Martyn Redstone🎗️
10.5K posts

Martyn Redstone🎗️
@mredstone
Head of Responsible AI. Specialising in AI governance and policy, responsible innovation and modernising employment in the new world of work.
St Albans, United Kingdom Katılım Ocak 2009
1.1K Takip Edilen902 Takipçiler

@florencemachine was incredible tonight at the O2
Moody, uplifting, energetic, kooky.
Amazing

English


WHAT RESPONSIBLE AI MEANS FOR RECRUITERS? x.com/i/broadcasts/1…
English

A personal update about Bot Jobs, by @mredstone
open.substack.com/pub/owntheconv…
English

@elonmusk Did you see my work on the LinkedIn algorithm and how recommendation systems can stack proxy bias? doi.org/10.5281/zenodo…
English

We know the algorithm is dumb and needs massive improvements, but at least you can see us struggle to make it better in real-time and with transparency.
No other social media companies do this.
Engineering@XEng
We have open-sourced our new 𝕏 algorithm, powered by the same transformer architecture as xAI's Grok model. Check it out here: github.com/xai-org/x-algo…
English
Martyn Redstone🎗️ retweetledi
Martyn Redstone🎗️ retweetledi

Includes a link to the 100+ page research from @mredstone. You can also watch the @EWMDnet webinar where the experiments and testimony were presented here: youtube.com/watch?v=IkzN9G…

YouTube
English
Martyn Redstone🎗️ retweetledi

Reproducible experiments and testimony suggest LinkedIn’s algorithms systematically reduce the visibility of anyone whose profile doesn't match the patterns the algorithm has learned to reward. Read about proxy bias accountability in AI-driven platforms - bit.ly/3NlyGxG
English

A fictional football match, a real-world ban, and a Chief Constable under pressure. Can we really blame AI?
The news regarding West Midlands Police yesterday is perhaps the most significant warning on AI governance we have seen in the UK.
Here is the reality of the situation:
• Fans were banned from a Europa League match between Aston Villa and Maccabi Tel Aviv.
• The justification included a report citing a clash between Maccabi Tel Aviv and West Ham United.
• That match never happened.
In a letter to the Home Affairs Committee, Chief Constable Craig Guildford admitted the information came from Microsoft Copilot, despite his prior belief that it was a verified Google search result.
This was not a harmless draft email. This was intelligence used to make a significant public order decision that is now causing political outrage and has led to calls for resignation.
For HR and Business Leaders, the lesson is stark:
1. Accountability cannot be delegated to an algorithm. The Chief Constable is the one facing calls to resign, not the software provider. When we use AI to support decisions (whether in recruitment, restructuring, or disciplinary hearings) the "human in the loop" is fully liable for the output.
2. "I thought it was Google" is not a defence. We have spent decades trusting search engines to retrieve facts. We are now using tools that generate text based on probability. Leaders must ensure their teams understand this fundamental difference. A hallucination looks exactly like a fact until you check the source.
3. The risk of 'Shadow AI'. If the Chief Constable didn't know Copilot was being used until after the error was exposed, it suggests a gap in governance. Do you know where and how your teams are using AI right now?
We cannot treat Generative AI as a search bar. It is a creative engine, and without rigorous verification, it can create fictions that cause very real damage.
Question for your teams this week: If you are using AI to summarise data for decision-making, who is checking the raw source material?

English

@rrmdp @alexchukovski @jobboardsrch I will absolutely be listing it on @jobboardsrch - one small step at a time :-)
English

@mredstone @alexchukovski LFG Martyn 💪🏻, hope you make a lot of $$$ and help people find jobs with it!
(And of course see it listed on @jobboardsrch 😄)
English

Pleased to announce the (soft) launch of AI Governance Jobs: jobs.ai-governance.network
With 2026 being the year that people start asking for more transparent, responsible AI, I believe that there will be a distinct need for these kinds of professionals.
cc: @rrmdp @alexchukovski
English

My new research reconstructs LinkedIn’s feed algorithm using only LinkedIn’s own technical papers (2023–2025).
Key finding: unequal visibility can emerge as a structural property of engagement-optimised AI systems, without intentional bias.
doi.org/10.5281/zenodo…
English
Martyn Redstone🎗️ retweetledi

Bondi Beach is not occupied territory. Yet a Jewish celebration there ended in blood. It is not within a military zone, not contested land, not an ‘open air prison’, but still, among civilians, on a day marked for celebration, Jews were once again slaughtered, picked off by a Muslim father and son who were motivated to kill as if it were their God-given right.
Jonathan Sacerdoti 🙏 @jonsac
English







