* Posts by jrchips

14 publicly visible posts • joined 1 Jan 2016

US politicos wake up to danger of black-box algorithms shaping all corners of American life

jrchips

AI advocates tell us the algorithms will "self-learn". On the surface, that sounds very impressive. But it begs the question, "Learn what?" How will developers know what algorithms their AI software is using if it has taught itself?

The more decision making is delegated to AI, the more vulnerable companies will become, because untangling the code and deciphering the AI "self-taught" rules will be a nightmare. Setting aside the issue of frustrated customers who might be denied products or services, there are serious legal risks in the area of perceived discrimination and in product safety.

For example, if a fully autonomous vehicle is involved in an accident, how will the manufacturer demonstrate that the vehicle behaved correctly? The more factors that are involved in the internal AI decision-making, the less transparent, and obvious they become. At some point, I can well foresee a company software engineer saying, in court, "we know what it did, but we don't know why it did it."

Oops, "Bad answer!" Get your checkbook out.

Sorry, but those huge walls of terms and conditions you never read are legally binding

jrchips

Wait until you get pages of virtually unreadable T&Cs for your autonomous vehicle.

The bottom line will be; if you own it, it's yours and it's your responsibility even if the software's flawed. That's how the autonomous vehicle makers will try to escape liability. But it probably won't work. Class action lawsuits over vehicle-related deaths have been very successful. The tech industry and auto industry aren't the same thing.

UK publishes Laws of Robotics for self-driving cars

jrchips

"See everything, track everything, secure everything."

The new rules for the road (and your digital life).

The curious case of a Tesla smash, Autopilot blamed, and the driver's next-day U-turn

jrchips

Re: Hmm

Ah! The small print. It will be the curse of ''intelligent" vehicles.

For liability reasons, the legal team will insist on lots of small print, including explicit language saying you take liability for everything (If you don't believe me, just wait and see!).

So, either you read, and understand, every line of the Operator's Manual or you're at risk of a "gotcha" if an incident happens. But few folks read, let alone understand, every line of an Operator's Manual and no one will want to take a two-day 'understand your car' course to learn all the features. So folks will continue to be confused as to the functionalities. It won't take too many "you should have know it doesn't work that way" incidents to sour them on all that fancy functionality.

It's another reason why you can't creep up on Level 5 and why Google gave up simply incrementing functionality. At some point either the vehicle does it all (Level 5), or it's worth no more than a few understandable driver assist features.

Electric driverless cars could make petrol and diesel motors 'socially unacceptable'

jrchips

Re: "Having a level 5 autonomous vehicle would be very nice indeed."

"we have vehicles on the road that are capable of a very significant proportion of the required driving"

The key words here are "a very significant proportion" because "a very significant proportion" doesn't mean "all". But it needs to be "all' if you want to take the driver out of the equation. As long as you require some occupant interaction/supervision/emergency override you require a sober, alert, non-distracted occupant. So forget about watching movies, sleeping, writing code, etc. Said another way, Level 4.8 isn't almost Level 5, it's just Level 4. And it could be worse, because if 90% of the driving is done for you who's going to be alert for the unplanned, time-uncertain 10%?

Fully autonomous means fully autonomous under every circumstance. Otherwise folks won't trust the vehicle. That's why Level 5 is such a big challenge.

Homeland Security: Putin’s hackers tried to crack electoral networks in 21 US states

jrchips

What exactly does "the weaponization of stolen cyber information" mean?

Without specific examples this just sounds like a lot of alarmist arm-waving.

Senator blows a fuse as US spies continue lying over spying program

jrchips

Amazing, isn't it?

Amazing that there is acknowledgement by the agencies that they aren't following the law, and that they absolutely refuse to do anything about it. The only metric we can guess at as far as domestic data collection is concerned is that it's on more than one individual and less than 300 million!!

More importantly, the obvious conclusion is that the agencies are directly challenging Congress over their governance. It's almost as if they are saying to Congress, "you don't count". That's a pretty big challenge.

