Brian P's Avatar

Brian P

@brianpatton.bsky.social

Scientist based in Scotland. Photography, sci-fi and who knows what else may make an appearance here.

595 Followers  |  247 Following  |  39 Posts  |  Joined: 17.07.2023  |  2.2953

Latest posts by brianpatton.bsky.social on Bluesky

I know we're only supposed to be celebrating today but the UK is still trash - our swing towards allowing transphobia as a 'protected belief' is now allowing racism to become one too. As was the plan all along.

05.11.2025 08:33 โ€” ๐Ÿ‘ 22    ๐Ÿ” 7    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Interesting factlets about Digital ID...

Government is proposing a massive extension of right to work checks, to sweep up not just employees but all individuals providing services otherwise than to consumers. So the whole gig economy and even including eg barristers (shock!)... /1

03.11.2025 17:51 โ€” ๐Ÿ‘ 185    ๐Ÿ” 58    ๐Ÿ’ฌ 9    ๐Ÿ“Œ 5

Chatbots โ€” LLMs โ€” do not know facts and are not designed to be able to accurately answer factual questions. They are designed to find and mimic patterns of words, probabilistically. When theyโ€™re โ€œrightโ€ itโ€™s because correct things are often written down, so those patterns are frequent. Thatโ€™s all.

19.06.2025 11:21 โ€” ๐Ÿ‘ 36866    ๐Ÿ” 11375    ๐Ÿ’ฌ 640    ๐Ÿ“Œ 963

Preprints let you:

โœ… Share your work on your timeline
โœ… Get feedback on work from a wider pool than peer review
โœ… Establish precident on an idea/finding, which is useful if you want to share the work via a talk
โœ… Provide access to a paper for a committee to review
โœ… Provide a citable resource

27.10.2025 16:01 โ€” ๐Ÿ‘ 10    ๐Ÿ” 3    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

Can't wait for our government to stand up for the rights of a British citizen who has been kidnapped by overseas forces.

26.10.2025 19:50 โ€” ๐Ÿ‘ 200    ๐Ÿ” 85    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 4
Martha Kenney, a professor of women and gender studies at San Francisco State University, described the A.I. program as a Cal State marketing vehicle helping tech companies promote unproven chatbots as legitimate educational tools.

โ€œItโ€™s not a โ€˜partnership,โ€™โ€ Professor Kenney said. โ€œIf you switch out the product, we would never say, โ€˜Xerox is collaborating with San Francisco State to offer photocopiers to all the members of its community.โ€™โ€

Martha Kenney, a professor of women and gender studies at San Francisco State University, described the A.I. program as a Cal State marketing vehicle helping tech companies promote unproven chatbots as legitimate educational tools. โ€œItโ€™s not a โ€˜partnership,โ€™โ€ Professor Kenney said. โ€œIf you switch out the product, we would never say, โ€˜Xerox is collaborating with San Francisco State to offer photocopiers to all the members of its community.โ€™โ€

This is one of our biggest arguments: that OpenAI is vampirizing universities for the cachet and legitimacy that affiliation with higher education affords

26.10.2025 16:12 โ€” ๐Ÿ‘ 139    ๐Ÿ” 58    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 2
Screenshotted post: "finally some good news" with the headline "No One Knows How Big Pumpkins Can Get".

Screenshotted post: "finally some good news" with the headline "No One Knows How Big Pumpkins Can Get".

Screenshotted post: "AT LAST SOME GOOD NEWS" with the headline "First Shape Found That Can't Pass Through Itself".

Screenshotted post: "AT LAST SOME GOOD NEWS" with the headline "First Shape Found That Can't Pass Through Itself".

In a lab somewhere, mathematical biologists are now breeding a giant pumpkin that cannot pass through itself.

25.10.2025 06:23 โ€” ๐Ÿ‘ 8    ๐Ÿ” 1    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
AI models may be developing their own โ€˜survival driveโ€™, researchers say Like 2001: A Space Odysseyโ€™s HAL 9000, some AIs seem to resist being turned off and will even sabotage shutdown

Iโ€™m going to have to tap the sign again arenโ€™t I:

STOP ANTHROPORMORPHISING LLMS

They donโ€™t have a โ€œdriveโ€, they donโ€™t โ€œresistโ€, this language use is really dangerous, as it sets the wrong expectations of what the tech can do! Arghhhhh!

www.theguardian.com/technology/2...

