* Posts by Mathman

22 publicly visible posts • joined 7 Dec 2016

Alarm raised after Microsoft wins data-encoding patent


How to handle patents

The practical approach to handle these kind of situations is:

If you are aware of prior art that invalidates the patent, then ignore it and continue with your implementation. If the lawyers send a letter then you reply by pointing out the prior art. They will need to respond by defending the novelty of their patent and if it is clear cut, will avoid taking it further and having their patent overturned. This simply requires an exchange of carefully worded letters and need not be expensive.

If your argument is weak you will need to license the patent at a fair price. This can not be set artificially high and for a business is generally a fairly small proportion of profits. If it's hurting too much then look for a workaround carefully examining exactly the claims in the patent. There's usually another way given a little time (meanwhile you license and pay what you owe).

The idea that big companies can just win because they have clever lawyers is bunkum. They have a lot of patents to defend and a finite budget - especially if you are essentially "small fry".

Big companies use patents to defend themselves when entering a market with other big companies.

GitHub's Copilot may steer you into dangerous waters about 40% of the time – study


Oops - there's a bug in their "bad code" assessment checklist. I suspect that whenever they see sprintf instead of snprintf it is clocked as "bad". Now I don't trust the 40% figure telling me how often not to trust Copilot.

We can't avoid it any longer. Here's a story about the NFT mania... aka someone bought a JPEG for $69m in Ether


Re: Blockchain 0, Mona Lisa 1?

Well you can own the copyright to an artist's back catalogue. But that is meaningful ownership which you can monetize. Whereas NFT doesn't seem to correspond to any meaningful sense of ownership. Certainly not for commercial gain other than "resale". Which makes it sound quite Ponzi like.

Third time's a harm? Microsoft tries to get twice-rejected encoding patent past skeptical examiners



This seems like a fundamental misunderstanding of the patent system. A granted patent means very little. Hobbyists and academics are free to do what they like as long as it's for experimental research. And commercial companies would need to be sued, at which point there is an opportunity to examine the legality of the granted patent in a legal setting. In reality a small company would simply point to prior art and if it's clearly a non inventive patent that is already seen in the prior art, the patent owner is likely to back down rather than face an expensive court case they are likely to lose.

Just cough into your phone, please... MIT lab thinks it can diagnose COVID-19 from the way you expectorate


Data set

I wonder if the dataset included people who had other infections such as coughs colds and flu. Wouldn't be surprised if the input was people who tested positive and people who who have no symptoms. So it may be no better than saying have you got a persistent cough? Also those who know they have Covid may be recording under different conditions (physical and mental). So it would be important to test on a random set of subjects and only then test them for Covid with a follow up test a week later.

Just like when you 'game over' two seconds into a new level... Facebook launches Gaming app without games on iOS


Re: What I don't understand ...

The risk that Apple don't get a cut?

Mind the airgap: Why nothing focuses the mind like a bit of tech antiquing


Re: Distractions...

Just set your status to "Do not disturb" and everyone will see your red light status and assume you are on another important call.

Former UK Labour deputy leader wants to know how the NHS's contact-tracing app will ensure user privacy


Re: It will ensure user privacy

How about an anonymous email address - just set up a "burner" email account and setup forwarding (Hey, you could use Hey). Can only be used to send you a message and can't be used to track your activity. You can delete the account and setup a new one at regular intervals.

Good luck using generative adversarial networks in real life – they're difficult to train and finicky to fix


Re: An intelligent networlk

Looks false? Do you think that means you could train up an AI system to decide if it really is true or false? OK, so can you train up another AI system that can beat your first system? Well maybe it can also defeat the human eye? Worth a try.

Just because you're paranoid doesn't mean Google isn't listening to everything you say


Cocaine Noodles

Apparently "Cocaine Noodles" used to work. The audio is broken down into phonemes and these are clustered so many phrases could be misinterpreted - especially as there's no context for activating (no prior conversation to help).

