothing whatsoever to do with the system described in this article. Really, I can't imagine how the OP thought they're at all relevant.

1
0

The Register - Independent news and views for the tech community. Part of Situation Publishing