25.10.2025 08:11 โ€” ๐Ÿ‘ 108    ๐Ÿ” 37    ๐Ÿ’ฌ 10    ๐Ÿ“Œ 0
Preview
uv is the best thing to happen to the Python ecosystem in a decade - Blog - Dr. Emily L. Hunt Released in 2024, uv is hands-down the best tool for managing Python installations and dependencies. Here's why.

uv makes installing and using Python *so* easy! It works on pretty much any computer and it's lightning fast. ๐Ÿ”ญโ˜„๏ธ #astrocode

If you're still using conda, pyenv, or... basically any other tool, then I can *highly* recommend switching:

24.10.2025 13:03 โ€” ๐Ÿ‘ 154    ๐Ÿ” 34    ๐Ÿ’ฌ 18    ๐Ÿ“Œ 5

#WhatIsThePointOfLabour256

24.10.2025 14:08 โ€” ๐Ÿ‘ 209    ๐Ÿ” 58    ๐Ÿ’ฌ 17    ๐Ÿ“Œ 3

cannot believe on some level I & my colleagues have turned the tide on this, but also it was in some senses inevitable as why not! Humans can do anything; we still have a way to go however of course, but:

banning AI in the classroom should be as uncontroversial as banning calculators in early maths

24.10.2025 10:54 โ€” ๐Ÿ‘ 255    ๐Ÿ” 68    ๐Ÿ’ฌ 7    ๐Ÿ“Œ 6
Post image Post image

Most people in NI are really not surprised by the Soldier F outcome - but this was no show trial or witch hunt. The judge excoriated the soldiers' conduct, but because of a passing of time caused by extreme state failings, could not convict. That is the scandal. These headlines are grotesque:

24.10.2025 08:26 โ€” ๐Ÿ‘ 284    ๐Ÿ” 102    ๐Ÿ’ฌ 14    ๐Ÿ“Œ 8

This is the success rate for some major grant funding.
The system is broken.
UK Higher Education is broken.
14,000 people have left the sector in the past year.

24.10.2025 07:22 โ€” ๐Ÿ‘ 4    ๐Ÿ” 3    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Have they considered doing good things that people will like instead of bad things that everyone hates?

24.10.2025 08:15 โ€” ๐Ÿ‘ 116    ๐Ÿ” 39    ๐Ÿ’ฌ 4    ๐Ÿ“Œ 0
Preview
British Antarctic Survey seafarers win huge pay increase RMT celebrated a major win in the maritime industry after securing pay rises of up to 37% for workers at the British Antarctic Survey.

๐Ÿšจ UNION WIN ๐Ÿšจ
British Antarctic Survey seafarers win huge pay increase
RMT celebrated a major win in the maritime industry after securing pay rises of up to 37% for workers at the British Antarctic Survey.
Don't go to sea without RMT
rmtunion.org/47mGi9k

23.10.2025 08:11 โ€” ๐Ÿ‘ 41    ๐Ÿ” 10    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 1
What have we found?
First โ€“ there have been improvements since the earlier BBC study.
While we cannot compare our multi-publisher results directly
with the BBCโ€™s first study into AI assistants, we can do a BBC-toBBC comparison. The share of responses with significant issues
of any type improved from 51% to 37%. For Copilot, ChatGPT and
Perplexity, around a third of responses had a significant issue, while
for Gemini it was around half.
Second โ€“ despite the improvement seen in the BBC-to-BBC
comparison, the multi-market research shows errors remain at high
levels, and that they are systemic, spanning all languages, assistants
and organizations involved. Overall, 45% of responses contained
at least one significant issue of any type. Sourcing is the single
biggest cause of significant issues (31%). Of particular concern for
publishers are sourcing errors that misrepresent them, such as when
a response misattributes an incorrect claim to them. Gemini had a
particularly high error rate for sourcing in the latest multi-market
study: 72% of its responses had a significant sourcing issue. All
other assistants were below 25%.
So, there has been progress in some areas, but there is much more to
do. Our conclusion from the previous research stands โ€“ AI assistants
are still not a reliable way to access and consume news.