India tells WhatsApp to add filters, ASAP


Maybe some compromise

It should be possible to detect messages that are being forwarded on to many people and add some kind of warning - without actually intercepting encrypted communications. e.g. Hash the message and if the same message is spreading like wildfire, slow down the spread and insert warnings to suggest the message may be fake.

Watch how Google's AI catches shoulder surfers spying on your phone


Could be useful

I should think it would work OK if the user has full control of when this feature is enabled. e.g. sending a personal message - turn on "no-peeking" feature. Then turn off to save battery.

Boffins craft perfect 'head generator' to beat facial recognition


Re: Obvious problem there

Nope - it's a random distortion of the key features so each in-painted face is different and you can generate an arbitrary number of them.

No, Samsung, you really do owe Apple $120m for patent infringement


Re: Who invented what?

Quick Google of Samsung patents reveals they filed more US patents last year than any other company.


Big-in-Japan AI code 'Chainer' shows how Intel will gun for GPUs


Neural Networks are a technique that allows software algorithms to appear to perform "intelligent" processing such as recognition tasks. This falls under the term "artificial intelligence".

Obviously there is more to AI than learning and NNs (or "deep learning").

Note that "artificial intelligence" never claimed to be something similar to human intelligence (or consciousness). Hence the term "artificial". It is just the study of systems that perform tasks that humans would regard as requiring some intelligence (even if none is needed. e.g. chatbots that blindly follow a simple set of rules or scrabble software that just brute force tries every possibility).

Londoners will be trialling driverless cars in pedestrianised area


Re: Glad I'm not a pedestrian in London

And yet statistically cyclists almost never hit pedestrians (unlike cars) despite the Lycra lout scaremongering.

Startup remotely 'bricks' grumpy bloke's IoT car garage door – then hits reverse gear


Re: Why would you need to control your garage door

Or just give your brother-in-law a spare key if you trust him. And trust yourself to lock up when you leave town. And stop checking up on your loved ones. Not convincing use cases really.

D'oh! Amber Rudd meant 'understand hashing', not 'hashtags'


Re: To be fair

Numeracy is a good start but it is also useful to understand the basics of probability (how likely you will be killed by a terrorist) and statistics (median salary v mean salary, error bars, significance tests).

So mathematics can help you better understand the world and the numbers that are reported in the news.

Android beats Windows as most popular OS for interwebz – by 0.02%


Not Dead Yet

Lots of commentards surmising the death of Windows. As Mark Twain quipped...

"The reports of my death are greatly exaggerated"

Windows OS is currently a close second to Android and still in a strong position (much higher than OS X for example).

Europe to push new laws to access encrypted apps data


The only way to enforce a ban would be make it illegal to run non-approved software on any device. Otherwise if you can side-load or install an executable then it could be an end-to-end encryption application or contain hidden somewhere in the interface the possibility of end-to-end encrypted communication.

The only "walled garden" where this is remotely achievable is on locked down Apple devices (iOS) - where even developers are required to acquire special certificates to test their own software.

In such a world, software development would be a potentially illegal and dangerous activity - especially if not being done for a "legitimate" approved organisation. Even then it would be difficult because any software application that supports a scriptable interface (including Javascript) is a potential encryption device. Spreadsheets would also be banned. Javascript would be ditched. Only "approved" code can be executed. All OS's would be locked down, Linux would be frozen. Open Source would be restricted.

Basically running or controlling software would be licensed under the sole auspices of the "authorities". In this dark world all developers would be vetted and regularly checked up on.

And all for what? So we can see the last "goodbye world" message sent by a deranged individual.

Autocomplete a novel phishing hole for Chrome, Safari crims


Re: lastpass-cli: a safe and open-source work-around

Or you can use the offline application Lastpass Pocket. I'm afraid you will have to live with a graphical interface though!


Crims using anti-virus exclusion lists to send malware to where it can do most damage


Sig check

Would have thought exclusion lists include signature hash of each executable. So should be difficult to spoof if not impossible.