Debating and thinking about AI in the workplace

The CEC produced a special report on AI in the workplace, which they called Organising in a changing economy.

an aisle in a computer hall

This is quite a long paper consisting of 23 pages. I made a personal wiki page, Notes on the GMB & AI. I passed the special report through two LLMs to summarise and deduplicate it, this final summary is available on the wiki page. Its headings are: Workers’ Rights and Collective Bargaining, AI Regulation and Worker Protection, Transition Support, Skills and Training, Apprenticeships and Industrial Strategy, and Employment Security. I am of the view that the summary could be further distilled.

One of the best ideas might be the demand to create a “Just Transition Fund” for workers displaced by automation or autonomous vehicles, funded through industry levies.

On reflection, the paper is not revolutionary, nor in reality particularly organisationally focused; AI adoption is clearly, a least to me, an issue where political power and strong state regulation is needed. It’s likely that the classic trade union approach of “what’s in it for me” is not enough, although personally I am a sceptic and am deeply unclear that there is a +ive ROI in using LLMs and that the biggest threat is the sunk finance cost in the rapidly ageing infrastructure,

Matt Williams of the CEC gave a powerful advocacy of how it can work, although given where he works, his employer, one of the contractors building the nuclear power stations, will be a late adopter of AI because AI gets things wrong which is unacceptable in nuclear power management, it can’t yet perform construction work and it may not be cheaper to run the project management software with AI support.

The debate was under special report rules, is available on the video and I spoke for London Region.

My words,

President, Congress, Dave Levy, London Region speaking in support of the Special Report.

Work is changing around us and changing fast.

New rights on paper will only matter if we turn them into real workplace power through strong organisation and updated agreements that reflect the modern economy.

We need to use the new freedoms in the Employment Rights Act to win better recognition and better agreements, otherwise workers will be left to navigate an underfunded tribunal system, while employers bank the benefits.

AI is already in our members’ workplaces and on our desktops – not as a distant threat, but as a daily reality.

The report is honest: AI is not necessarily the enemy – it is the boss’s intentions that matter as has always been true.

Used with workers, AI can cut drudgery and improve services; imposed on workers, it risks discrimination, surveillance, deskilling and displacement.

We demand consultation when tasks are automated, guarantees of redeployment and paid retraining, and stronger legal protections on data, bias and automated decision‑making building on the rights established by the EU’s GDPR and its UK implementations.

Our existing agreements were written for an analogue world, not one of algorithms, data centres, continuous surveillance and automated vehicles.

We need new agreements that say that no dismissal or disciplinary decision should be based solely on an algorithm; and AI should not be introduced without genuine consultation, risk assessments that cover health and safety and equality and data protection before anything goes live. We need to engage with the GDPR data privacy impact assessments.

As a union we need to link AI to the wider fight on skills and industrial strategy.

The report reminds us that rights are meaningless without jobs – and that without serious investment in skills, technological change will deepen regional inequality and de‑unionisation.

The report is very clear: we should treat AI as another organising issue.

I do have a concern that maybe it’s not the next big thing, and we should be arguing politically for greater regulation, and a cautious planning approach for the data centres.

But Let us leave Blackpool determined to organise harder, bargain smarter, and go back to every employer we deal with and say – our agreements must change, because the world of work has changed, and workers deserve protection, power and a say in every decision that shapes their working lives.

The debate was recorded on video and several of the speeches are really quite good. I have said previously that the “special report” format will often reduce debate in the union especially on a topic like this where negotiating ideas and demands are only just being tested .


At lunch time, the London Region held a fringe on the topic of AI. The speakers were Dr Antonio Weiss, an independent expert and author, Jenni Tennison of Connected by Data, Dawn Davidson of Whistleblowers Uk, Aparna Surenda, an employment solicitor from AWO.

There was massive interest, by which I mean that all the speeches/questions from the floor were well informed and seemed to come from workplace engagement not a hobbyist’s interest. Given the restrictive nature of the debate format in the Congress Hall, it was good to hear from shop stewards and reps that were engaging. The promise by Steve Garelick of London Region to continue to engage with the shop floor on this issue is hopeful but perhaps it needs a national committee.

The panel introduced me to the term bossware, I hadn’t heard it before but the GMB and many of its members particularly in Amazon and Uber will be used to it. It’s an enhanced form of surveillance and needs to be resisted as it is dehumanising for the workers and dangerous for their customers. While automated/autonomous vehicles were not raised at Congress or in the fringe, there will be pressure from the US software/vehicle authors to permit them to be used.

We briefly spoke of how to organise in the hyperscaler service providers and noted such tools as the corporate compliance requirements established by the GDPR, such as whistleblowing to the DPO, using the GDPR & engaging with Data Privacy Impact Assessments. I may have mentioned European law and the digital market, the GDPR and the AI act. The GDPR adequacy agreement is one means by which we can acquire enforceable rights from European law, extending the Trade & Co-operation agreement and the “Reset” would be others. Full membership of the EU single market or political union is a third.

If these laws aren’t sufficient, we can and should use Health and Safety laws, although H&S enforcement hasn’t helped us in the Amazon warehouses.

One of the platform speakers raised the issue of copyright, framing it as stealing to fuel the training processes. It’s not, such propositions misunderstand the ‘training process and fail to understand that we probably need different intellectual property laws, for science publishing and art.