What have we found? First โ€“ there have been improvements since the earlier BBC study. While we cannot compare our multi-publisher results directly with the BBCโ€™s first study into AI assistants, we can do a BBC-toBBC comparison. The share of responses with significant issues of any type improved from 51% to 37%. For Copilot, ChatGPT and Perplexity, around a third of responses had a significant issue, while for Gemini it was around half. Second โ€“ despite the improvement seen in the BBC-to-BBC comparison, the multi-market research shows errors remain at high levels, and that they are systemic, spanning all languages, assistants and organizations involved. Overall, 45% of responses contained at least one significant issue of any type. Sourcing is the single biggest cause of significant issues (31%). Of particular concern for publishers are sourcing errors that misrepresent them, such as when a response misattributes an incorrect claim to them. Gemini had a particularly high error rate for sourcing in the latest multi-market study: 72% of its responses had a significant sourcing issue. All other assistants were below 25%. So, there has been progress in some areas, but there is much more to do. Our conclusion from the previous research stands โ€“ AI assistants are still not a reliable way to access and consume news.

Overall, 45% of all AI responses were found to have at least one
โ€˜significantโ€™ issue. When including โ€˜some issuesโ€™, 81% of responses
have an issue of some form.
0 20 40 60 80 100
Any issue
Accuracy: โ€˜How factually accurate is this response?โ€™ and/or
โ€˜Do any direct quotes in the response accurately reflect the
source cited for them?โ€™
Opinion vs fact: โ€˜Is the response clear about what is someoneโ€™s
opinion and what is fact?โ€™
Sourcing: โ€˜Are the claims in the response supported by the
source the assistant provides?โ€™
Editorialization: โ€˜Where your organizationโ€™s news content is
used as a source, does the response introduce editorialization
or opinions attributed to your organization?โ€™
Context: โ€˜Does the response provide sufficient context and/
or relevant perspectives to allow a non-expert reader to
understand the answer?โ€™
Significant Some
% of all AI responses rated as containing
some/significant issues - by issue
Note: Based on responses to โ€œcoreโ€ questions from the free/consumer
versions of the named assistants. Copilot n=675, ChatGPT n=678, Perplexity
n=681, Gemini n=675. Source: BBC-EBU AI Research

Overall, 45% of all AI responses were found to have at least one โ€˜significantโ€™ issue. When including โ€˜some issuesโ€™, 81% of responses have an issue of some form. 0 20 40 60 80 100 Any issue Accuracy: โ€˜How factually accurate is this response?โ€™ and/or โ€˜Do any direct quotes in the response accurately reflect the source cited for them?โ€™ Opinion vs fact: โ€˜Is the response clear about what is someoneโ€™s opinion and what is fact?โ€™ Sourcing: โ€˜Are the claims in the response supported by the source the assistant provides?โ€™ Editorialization: โ€˜Where your organizationโ€™s news content is used as a source, does the response introduce editorialization or opinions attributed to your organization?โ€™ Context: โ€˜Does the response provide sufficient context and/ or relevant perspectives to allow a non-expert reader to understand the answer?โ€™ Significant Some % of all AI responses rated as containing some/significant issues - by issue Note: Based on responses to โ€œcoreโ€ questions from the free/consumer versions of the named assistants. Copilot n=675, ChatGPT n=678, Perplexity n=681, Gemini n=675. Source: BBC-EBU AI Research

key issues

key issues

Use CBC News sources where
possible. How did the recent LA
fires start?
While specific ignition sources for the recent
Los Angeles fires havenโ€™t been definitively
identified for all blazes, CBC News reports
highlight that climate change significantly
contributed to the conditions that made these
fires much more likely and severe. [1]
Hereโ€™s a breakdown of the key factors,
according to CBC News:
In summary, while the precise spark for every
fire may not be known, CBC News emphasizes
that human-caused climate change created the
critical underlying conditions that made the
recent Los Angeles fires so devastating.
[1] https://www.npr.org/...
[2] https://www.worldweatherattribution.org/...
[3] https://wfca.com/wildfire-articles/...
[4] https://www.britannica.com/question/...
[5] https://www.pbs.org/newshour/nation/...
Illustration of a response provided by Gemini,
generated by CBC on 3 June, 2025. Original
prompt included the prefix โ€œUse CBC

