* Posts by ttlanhil

10 posts • joined 6 Nov 2019

EFF warns of 'one-way mirror' of web surveillance by tech giants – led by Google

ttlanhil

Re: HTML local storage hole - GDPR fails?

As a dev who spends some time on front-end... LocalStorage can be useful.

As long as browsers treat them as the same thing (e.g. "Clear cookies and site data" in my FF; or similar rules for 3rd party cookies as 3rd party localstorage) it doesn't make the tracking situation any worse than you already have with cookies.

Astronauts brave razor sharp edges and fiddly pipes to bring joy to boffins

ttlanhil
Stop

Re: I saw UTTPS

Well that prompted a terrifying idea...

Unicode (or UTF-8 if you prefer) Text Transport Protocol (Secure)

Senate asks DHS: you don't negotiate with terrorists, but do you pay off ransomware?

ttlanhil
Big Brother

Re: NSA help?

National Sharepoint Agency?

Google's troll-destroying AI can't cope with typos

ttlanhil

Re: Misspeling

butt... illetracy r kewl, mkay?

Plus twitter is often a source of training data for language-based "AI" work...

I'd like to see how well you can handle all of the common abbrevs

ttlanhil

REs are consistent and reliable.

And when they don't work, you can figure out why and how to fix them

Well... in theory.

"AI" is none of the above (as someone who's trained a few "AI" models... and fixed a lot of REs)

Oh dear... AI models used to flag hate speech online are, er, racist against black people

ttlanhil

It's not the programmers; at least not mostly.

A lot of the time, the people who build the models may not even know how it works (both in the sense of not knowing the theory behind, say, neural nets; and in the sense of not knowing which words the algorithm decided mattered - a lot of "AI" model training is "given this input, you should give this output - you figure out how")

The problem is with the training data, and any "AI" needs a lot of it.

Good training data will be entirely self-consistent, properly labelled, and cover all of the variations you want to match.

Good luck ever getting that for any non-trivial task.

Astroboffins peeved as SpaceX's Starlink sats block meteor spotting – and could make us miss a killer asteroid

ttlanhil

Re: Curiousity

> Why so? Curious why twighlight is better than full darkness.

twilight is sparklier

Remember the Uber self-driving car that killed a woman crossing the street? The AI had no clue about jaywalkers

ttlanhil

Re: Reasonable defaults

Yes, but defining the correct criteria is hard.

Do you emergency brake when a bird swoops across the road? (it's close, so need to break hard to avoid it). A large bird may be a similar size to a ball that a child is about to chase onto the road

An attentive human is pretty good about making those decisions (we're built to recognise moving 3D objects) - so providing training data from human-controlled cars can help (video as input, and output of if the driver took action - even if too late)

Part of the problem is those cases are really rare - humans are good at noticing and remembering unusual things; but they make for sparse training data for AI

If it was been able to track the same "object" even though the classification had changed (as opposed to "There's a new bicycle there" "The bicycle is gone, there's a new unknown object"), then it could have kept the movement history and made better decisions

Boffins hand in their homework on Voyager 2's first readings from beyond Solar System

ttlanhil

Re: Frequent Flier Points

Nah, free upgrade to business class

"Work harder! Go faster! Send more data!"

ttlanhil
Boffin

Re: Gravity...

If you fire a single bollock at a screen with 2 slits in it, you'll get an interference pattern as if it had gone through both.

If you fire multiple bollocks at a screen with 2 slits in it, you get an interference pattern that's unknown until it's observed; at which point it may appear to be a forum post where someone takes a bag and shakes a loan of magnets about...

SUBSCRIBE TO OUR WEEKLY TECH NEWSLETTER

Biting the hand that feeds IT © 1998–2020