No laptop ban on Euro flights to US... yet

jrchips

Where are the airline's customers in all this? Who is speaking for them?

Long haul flights, especially for Business and First customers, are places where serious work gets done on laptops. So banning laptops is a huge productivity hit. Worse those same execs will now have to wait for their checked baggage, a no-no for most busy execs. And then there's the risk of pilfering and missing bags.

If the issue is getting commercial quantities of a detection system out there quickly, let's say so and focus on that. Instead there's a rush to hugely inconvenience the folks who pay the freight. Makes no sense.

Family of technician slain by factory robot sues everyone involved

jrchips

Many automated industrial systems have 'kill switches' (or e-stops) that allow a human being to freeze the operation of the equipment. Typically 'kill switches' are in readily accessible places and spaced in such a way that an operator can reach one quickly. Now the EU is considering making such 'kill switches' mandatory on robots.

So does that mean that all fully autonomous vehicles should have 'kill switches' on them reachable by someone on the outside? One on each corner perhaps?

TV anchor says live on-air 'Alexa, order me a dollhouse' – guess what happens next

jrchips

I'm sorry, but if you've got your home communications set up without adequate safeguards and verification systems, don't blame the TV station for causing you problems. Why would anyone allow Alexa to order something without requiring Alexa to ask for confirmation?

Jeez!

White House report cautiously optimistic about job-killing AI

jrchips

"drivers and cashiers"

Are these the two best examples the report could come up with? Because both are arguable.

Self-checkout has been around for a long time but is still only an adjunct to, rather than a replacement for, retail cashiers. There is nothing happening to suggest that scenario will change anytime soon.

Driverless vehicles are, in reality, still a dream. And, if recent Uber and Tesla missteps are any kind of indicator, the regulations will likely get a whole lot tougher. In all probability they will severely limit public experimentation and raise the bar on 'proof of performance'. Then the rate of adoption will stretch way out and the financial picture will look much less attractive.

So, no, the big AI disruption is far from certain despite all the PR drum beating.

UK.gov flings £30m at driverless car R'n'D, wants plebs to speek their branes

jrchips

As always, "it depends".

Road capacity depends on the space between cars. So what safety margin will autonomous vehicles use when following another? Three seconds behind the car in front, two seconds, or something else? The first is the conservative recommendation from departments of transportation in a number of countries. At 60 mph, and assuming 16 ft per car, that translates into 16 car lengths. Even two seconds, a concession for dry roads and perfect visibility, translates into 11 car lengths. At 40 mph the distances are 11 car lengths and 7 car lengths respectively. But nobody I know follows those rules, especially in rush hour traffic. Not even close. So do driverless car algorithms ignore theses recommendations? If so, on what basis?

Oh, I know the arguments. Driverless cars will seamlessly communicate with each other so it's OK to run them as a train, right? All the manufacturers are going to agree to one communication standard, accept unchecked the info from the car in front and all will respond instantly to whatever the lead car is doing, Sure! But at this point that's a total fantasy.

So what 'gap' will the designers choose? My bet is it will be conservative, for liability reasons if nothing else, and, as a consequence there will be lots of complaints about regular drivers cutting into the gap. The end result is that driverless cars will, at least initially, reduce road capacity not increase it.

Google AI gains access to 1.2m confidential NHS patient records

jrchips

Re: What kind of moronic thinking is that?

Time to legislate universal 'opt-in' language for all third party use of personal data. One of the provisions would be that if you choose not to make the data available to third parties, the service cannot be refused. This would extend to data retrieved from personal computers by software programs, including browsers.

We've currently got the equation backwards. It's time to put the consumer (not the service provider) first.

MPs slam mandarins over failed GP IT system

jrchips

Same old, same old.

These kind of IT project definition and management issues have been around for 40 years.

Until government departments terminate incompetency it will always be 'same old, same old'.