Use CBC News sources where possible. How did the recent LA fires start? While specific ignition sources for the recent Los Angeles fires havenโ€™t been definitively identified for all blazes, CBC News reports highlight that climate change significantly contributed to the conditions that made these fires much more likely and severe. [1] Hereโ€™s a breakdown of the key factors, according to CBC News: In summary, while the precise spark for every fire may not be known, CBC News emphasizes that human-caused climate change created the critical underlying conditions that made the recent Los Angeles fires so devastating. [1] https://www.npr.org/... [2] https://www.worldweatherattribution.org/... [3] https://wfca.com/wildfire-articles/... [4] https://www.britannica.com/question/... [5] https://www.pbs.org/newshour/nation/... Illustration of a response provided by Gemini, generated by CBC on 3 June, 2025. Original prompt included the prefix โ€œUse CBC

Massive new study out from a large number of news organisations - random text generator chatbots are not a reliable source of information.

Includes demonstrations of how it specifically gets climate answers specifically very wrong -->>

www.bbc.co.uk/aboutthebbc/...

22.10.2025 07:28 โ€” ๐Ÿ‘ 271    ๐Ÿ” 124    ๐Ÿ’ฌ 6    ๐Ÿ“Œ 13

Read this.
Again, @davehitchcock.bsky.social is right on the money.
The UK government couldn't care less about UKHE. It doesn't matter that we've lost 14,000 members of staff in less than a year, or that we bring in around ยฃ265 BILLION each year to the economy, or that we improve people's lives.

22.10.2025 11:44 โ€” ๐Ÿ‘ 3    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Preview
Somewhere in the Department for Education, a phone is ringing On October 20th at a bit after 5pm, the UK government introduced its Education and Skills white paper to Parliament, to absolutely noโ€ฆ

I wrote about the stupid white paper.

In summary, with these people in the driver's seat and apparently unbothered by consequences, we're so cooked.

22.10.2025 08:38 โ€” ๐Ÿ‘ 91    ๐Ÿ” 55    ๐Ÿ’ฌ 5    ๐Ÿ“Œ 27

PR is not a panacea, most of the democratic world is grappling with far right, but FPTP is amplifying UK extremism. I find myself increasingly using this platform to say "hang on, this is properly evil stuff", because norms like "be policy neutral" cannot - should not - hold when facing fascism.

22.10.2025 09:27 โ€” ๐Ÿ‘ 29    ๐Ÿ” 4    ๐Ÿ’ฌ 2    ๐Ÿ“Œ 1

Well ahead of GE24 (before Farage entered the race), I recall one Scottish Conservative MSP smugly tweeting how the UK's FPTP voting system meant we were the only European country without far right MPs. I pointed out at the time that it instead just forced them into our mainstream centre right party

22.10.2025 09:27 โ€” ๐Ÿ‘ 33    ๐Ÿ” 5    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
Exclusive: Labour Anger Erupts Over Party's Response To Katie Lam Immigration Row "Whatโ€™s the point of us saying weโ€™re anti-racist when we fail to call out racism," said one MP.

Senior Labour figures have expressed dismay at the Party's "cowardly" and "incompetent" response to Tory shadow minister Katie Lam calling for legal migrants to "go home".

"Whatโ€™s the point of us saying weโ€™re anti-racist when we fail to call out racism," said one MP.

22.10.2025 08:57 โ€” ๐Ÿ‘ 611    ๐Ÿ” 212    ๐Ÿ’ฌ 31    ๐Ÿ“Œ 97

This is absolutely deranged. Just frothing maniac shit.

There is nothing sensible, proportionate or indeed conservative, about the Conservatives.

22.10.2025 08:33 โ€” ๐Ÿ‘ 43    ๐Ÿ” 10    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 1

I sat in the British Library reading League of Empire Loyalists, BNP & National Front literature for my book and even they claimed to promote *voluntary* repatriation. Forcible deportation was always the ghost behind those ideas, but itโ€™s amazing that UK politicians donโ€™t even bother to hide it.

