UK watchdogs ask how they can better regulate algorithms
- Reference: 1651226470
- News link: https://www.theregister.co.uk/2022/04/29/drcf_algorithm_regulation/
- Source link:
While "algorithm" can be defined as a strict set of rules to be followed by a computer in calculations, the term has become a boogeyman as lawmakers grapple with the revelation that they are involved in every digital service we use today.
Whether that's which video to watch next on YouTube, which film you might enjoy on Netflix, who turns up in your Twitter feed, search autosuggestions, and what you might like to buy on Amazon – the algorithm governs them all and much more.
[1]
While that all sounds pretty benign, regulators are hip to the fact that algorithms don't always work for the benefit of consumers.
[2]
[3]
"Algorithmic systems, particularly modern Machine Learning (ML) or Artificial Intelligence (AI) approaches, pose significant risks if used without due care," [4]the group said . "They can introduce or amplify harmful biases that lead to discriminatory decisions or unfair outcomes that reinforce inequalities. They can be used to mislead consumers and distort competition."
The DRCF – comprised of the Competition and Markets Authority (CMA), the Information Commissioner's Office (ICO), and the Office of Communications (Ofcom) – has set out a workplan for the coming year in which it aims to:
Better protect children online
Promote competition and privacy in online advertising
Support improvements in algorithmic transparency
Enable innovation in the industries they regulate
However, one does not simply pop the hood on an algorithm and say "yep, this checks out." We're talking about vastly complex mathematics that can only really be assessed after they've done the job they were built for if you're looking to root out (un)intentional biases and other nastiness.
We're also talking about information that the companies who own them consider proprietary – as evidenced by the [5]lawsuits launched over this fact .
[6]
While digital secretary Nadine Dorries – who reportedly asked Microsoft [7]"when they were going to get rid of algorithms" – may not understand them, at least the DRCF has opened the floor to people who... might ?
Gill Whitehead, DRCF chief executive, said in a statement: "The task ahead is significant – but by working together as regulators and in close co-operation with others, we intend for the DRCF to make an important contribution to the UK's digital landscape to the benefit of people and businesses online.
"Just one of those areas is algorithms. Whether you're scrolling on social media, flicking through films or deciding on dinner, algorithms are busy but hidden in the background of our digital lives.
[8]
"That's good news for a lot of us a lot of the time, but there's also a problematic side to algorithms. They can be manipulated to cause harm or misused because firms plugging them into websites and apps simply don't understand them well enough. As regulators, we need to make sure the benefits win out."
[9]Algorithm can predict pancreatic cancer from CT scans well before diagnosis
[10]Uni team demo algorithm to shield conversations from eavesdropping AI
[11]Broken password check algorithm lets anyone log into Cisco's Wi-Fi admin software
[12]Complaints mount after GitHub launches new algorithmic feed
[13]Twitter's machine learning algorithms amplify tweets from right-wing politicians over those on the left
Stefan Hunt, the CMA's Chief Data and Technology Insight Officer, added: "Much work has already been done on algorithms by the CMA, FCA, ICO and Ofcom but there is more to do. We're asking now, what more is needed, including from us as regulators and also from industry?"
As for what that work might be, two discussion papers from the DRCF on the benefits and harms of algorithms, and the landscape of algorithmic auditing and the role of regulators therein can be viewed in detail [14]here .
Both point out: "This discussion paper is intended to foster debate and discussion among our stakeholders. It should not be taken as an indication of current or future policy by any of the member regulators of the DRCF."
It may be that we're just too far gone to do anything useful about harmful algorithms. It's not like the tech giants, for whom proprietary software prints money, are going to suddenly bin the programming that underpins their products and services.
And while one compsci boffin may be an "algorithm expert," there's no guarantee that they will be able to look at Twitter's and have a clue what any of it means or refers to.
The opportunity to comment is open until June 8. Opinions are to be submitted to DRCF@ofcom.org.uk.
To the regulators, we say good luck. And don't listen to any views from journalists. ®
Get our [15]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YmwLq9x80h3Kqn4e23sckwAAAMU&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YmwLq9x80h3Kqn4e23sckwAAAMU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YmwLq9x80h3Kqn4e23sckwAAAMU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://www.gov.uk/government/news/uk-s-digital-watchdogs-take-a-closer-look-at-algorithms-as-plans-set-out-for-year-ahead
[5] https://www.theregister.com/2020/03/11/facebook_ai_chips/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YmwLq9x80h3Kqn4e23sckwAAAMU&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.politico.eu/article/nadine-dorries-digital-minister-big-tech/
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YmwLq9x80h3Kqn4e23sckwAAAMU&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[9] https://www.theregister.com/2022/04/27/ai_pancreatic_cancer/
[10] https://www.theregister.com/2022/04/22/new_tech_uses_realtime_ml/
[11] https://www.theregister.com/2022/04/14/cisco_wireless_controller_flaw/
[12] https://www.theregister.com/2022/03/23/github_for_you/
[13] https://www.theregister.com/2021/10/25/twitters_political_bias/
[14] https://www.gov.uk/government/publications/findings-from-the-drcf-algorithmic-processing-workstream-spring-2022
[15] https://whitepapers.theregister.com/
Re: Algorithms?
