Back again

It’s been just over 10 years since I last wrote this blog. I wouldn’t be here again if I wasn’t missing something in the mediasphere. But the politics around digital are moving so swiftly, and in so many directions, that it is hard to keep track of it all. I don’t promise to cover every story, but I am going to follow two main themes: law-making, and law-breaking in the digital world.

Ten years ago, one could see the threads of what is now a much denser tapestry. The digital world, for all its variegated splendour, has resolved into the enormously powerful platforms run by Alphabet (Google), Apple, Meta (Facebook), Amazon, and Microsoft, collectively and colloquially known as Gafam. Virtually everything we do online is mediated by one or more of these firms.

Ten years ago, these firms were telling legislators, that it was “too soon” to regulate them, for fear of strangling the “innovation” that was going to dig us out of the economic abyss that the 2007 financial crisis tipped us into. They are now repeating themselves on the subject of Artificial Intelligence. It was bullshit then; it is bullshit now.

But this time, legislators, at least those that depend on voters electing them, appear more willing to heed other voices. As a result, we saw in October President Joe Biden issue an Executive Order on AI safety, followed closely by UK Prime Minister Rishi Sunak’s AI safety summit. But these events contained only invited guests, mostly men with agendas appealing for freedom to develop general AI, despite what they claim is an existential threat to the human race.

Important dissenting voices, many of them women’s voices, were excluded from the US and UK AI meetings. Those missing voices are mostly concerned with the harms that algorithmic decisions are causing now.

Be that as it may, the last 10 years has seen politics around the world lurch to the right. Politics, at least in the West, seems to be in danger of following many industries, not least those in the digital space, in terms of centralisation of power.

This may not be accidental. Those digital powerhouses and other special interest groups like Big Oil, Big Banks and Big Pharma have spent tens of millions lobbying legislators to pass laws that enabled them to consolidate their positions, usually to the detriment of competitors and customers. It is not stretching the imagination to think these firms would prefer the devils they know rather than take their chances with a fresh set of legislators and, dare I say it, jurists (see the revelations about US Supreme Court Judges Thomas, Alito, Gorsuch, and Sotomayer.)

So that pretty much sets out the stuff of this blog — politics, Big Tech, law enforcement, privacy, surveillance and monopolies are all grist for this mill.

Come along for the ride.

Law-making and law-breaking in the digital world

Why you don’t want to use generative AI for real work

Punch cards used to program early computers. Copyright IBM.

I have been interested in artificial intelligence ever since I was introduced to expert systems in the mid-1980s by Stuart Savory. Savory led the team that developed such a system for fault-finding and repair planning at Nixdorf, then Europe’s fourth largest computer company.

Obviously, things have changed in the intervening years, but the principles that underpin today’s “artificial intelligences” haven’t.

The absolute bedrock that supports today’s overhyped large language models (LLMs, or LMs for short) like ChatGPT is the Bayes Theorem. Simply put, Bayes Theorem gives us a way to calculate the probability of something happening if we know something else (that might be related) is happening or has happened.

This is how we can work out the chances of a tornado ravaging Texas if we know a butterfly flapped its wings in Brazil.

This “Butterfly Effect” is Layer 2 of today’s AI – chaos theory, which is usefully described as the behaviour of systems that show a “sensitive dependence on initial conditions”, or SDIC.

US mathematician Edward Lorenz defined chaos as a state where the present determines the future, but the approximate present does not approximately determine the future. The world abounds with chaos as so defined – the weather, the shape of clouds and trees, how water flows out of a tap, traffic, Jupiter’s Great Red Spot, and even the fractal Mandelbrot Set so beloved of old hippies.

When you apply these principles to real life examples – like weather and traffic forecasting, language translation, facial recognition, the likely state of the NASDAQ at any point in time – two things become obvious very quickly. The first is that the more data you have, the more accurate the result, and second, the absolute importance of the question you are asking, because of SDIC. This is why ChatGPT can and does give very different answers to similarly-framed questions.

The need for data is why LLM developers have trawled the Internet looking for examples of how words are used, in what sequence, and in what context. This allows them to use Bayes Theorem to predict the likely next word in a sequence, given the previous word.

This can work spectacularly well, as recent reports about ChatGPT 4 have shown. But the many odd answers it and others provide are a direct result of the algorithm picking an unexpected word or sequence, but one that merely has a high probability of being right. In effect, all the algorithm is doing is matching patterns of words and sentences and contexts, given its seed word.