The notes I took when composing my intervention, are that,

  • It is a massive threat to privacy.
  • There is no evidence for an ROI through job loss.
  • It’s a bubble based on circular debt, which maybe the biggest risk.
  • The response to these threats must be political.
 …

AI’s black hole

People are interested in what they’re interested in and it seems that I can’t put copyright down. I was at in conference on artificial intelligence over the weekend, organised by global justice now (GJN).

by NASA via unsplash

The question I want to ask though is, that if AI, is meant to be so clever, how does it acquire all the science, social science and humanities white papers currently held behind copyright enforced paywalls or on DNS blocked repos.

It seems they don’t, unless they have “Partnership agreements” with the copyright holders for which they pay.

I’m concerned about this hole in human knowledge that is unavailable to the LLMs. Are we really betting all this money on only part of what we know?

The general intellect it is not! …

How important is profit for software projects?

a robot using a laptop

I have just posted on LinkedIn, “Does AI actually impact the bottom line in a good way?”. I look at an article from the Register based on a Gartner report which reports that in their survey AI projects “sucess rates were “only 28 percent of use cases fully succeed and offer return on investment (ROI).”

I look at at arguments from Perez’s Technological Revolutions and Financial Capital and remember a late 20th century paradox that while many IS projects had poor and even negative ROIs, firms that didn’t invest failed.

 …

About the AI Boom

an AI chip on a board

Five things

  1. There’s no positive business model, they can’t replace staff, or at least not without new supervisors.
  2. They are now borrowing to invest in each other.
  3. Hallucinations are a feature, it gets things wrong,
  4. The rule base’s ownership is obscure and Code is Law
  5. It’s [deliberately] wasteful of resources which people need to live.

It’s useless as a pillar of a growth/industrial policy. …

The downsides of the A.I. industry.

Karen Hao in front of a book shelf

I have recently, watched the YouTube video, “How tech CEOs are lying to you”, where Karen Hao is interviewed by  Aaron Bastani on Novara Media's channel. Ms Hao has just published a book, “Empire of AI” and the interview covers the topics of the book.

At the centre of her arguments are, that large language modules aka AI, and their use of resources, water, power, land and rare metals are a choice, and one that society cannot afford. She questions the business model of the AI industry, sees it as a threat to [US] wealth and notes its scofflaw approach to its own regulation and its oppression of poor and vulnerable communities. She is highly critical of the motivations of the oligarchs funding the AI bubble.

The remainder of this article, which is over 2000 words is overleaf, use the "Read More" button ....

On the necessary regulation of AI

a robot issuing a parking ticket, generated by deepai.org

I wrote a piece for Chartist on AI & its regulation, which I have signposted on LInkedin. I look at its likely macro-economic effects and the essential defence of Article 22 of the GDPR, where I say,

… the most important defences that we as citizens, workers, and consumers have is the EU’s GDPR, which in Article 22 & Recital 71 establishes what they call a right to “freedom from profiling”. This, through the rulings of the CJEU, has become quite extensive and now prohibits such things as ‘general monitoring’, a legal protection brought forcefully to light by the French supervisory authority fining Amazon €32m for violations of the GDPR within their workforce management regime.

In the article, I talk about the problem of Authority vs Popularity, the need for open source, and source citation. I also review the need for some innovators for privacy and competitive advantage and the possible future of regulation of AI to ensure decency and accountability. I also look at the patchy European response and the paradoxical attitude of the US.

I conclude.

In summary, there are plenty of laws to ensure that AI and its owners behave decently, and in some European countries, the will and resources to enforce them, but it’s not universal. Also, there are important economic countervailing forces opposing the creation of a privately owned “Global Intellect” even if the current technology is capable of such a task.

 …

On AI at GMB 25

On AI at GMB 25

There were two debates on AI, calling for Transparency, Accountability and Job Security. I briefly review the and reference the two motions, point at the TUC proposals, the French SA's fining of Amazon, ORG's critique of the Data Use and Access Act and cross reference other articles I have written on the coming impact of AI/LLMs. The full article is overleaf ....

AI ethics and accountability

a silhouette of a man in fron of computer code

I attended a lecture on AI Ethics and blogged on Linkedin and mirrored it on Medium, I catalogue the issues as presented by Dr Hung and used Google to see if there were any obvious gaps.

I look at Garbage In, Garbage Out problem, repeat the calls for transparency, I repeat my arguments about authority vs popularity and the role of the GDPR, and I look at copyright and the four [software] freedoms.  …

An AI prosecutor?

An image of a robot in black and white

I wrote a Linkedin an article called an AI prosecutor. In it I say,

The problem with modern software is much of it is inference, and completely unsuitable for “beyond reasonable doubt”. It’s also opaque and likely to fail the tests around if it returns popular vs accurate and authoritative results. It’s often wrong and arguably a bullshitter. The EU’s GDPR introduced the right to freedom from profiling, which means a freedom from being processed automatically by computers. This is an important barrier.

This is my first written declaration that that the GDPR’s “freedom from profiling” is a crucial defence of humanity against the machines.

My alarm about the consultation was probably unnecessary.  …

Whose jobs are AI coming for?

Whose jobs are AI coming for?

McKinsey have produced a report on the role of generative AI on productivity and the future economy. The white paper can be found on their website. They launched the paper with a series of webinars, one of which I attended. The rest of this article describes my notes and thoughts ... I made a blog post on linkedin which I mirrored here, to see the full article, either "Read More", or click the linkedin hyperlink.