Agreed, I go further in so much as you cannot withhold a service/product or discriminate against a customer by offering discounts if the customer wants to opt out and opt out should be the default setting for data collection.
Re: Algorithms?
There is a very strong tendency to misunderstand what algorithms are, and what they should be for. The statistician Sir David Spiegelhalter referring to predicting mortality rates in older people said that the model "fitted the data very well" about increased mortality due to factors such as smoking, drinking, diabetes etc. It is reality that exists, and algorithms which are used to implement models of reality to predict or explain events.
Many people seem to think that once you have a statistical model, it is 'reality' whereas in fact it is a model, and any computer program which 'implements' an algorithm does it only to an approximation of 'reality'.
Re: Algorithms?
A statistical model is only as good as the data you fed it, and can be totally, ridiculously wrong (I have names...) when you forget to take in account some required factor (either because you didn't think or knew it mattered). Unfortunately, once such a model is built people tend to automatically consider it as all-knowing and infallible.
I blame the old SciFi movies where "The Computer" was always omniscient and always had the right answers. People haven't yet learned real computers are just glorified pocket calculators and not some omnipotent mix between an old sage and an oracle.
Re: Algorithms?
Absolutely. Every word out of that woman's mouth proves she does not understand her remit in the slightest and has no interest in trying to understand it. Which tbh makes her the perfect person for a government job.
If they pass some stupid rule that we have to publish all algorithms before they are used then how will we fix bugs, write new software, etc?
"but there's also a problematic side to algorithms"
This is like saying there's a problematic side to breathing, because living people could do something harmful.
God forbid they should ever send a minster for "digital" on a one day course explaining what digital computing actually is, they might stop thinking that it all runs on magic, and realise that algorithms and computer code are the same thing. I await Mad Nad's law mandating that all code be sent to the Algorithm Police so it can be checked for unbadness before compiling it, and being hauled in for questioning as to what harm this suspicious add(x, y) function is going to cause.
Re: "but there's also a problematic side to algorithms"
Algorithms are just a small part of the problem when considering programs which are data-driven. Who is going to review gigabytes of training data? Another algorithm? Or Britain's arts graduate civil servants?
I have very little faith in these chair polishers
First they would need to understand what they want to regulate. So they are already stuck.
Re: I have very little faith in these chair polishers
Come on, since when do politicians try to understand anything?
Even if they could, it's so much easier, quicker and more gratifying to force the perceived solution upon the Great Unwashed, to show them who is boss, and how clever you are. That's how pi = 4 and other similar genius ideas come to be.
OMG
> While digital secretary Nadine Dorries – who reportedly asked Microsoft "when they were going to get rid of algorithms"
Kill me NOW!
Re: OMG
There was an equally dumb question asked by a UKIP representative. What happens when renewable energy runs out?
These people walk among us!
Re: OMG
I don't mind them walking, as long as they aren't the ones supposed to rule my life!
I would say thank [deity] I'm not British, but there are specimen just as bright over here...
Re: OMG
Solar power is in short supply every night. There again, do politicians understand that?
Re: OMG
Politicians?
Do they understand things?
What things do they understand?
Let's find out!
That would have been a tv show I'd watch. Whether J D Salinger was producing it or not.
Re: OMG
It does run out. Wind stops blowing, clouds stop sun shining, in fact it stops shining once a day regular as clockwork. Renewable energy runs out more frequently than fossil fuels do. Fossil fuels have been predicted to have ran out by now since when I was at school in the 70s. Bar the occasional power cut the gas is still flowing and the lights are still coming on.
Re: OMG
TahtOne: "Kill me NOW!"
No chance, if I've got to suffer this lot, so have you!
Stiff upper lip and all that, don't you know.
Apparently it never rains but it pours ... and oft can there be deluges of epic biblical proportions
When things are bad they can be very bad. When things can be very good they are guaranteed to be very bad for things that are bad and/or not good and great. The problems and battles that you might see and realise are all before you now with the likes of algorithms, and you are racing headlong blindly forward to try to control and mitigate with regulation in fields that are beyond your command, are only the tip of the titanic dilemma you will have to deal with and submit to in order to survive and prosper as if in the lead with changing fundamentals presenting events/future views elsewhere ..... for there is no other available option to choose.