How this works is set out in a 2021 paper, On the dangers of stochastic parrots – can language models be too big?, by Emily Bender, Timnit Gebru, Angelina McMillan and Shmargaret Shmitchell.

(Stochastic comes from the Greek word for aim or guess, and is now synonymous with random.)

Language models, they say, refers to “systems which are trained on string prediction tasks: that is, predicting the likelihood of a token (character, word or string (of words)) given either its preceding context or…its surrounding context. Such systems are unsupervised and when deployed, take a text as input, commonly outputting scores or string predictions.”

They stress that “no actual language understanding is taking place in LM-driven approaches… Human communication relies on the interpretation of implicit meaning conveyed between individuals… The training data never included sharing thoughts with a listener, nor does the machine have the ability to do that.”

Small wonder then that some knowledgeable people describe current LLMs as “spicy Autocomplete”. It’s as if the algorithm builds a house from a brick that you give it, based on pictures of other bricks and stones and houses it has stored in its memory. But you probably wouldn’t want to buy it, because you can’t see the plumbing or the wiring or the roof trusses, still less the building certificates. And besides, it might be ugly.

EFF founder to seek protection from US govt snoopers for non-US citizens

The Electronic Freedom Foundation (EFF)  will do more to seek protection for non-US citizens from the US government’s electronic snoopers , co-founder John Perry Barlow said yesterday.

The commitment came in the wake of a week of revelations of widespread collection and trading of private electronic messages by US and UK security agencies.

Barlow, EFF co-founder with former Lotus Corp CEO Mitch Kapor, committed the digital rights organisation to the action under pressure from delegates to ORGcon13, the Open Rights Group‘s annual conference in London.

It will seek to persuade the US to extend Fourth Amendment rights (no search without a warrant) to non-US citizens.

The pressure came particularly from Caspar Bowden, privacy advisor to Microsoft from 2002 to 2011, who found that the US Foreign Intelligence Service Act Amendment Act 2008 (FISAAA) removes all ‘safe harbour’ protection of non-US citizens’ data from US global surveillance.

This makes all non-US citizen’s data stored in the cloud vulnerable to a Section 1881(a), also known as a S.702 order. It allows US agencies to scan, without a warrant, all communications traffic entering or leaving the US, and to target anyone outside the US. Most of the traffic is likely to relate to terrorist, criminal or political content. However, there are no safeguards against commercial or other intellectual property data.

Bowden has spoken of this before, but been roundly ignored by mainstream media.

Barlow, who consults to the US National Security Agency (NSA), said the NSA and GCHQ, the UK’s electronic surveillance agency, have swopped information on each others’ citizens for decades. This is a ‘gentlemens’ agreement’ that allows both agencies to get information that would be illegal for either to acquire directly.

Barlow was at pains to say the people he deals with at NSA are deeply concerned that some of the work it is asked to do borders on illegality.

He said one of the problems was to get the courts to recognise the EFF’s “standing” as a representative of non-US interests. “I will see that we say that a lot more obviously and visibly, and here’s me saying it now,” he said to applause.

Sir Paul Kennedy, the UK commissioner of communications interception, said in his latest annual report (for 2011), that under the UK’s main interception legislation (RIPA 2000) there were nearly 500,000 requests for communications data, 11% down on 2010. “Communications data must only be acquired for the purpose of preventing or detecting crime and where there is an intention to gather evidence for use in legal proceedings,” he said.

Soundbites from the conference will be available later.

EFF founder to seek protection from US govt snoopers for non-US citizens

The Electronic Freedom Foundation (EFF)  will do more to seek protection for non-US citizens from the US government’s electronic snoopers , co-founder John Perry Barlow said yesterday.

The commitment came in the wake of a week of revelations of widespread collection and trading of private electronic messages by US and UK security agencies.

Barlow, EFF co-founder with former Lotus Corp CEO Mitch Kapor, committed the digital rights organisation to the action under pressure from delegates to ORGcon13, the Open Rights Group‘s annual conference in London.

It will seek to persuade the US to extend Fourth Amendment rights (no search without a warrant) to non-US citizens.

The pressure came particularly from Caspar Bowden, privacy advisor to Microsoft from 2002 to 2011, who found that the US Foreign Intelligence Service Act Amendment Act 2008 (FISAAA) removes all ‘safe harbour’ protection of non-US citizens’ data from US global surveillance.

