
By @jamesks to support courses on data, internet and society at Edinburgh Uni. Focus on mainstream news on controversies in AI, Internet culture, Policy etc.
Edinburgh, Scotland
Joined October 2018
- Tweets12.3K
- Following1K
- Followers731
- Likes808
datacontroversies@UoE retweeted
Massive and critical data leak of Tea app exposes 72,000 user verification images including driver's licenses, selfies, and ratings/comments about men. This breach enables identity theft and doxxing of both rating women and rated men (the app lets users create public profiles for men where anyone can post information - truthful or not - without the target's knowledge). The exposed database lacked basic authentication making the database of private data effectively publicly accessible. Map of driver licenses below
Linked Data: Unlocking public good from citizens personal information with Dr Allison Kurpiel, Dr Ana Morales-Gomez, Dr Magdalena Getler @SmartDataFdry
28 Feb, 3pm - 5.30pm Part of CONTROVERSIES in the DATA SOCIETY Seminar Series
Hybrid seminar bit.ly/CDS-Unlocking
2
81
Can AI civic technology invigorate democracy in Scotland? 15:10 Friday 7th Feb - two talks and discussion with Dr Stuart Lynn and Mr Iñaki Goñi as part of the Controversies in the Data Society Series. In person or online. sps.ed.ac.uk/news-events/ @ScotGovEngage
1
46
Controversies in the Data Society Full Programme now available! We have a fantastic lineup of speakers and guests addressing topics from egovernment and digital democracy to climate change resilience and creative authenticity. eventbrite.com/cc/controvers…
And the full Programme Website sps.ed.ac.uk/science-technol…
10
datacontroversies@UoE retweeted
Today, the UK government announced a proposal to change copyright law - for the benefit of AI companies - that would cause huge, irreversible harm to creators.
More info below, but most importantly here's what you can do (wherever you live):
1. Email your MP. If you're in the UK, here's a template letter you can copy - docs.google.com/document/d/1… - and you can find your local MP's details here - members.parliament.uk/member….
2. If you're a creator, email your representatives. Your publisher, record label, union - whoever represents you. Here's a second template letter you can copy - docs.google.com/document/d/1…. You need to know your representatives are representing your views.
3. Respond to the consultation. You can do so by emailing [email protected]. If possible, write your own response, and go into all the detail you can. Feel free to use info / data I've prepared here - docs.google.com/document/d/1….
4. Share these template letters with anyone you know. The more people get involved, the more the government will get the message that a broad copyright exception is the wrong path to be pursuing.
--
And here's a summary of the government's proposal, and why I think it's so problematic:
- Broad new copyright exception for commercial generative AI training. AI companies will be able to train on British copyrighted work without a licence, even if the AI model is designed to compete with the creators whose work is trained on. This would make the UK one of the most punitive jurisdictions for creators in the world.
- Rights holders can 'reserve their rights', i.e. opt out. But opt-outs don't work (you can't successfully opt out downstream copies of your work), most creators miss the chance to opt-out, doing so is a huge admin burden, etc. AI companies should be getting opt-in consent - it's unfair to shift the burden to creators.
- AI companies must offer some level of transparency over their training data. This would be good if presented on its own, but it's much less helpful if you're packaging it up with a broad copyright exception that lets AI companies train on most of the UK's creative output with impunity.
The consultation on these proposals lasts for 10 weeks. Anyone who cares about this issue should do whatever they can to make their views known to government now - there will only be one chance.
datacontroversies@UoE retweeted
Interesting small-scale study on accuracy of diagnosing illness:
- Human doctors: 74%
- Human doctors using ChatGPT: 76%
- ChatGPT alone: 90%
Takeaway seems like vast potential for AI to help with diagnosis, but need better human <> AI teamwork:
nytimes.com/2024/11/17/healt…
Try following me on the other one bsky.app/profile/datacontrov…
24
cloud.google.com/blog/topics… @JohnnyHistone
19
Meta sees AI spending accelerating as earnings top forecasts - on.ft.com/3YIxast via @FT
19
Palantir becomes a ‘Trump trade’ as investors bet on higher defence spending - on.ft.com/3Z0Z2Hp via @FT
39
AI and globalisation are shaking up software developers’ world
economist.com/business/2024/…
from The Economist
17