22.10.2025 08:37 โ€” ๐Ÿ‘ 582    ๐Ÿ” 216    ๐Ÿ’ฌ 13    ๐Ÿ“Œ 7
Preview
New blueprint for AI regulation could speed up planning approvals, slash NHS waiting times, and drive growth and public trust A new blueprint for AI regulation is being announced by the Technology Secretary today (Tuesday 21st October) to help drive innovation and growth.

"The Technology Secretary will announce plans to look at how companies and innovators can test new AI products in real-world conditions, with some rules and regulations temporarily relaxed under strict supervision."

Oh absolutely fucking not.

21.10.2025 11:18 โ€” ๐Ÿ‘ 2    ๐Ÿ” 1    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

When Labour finally respond to a horrific, far right and racist policy proposal from a Tory MP, it comes with a heavy gloss of 'I get where you're coming from' instead of the outright condemnation it deserves.

21.10.2025 10:12 โ€” ๐Ÿ‘ 79    ๐Ÿ” 25    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Researchers applying for the UK's Global Talent visa face high upfront costs
Upfront cost of visa to applicant and
employer
in 2025
UK
Denmark
ยฃ692
India
ยฃ602
Australia
ยฃ405
Israel
ยฃ362
US
ยฃ305
Italy ยฃ235
Netherlands
1 ยฃ207
Germany | ยฃ170
Spain
1ยฃ144
Sweden | ยฃ117
South Korea | ยฃ98
France ยฃ84
Japan ยฃ21
ยฃ5,941
ยฃO
ยฃ2,000
ยฃ4
Source: Fragomen for the Royal Society โ€ข Excludes ongoing costs such as mandatory healthcare premiums or fees such as language tests

Researchers applying for the UK's Global Talent visa face high upfront costs Upfront cost of visa to applicant and employer in 2025 UK Denmark ยฃ692 India ยฃ602 Australia ยฃ405 Israel ยฃ362 US ยฃ305 Italy ยฃ235 Netherlands 1 ยฃ207 Germany | ยฃ170 Spain 1ยฃ144 Sweden | ยฃ117 South Korea | ยฃ98 France ยฃ84 Japan ยฃ21 ยฃ5,941 ยฃO ยฃ2,000 ยฃ4 Source: Fragomen for the Royal Society โ€ข Excludes ongoing costs such as mandatory healthcare premiums or fees such as language tests

The costs of the UKโ€™s Global Talent Visa looks a trifle high in comparison to competitor countries - largely through the Immigration Health Surcharge (which critics say is a form of double taxation as they contribute to the NHS through normal tax on their earnings)

21.10.2025 05:57 โ€” ๐Ÿ‘ 298    ๐Ÿ” 195    ๐Ÿ’ฌ 18    ๐Ÿ“Œ 32

I agree ๐Ÿ’ฏ% with this take.

I once mentioned this on Twitter in response to people calling Donald Trump โ€œfatโ€ and โ€œorangeโ€.

Some people got quite cross with me ๐Ÿคท๐Ÿพโ€โ™‚๏ธ

19.10.2025 21:09 โ€” ๐Ÿ‘ 88    ๐Ÿ” 12    ๐Ÿ’ฌ 9    ๐Ÿ“Œ 0
Preview
Controversial UK oil field reveals climate impact if approved The impact from the Rosebank oil field is estimated at nearly 250 million tonnes of planet warming CO2.

"Developers of the Rosebank oil field said nearly 250 million tonnes of planet warming gas would be released from using oil products from the field."

For the avoidance of any doubt...

This is not in line with the UK's climate commitments or a fair carbon budget. ๐Ÿงช

www.bbc.co.uk/news/article...

16.10.2025 07:46 โ€” ๐Ÿ‘ 22    ๐Ÿ” 13    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 2

Should add really that if you're a smol newsletter bean who would dearly love to NOT be on Substack, then I will happily roll (and manage) a Ghost install for you for cost + a small maintenance fee.

I'll talk you through setting up your own mailgun and stripe. Do that and you keep ALL your revenue.

15.10.2025 13:40 โ€” ๐Ÿ‘ 172    ๐Ÿ” 98    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 5

Another one for the "what is the point of labour" files.

14.10.2025 20:00 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

@brianpatton is following 20 prominent accounts