This makes all non-US citizen’s data stored in the cloud vulnerable to a Section 1881(a), also known as a S.702 order. It allows US agencies to scan, without a warrant, all communications traffic entering or leaving the US, and to target anyone outside the US. Most of the traffic is likely to relate to terrorist, criminal or political content. However, there are no safeguards against commercial or other intellectual property data.

Bowden has spoken of this before, but been roundly ignored by mainstream media.

Barlow, who consults to the US National Security Agency (NSA), said the NSA and GCHQ, the UK’s electronic surveillance agency, have swopped information on each others’ citizens for decades. This is a ‘gentlemens’ agreement’ that allows both agencies to get information that would be illegal for either to acquire directly.

Barlow was at pains to say the people he deals with at NSA are deeply concerned that some of the work it is asked to do borders on illegality.

He said one of the problems was to get the courts to recognise the EFF’s “standing” as a representative of non-US interests. “I will see that we say that a lot more obviously and visibly, and here’s me saying it now,” he said to applause.

Sir Paul Kennedy, the UK commissioner of communications interception, said in his latest annual report (for 2011), that under the UK’s main interception legislation (RIPA 2000) there were nearly 500,000 requests for communications data, 11% down on 2010. “Communications data must only be acquired for the purpose of preventing or detecting crime and where there is an intention to gather evidence for use in legal proceedings,” he said.

ata traffic to travel privately on existing networks.

Soundbites from the conference will be available later.

Stewart’s £150m reward for rural correctness

Being a pain in the arse works sometimes.

So, hats’ off to Penrith and The Border MP Rory Stewart, who got  George Osborne to set aside an extra £150m so that up to six million people living in mobile not-spots might actually get a signal.

Stewart organised a Commons debate, possibly still the best-supported debate ever, to get Ofcom to increase its mobile coverage target, and with it, high speed 4G coverage, from 95% to 99%. The motion passed unanimously.

But it came at a cost. Shortly after the debate Stewart went public with tales of being told in dark corridors that uppity first-time MPs needed to shut up and watch the big boys at work.

Obviously, it took more than “a few words to the wise” to intimidate a man who first walked 6,000 miles in two years through Pakistan, Iran, Afghanistan, India and Nepal, and later ran an Iraqi province during the height of the military adventure there.

As I recall, Stewart worked out that it would cost an extra £210m to build the masts to provide 99% coverage, so Osborne has not been over-generous to Stewart. But it is proof of Stewart’s resilience and his ability to work behind the scenes as well as in front.

While this is a coup for Stewart, it’s too soon to call it victory in the battle to get the UK a globally competitive broadband infrastructure, if today’s BT broadband outage, which affected large parts of the UK for hours, is any sign.

 

 

How to drive economic growth in the UK

Economic growth, or rather the search for it, is the big buzzword these days.

It’s always good to look at the numbers, because as Robert Townsend, author of Up the Organisation, the best business book I ever read, said, if you aren’t in it for fun or profit, what are you doing here?

The Office of National Statistics has published some business demographics data. It found that at the start of 2010, the 4.5 million UK private sector enterprises employed some 22.5 million people, and had an estimated combined annual turnover of £3,200bn.

Small and medium enterprises accounted for 59% of the jobs and 49% of the turnover, but they represent more than 99% of enterprises.

Number of enterprises in the private sector, UK, start of 2010.

	Enterprises	Employment	Turnover	Average turnover per employee
		thousands	£millions
All enterprises	4,484,535	22,514	3,212,315
SMEs (0-249 employees)	4,478,595	13,316	1,561,793
All employers	1,193,965	18,982	3,000,522
With no employees	3,290,570	3,532	211,793	59,964
1-9	989,845	3,717	431,378	116,055
10-49	170,410	3,363	482,225	143,391
50-249	27,770	2,703	436,397	161,449
250 or more	5,940	9,198	1,650,522	179,444
Source: ONS

ONS found that of the 280,000 business that started up in 2004, only two-thirds survived to three years, and less than half (46.8%) were left after five years. The figures since the banking collapse may prove much worse.

One can argue whether the start-up rate of 6.2% (start-ups vs existing businesses) is good enough. Probably it is not. But it may be more important to support existing SMEs for the following reasons.

Almost all are one-man bands, sole proprietors and self-employed. They will turn over, on average, just under £60,000.

