Marzia Briel (@TrustOpenAccess) header
Marzia Briel (@TrustOpenAccess) profile picture

Governance & Regulation Specialist, solutions to commercial exploitation of research, data, public funds and services by private/state actors (personal views)

Joined October 2020
Thank you to my co-panellists #OASPA Lisbon. We unpicked the research integrity issues in scholarly publishing ranging from a transparently labelling research metadata, to defining the value of open peer review to looking at the impact of incorrect/false research data in LLM's.
241
Marzia Briel retweeted
Timelapse of a sunflower bloom
69
602
54
4,209
184,662
💚 CHANGE IS HAPPENING! 💚 Three groundbreaking legal rulings have pushed back against polluting factory farms proving that the law can protect people, nature, and the planet. 🔗 - pulse.ly/vgepwnykx8 💩 Manure declared as waste in Herefordshire 🌍 Climate emission grounds in Norfolk 🐔 Cumulative pollution ruled unlawful in Shropshire In her latest piece for The Planner, our Head of Legal, Emma Dearnaley, explains how these decisions are reshaping planning law and what they mean for the future of factory farming in the UK.
3
20
44
1,098
Still available, in e-book (only £4.19!) and print, #ChangeEverything, my accessible, short book explaining #Green political thinking. If you're in the UK, purchase here supports independent bookshops. hive.co.uk/Product/Natalie-B…
8
35
1
73
32,588
Marzia Briel retweeted
How data was stored in 1982.
77
282
27
1,329
145,867
Marzia Briel retweeted
Nearly 40% of new US homes were built in climate high-risk zip codes at the end of 2024 on.ft.com/41v0rbz
5
20
38
18,845
The perfect match @TimothyWSchwab for @unherd Gates strategic creation of conflict of interest: the“Bill chill” — the chilling effect that makes (Universities) reluctant to criticise the man who so many look to for funding unherd.com/2025/02/what-is-b…
67
Enjoyed the latest @FoundSciTech Governing AI for Humanity event, though the conversation centred around Governing AI for commercialisation.
47
Marzia Briel retweeted
New report by leading semiconductor analyst Dylan Patel shows that DeepSeek spent over $1 billion on its compute cluster. The widely reported $6M number is highly misleading, as it excludes capex and R&D, and at best describes the cost of the final training run only.
SemiAnalysis published an analysis on DeepSeek, addressing recent claims about its cost and performance. $NVDA The report states that the widely circulated $6M training cost for DeepSeek V3 is incorrect, as it only accounts for GPU pre-training expenses and excludes R&D, infrastructure, and other critical costs. According to their findings, DeepSeek’s total server CapEx is around $1.3B, with a significant portion allocated to maintaining and operating its GPU clusters. The report also states that DeepSeek has access to roughly 50,000 Hopper GPUs, but clarifies that this does not mean 50,000 H100s, as some have suggested. Instead, it’s a mix of H800s, H100s, and the China-specific H20s, which NVIDIA has been producing in response to U.S. export restrictions. SemiAnalysis points out that DeepSeek operates its own datacenters and has a more streamlined structure compared to larger AI labs. On performance, the report notes that R1 matches OpenAI’s o1 in reasoning tasks but is not the clear leader across all metrics. It also highlights that while DeepSeek has gained attention for its pricing and efficiency, Google’s Gemini Flash 2.0 is similarly capable and even cheaper when accessed through API. A key innovation cited is Multi-Head Latent Attention (MLA), which significantly reduces inference costs by cutting KV cache usage by 93.3%. The report suggests that any improvements DeepSeek makes will likely be adopted by Western AI labs almost immediately. SemiAnalysis also mentions that costs could fall another 5x by the end of the year, and that DeepSeek’s structure allows it to move quickly compared to larger, more bureaucratic AI labs. However, it notes that scaling up in the face of tightening U.S. export controls remains a challenge.
Readers added context they thought people might want to know
It was explicitly stated in the original DeepSeek report that the ~$6M is a calculation based on the compute time of the final training run using a typical rental rate (without claiming HW was rented). It was never claimed that this would be cost of full development or such.arxiv.org/pdf/2412.19437… p5
958
1,807
488
12,786
2,509,861
open.spotify.com/track/6yLuf… As our world is dying, a new beautiful world is being born, starting with our own individual transformation
30
Trust Lawyers going to be busy teapawt.org/ClarksonsFarm1/status/…
Jeremy joining protesters today against the new inheritance tax regulations for farmers
41
The cost of mandating Open Science (and giving up IP) to the UK University sector is a staggering GBP 287 703 000 (approx Two hundred eighty-seven million pounds) over 9 years - an average of about GBP 32, 000, 000 (Thirty-two million) pounds per year. hesa.ac.uk/data-and-analysis…
36
Some interesting stats on IP commercialisation in Universities. Oxford head and shoulders above the rest - Sheffield interestingly does well too. hesa.ac.uk/data-and-analysis…
42
I met Tim in London at a conference earlier this year and everyone in the room, including Gates funded academics, left enlightened. Philanthrocapitalism is entrenched in academia.
Now I tell my story: I was invited to participate in this Netflix series, but refused b/c I didn't trust the filmmakers. Here's what I found out about their funding. No paywall: timschwab.substack.com/p/why… Please subscribe (paid or unpaid) to my newly launched Substack!
41
Hi, please sponsor our @UniRDG_Law School walk in Reading next week to provide legal support for local communities londonlegalsupporttrust.enth……
38
So wonderful to have connected back to my roots. The dry winter weather force all the predators to move around to find food so it was much easier to see them.
42