The likes of the following, which is available in all sorts of custom flavours dependent upon the assets and liabilities of targeted customers/valued clients, is that which confronts you and should you try to deny exists, compete against injudiciously or oppose wholeheartedly, can simply decide to thwart you via ITs very strange means and virtual memes. IT and AI are very good and great at that.
amanfromMars [2204290825] ....... saying more on [1]https://www.nationaldefensemagazine.org/articles/2022/4/28/winner-takes-all-for-uk-armys-training-simulation-contract
[ Thank you. Your comment will be displayed soon after reviewing. ]
Well, well, well, Stew Magnuson, that's quite another AWEsome type development surely [UKGBNI Army Warfighting Experiment] ...... so maybe not anything really new to that and those with more than just experimental experience in the field with its many very sensitive, highly proactive virtual arenas of relatively anonymous engagement realising that having leading skin in the game is akin to being able to name one's own price for what one would know and be prepared to share about how to enable and/or disable vital elements and virulent components within it whenever one likes ...... which to be perfectly honest with all, one needs to be thinking and realising easily costs failing opposition and trailing competition crippling sums in the trillions, given the nature of that which is discovered as one delves ever further and deeper into that which is uncovered and energised.
However, it may very well be the case, that expanding command and control of such developments across all militarised and para-militarised services is not something which can similarly scaled up and expanded, and such will thus always remain in the secure and failsafe grip of a very select few indeed ........ which I suppose is quite normal and therefore fully to be expected ..... and such will always be, as has always been the case, a private/pirate sector/body leading from way out in front of all others in the public and military and paramilitary sectors which will always be dependent upon that which is fed back down to them from that which is phormed and active way up ahead of them.
And as more information and intelligence escapes and merges in these times and spaces of 0days and Remote ACTive [Advanced Cyber Threat] Access Trojans will there be more instances of evidence in support of a drift towards a unifying singularity of purpose and destination, for the following which was also very recently shared with you here on National Defense Magazine in a comment on your article entitled " JUST IN: Startup Creates Simulator for Info-Warfare Ops " 4/27/2022 says very much the same as all in the above ......
“A U.K. startup has introduced computer-based training software that simulates the information environment warfighters encounter during conflict.”
To only imagine that as something relatively novel and not also realise it long active as a well practised model [modi vivendi/operandi] deployed virtually everywhere and exercised by all manner of private and pirate and public operations/operands, some renegade rogue and malevolent and other not so, identifies one as being catastrophically vulnerable to the programs presented by the very best of the best in their class in the Live Operational Virtual Environment/Virtual Information Environment chosen for remote manipulation and sublime alteration/metadatabase adjustment.
And its success, and one would be very wrong to deny such developments are remarkably successful, has one having to admit that the true nature of one’s human existence is led by remote spookily strange entangled virtual stimulation of thoughts for activation in simulation arenas copied to/from Earth ……. which has one pondering on whether it be a human Earth based project, which maybe can be interrupted or disrupted and prevented or an alien space based program which maybe cannot be interrupted or disrupted and prevented?
[1] https://www.nationaldefensemagazine.org/articles/2022/4/28/winner-takes-all-for-uk-armys-training-simulation-contract
I don't see why they need to regulate Al Gores musical ability.
Seriously though, as stupid as my comment is, the policing or analysing of an algorithm that might be built from thousands of pieces of data is actually stupider. Especially something that could potentially be evolving based on new inputs. Whilst we're on this subject could someone please own up to buying the same thing from Amazon multiple times? Someone must have for that algorithm. I can only use one spatula. Why do I need another? Exception for fish spatula of course.
> the policing or analysing of an algorithm [...] is actually stupider
It's the use of algorithms which needs some regulation. Like for instance for those used to hire people, and which might cause some classes of people to be totally unable to get a job. It is necessary, since algorithms are such blunt tools, to make sure there are safeguards built in the process.
.
> someone please own up to buying the same thing from Amazon multiple times
The issue is, as usual, that you think it's all about you... It isn't. Those ads aren't paid by you, but by the ad agency, and the ad agency pays more for "targeted" ads. It doesn't matter if those ads are relevant to you, as long as Amazon can bill a premium for showing them to you.
How is an algorithm going to decide who to hire? On what basis do you train it? Sure it can look for keywords in CV's but unless someone is purposely making it biased then bias shouldn't be possible.
As for Amazon, Amazon makes money whoever I buy from. The idea behind those emails is to try and show me something I want yet they show what I already have. I'm sure they do get paid by ad agencies but not as much as a sale would bring them by working out what I want. They have actually got better at this over the years or maybe I'm just addicted to kitchen gadgets.
Someone here might have bought more than one chainsaw recently
Algorithms?
I think the use of the word "algorithms" shows the lack of understanding of computer systems by "those in power".
Instead of the implicit assumption that "algorithms are bad!" (which is a logical equivalent to "computer programmers are bad") perhaps the more sensible target description should be something more like "machine driven personal profiling and targetting by corporate entities"?