Companies with up to nine employees do much better, earning on average just over £116,000 per employee. Firms with more than 250 staff enjoy turnovers of almost £180,000 per employee. Which provides evidence for Adam Smith’s argument that the division of labour and task specialisation leads to economic wealth.

So if the government wanted to encourage economic growth, it would encourage existing SMEs to hire more people.

Telcos wrong to charge for delivery

It would be wrong to let internet service providers charge content providers like Google, Netflix and iTunes for delivering content to customers, says Communications Chamber partner Rob Kenny. (This item is cross-posted from Br0kenTeleph0n3)

In a point by point critique of a telco-commissioned AT Kearney paper, Kenny demolishes arguments telcos are using to persuade regulators and politicians to let them to charge content providers for delivery.

Source: R. Kenny – 10% of users cause 55% of net traffic.

In doing so, he strengthens the argument for other countries to follow the Dutch and pass laws that guarantee net neutrality, in other words, that telcos should not discriminate between different types of traffic, and continue to use their “best efforts” to get the bits to their destination as fast as possible.

Kenny’s most telling line is a quotation from Gary Bachula of Internet2, a US non-profit consortium of government and academic network researchers intent on building the next generation internet.

Bachula told the US Senate in 2006, that, based on seven years’ experience of running advanced broadband networks for five million users, “… we seriously explored various ‘quality of service’ schemes, including having our engineers convene a quality of service working group. As it developed, though, all of our research and practical experience supported the conclusion that it was far more cost-effective to simply provide more bandwidth.”

In addition, Kenny points out that the top 10% of internet users are responsible for 55% of the traffic. Why charge content providers when users generate the traffic, he says.

Kenny also shows that peer to peer traffic represents more than 30% of internet traffic. Charging content providers without addressing file-sharing between users would be “incomplete”, he says.

Besides, large content providers use massive server farms close to users to improve customer experience. This lowers the impact on the telcos’ core network, reducing their claim that content providers are free-riding and should therefore be liable for charges.

Kenny’s critique supports evidence gathered by German consulting firm WiK Consult that showed the incumbent telcos will only invest in new infrastructure (eg fibre networks) where threatened with competition, mostly from cable TV operators.

Kenny shows that the support the AT Kerney report offers telcos is deeply flawed, and thinly disguises their determination to stave off competition, keep their monopoly over access to end users, keep prices high and the market inefficient.

Unless governments and regulators ignore these special pleadings, incumbent telcos are unlikely to play a meaningful role in achieving the ambitions set out in Europe’s Digital Agenda and, closer to home, Digital Britain (revised).

WMFFP4EANA4B

UK cities to meet on EU digital single market

How to remove barriers and create a true pan-European digital single market will the subject of a series of meetings between EU officials and local representatives in October and November in the UK and Europe.

There will be two London meetings on 24 and 25 October. The first will discuss Londoners’ response to the Digital Agenda.

Rural broadband will be high on the agenda at the Cardiff and Edinburgh meetings, with Cardiff also discussing eHealth, eGovernment and other uses and applications, the participation of Ofcom Wales, ICT research and innovation prospects.

City councillors from Birmingham, Manchester, and Bristol will meet local stakeholders and provide feedback to the EC director-general Robert Madelin at a London meeting.

Times and venues are still vague, as are whose opinions are deemed worthy of attention.

The meetings, intended to chart local courses of action, are a follow-up to a similar series a year ago just after the European Commission published its Digital Agenda.

A report on those meetings just published found that 1900 civil servants, regulators, industry representatives, members of parliaments, universities, NGOs and consumer associations attended, averaging 75 per country.

The report said press coverage was not an objective of the exercise as such, adding “some (countries) actively decided not to seek media attention”.

To get on the delegate list contact John Doyle at EC INFSO, Belgium, and Jon Zeff in the media directorate at the department for media, culture and sport, or the office.

Newzbin block will not stop online piracy

The UK high court’s decision to instruct BT to block access to Newzbin, a website that links to allegedly illegal copyright material, will do little to dampen either online piracy or emotional debate about it.

BT said in a statement, “This is a helpful judgement, which provides clarity on this complex issue. It clearly shows that rights holders need to prove their claims and convince a judge to make a court order. BT has consistently said that rights holders need to take this route (through the courts). We will return to court after the summer to explain what kind of order we believe is appropriate.”

BT and TalkTalk are appealing an earlier decision to reject their appeal against their loss of a judicial review to set aside parts of the UK’s controverisal Digital Economy Act that deal with online piracy.

Reacting to the Newzbin decision Peter Bradwell, copyright campaigner at the Open Rights Group said, “Website blocking is pointless and dangerous. These judgements won’t work to stop infringement or boost creative industries. And there are serious risks of legitimate content being blocked and service slowdown.

“If the goal is boosting creators’ ability to make money from their work then we need to abandon these technologically naive measures, focus on genuine market reforms, and satisfy unmet consumer demand.”

ISPA secretary general Nicholas Lansman said, “ISPA has long maintained that this is an issue that rights holders should seek to address in court, rather than through voluntary means, and today’s ruling should go some way to offering clarity on what is a complex area.”

Landsman said he was concerned about over-blocking, the ease of circumvention and increased use of encryption. These were ” widely-recognised”, which meany blocking would not stop online copyright infringement.

“Rather, as the government-commissioned Hargreaves Review recently found, there should be more focus on offering innovative, fully-licensed content services to give consumers what they are clearly demanding,” he said.

Pirate Party UK, which campaigns for an “open internet”, said on Twitter, “Hollywood’s win in the Newzbin case is a disaster for ordinary UK internet users. This could open the internet censorship floodgates.”

The decision comes less than a week after EMI’s former COO of new music and president of digital business, Douglas Merrill, reportedly claimed that EMI’s own research showed the peer to peer downloaders were in fact the music company’s best clients.

According to reports by ComputerWorld Australia, the TorrentFreak blog and Cory Doctorow on the Boing Boing blog, Merrill, a former Google CIO, said they used illegal download sites to sample music before buying the CDs of those they liked best.

The Newzbin case relates to videos and films, rather than to music, but the trade bodies that represent both industries have worked together in the past, notably on the Anti-counterfeiting Trade Agreement (Acta).

According to a new assessment of the legality of Acta under European law, “Acta was controversial both in terms of the process and the substance of the neotations. The decision to maintain secrecy (from October 2007) until the release of draft text in mid-2010 was to prove a significant handicap to public understanding and support of the treaty.”

The study, by the policy department of the European Commission’s external policies department,  found that the provisions against online piracy had been very watered down from the original proposals.

However, it found that ACTA extended criminal measures of indirect commercial benefit. This may contradict the European parliament’s position that acts “carried out by private users for personal and not-for-profit purposes” were to be excluded, it said.

The assessment also said there is no “three-strikes” provision that calls for internet service providers to cut off internet access to infringers, but ISPs are required to provide information about subscribers to rightsholders on request.

“However, this provision is not mandatory. There is no requirement for takedown notices. However, the normal producdures for injunctions and provisional measures will be applied in the digital environment,” it said.

Digital underclass mocks faster broadband

What responsibility and accountability is the UK government willing to take on in the areas of digital access, skills and engagement?

This is the question that must be answered before any other, says the London School of Economics media policy project.

In a dossier published today, the project makes clear that the targets for Broadband Britain remain unquantified (beyond a ‘best in Europe’ statement), and, despite Martha Lane Fox’s chumminess with cabinet ministers, her Race Online 2012 initiative to get the nine million Britons who have not yet tried the internet to go online “is independent of, and therefore not accountable to, government or policy strategies.”

The project argues that in addition to the question posed above, there are three others that shape the actions of market and other players. These are

1. How and at what pace can universal access to high quality, high speed broadband be realistically implemented? Is an aim of 100% access realistic or should a policy strategy consider incorporating concrete plans for intermediaries?
2. Which policy goals in relation to digital skills can be set up and which guarantees can be built in so that the most disadvantaged younger and older individuals are reached by these?
3. Which priorities can the government set for broad and sustainable engagement with ICTs by all sections of the population, (and) which sectors besides education and employers should be involved in this?

The answers are not yet in the public domain.

The project argues that the answers need to be informed by research results also published today that show the emergence of a growing “digital underclass” in Britain.

The LSE’s Ellen Helsper found digital exclusion of the most vulnerable groups has become “entrenched”.

“Gaps based on education and employment persist independent of age or other characteristics. They therefore represent a problem that is unlikely to go away even with better infrastructure or as younger generations grow up,” she said.

Helsper warned that people in these groups were the least likely to access public services as public services became ‘digital by default’, now and in future, despite needing it most.

“The market or third sector will not be able to deal on their own with an increasingly entrenched group of digitally excluded individuals,” the project said. “The policy answers have to be based on and integrated with other (social and economic) policy areas.”