Notices by danmcquillan (danmcquillan@kolektiva.social), page 2

"Med-PaLM [is] a large language model designed to provide high quality and authoritative answers to medical questions" https://youtube.com/watch?v=3Ud-BMOCkDI Pairing 'LLM' and 'authoritative answer' is silly under any conditions, let alone when dealing with people's health 🧑⚕️🤦♂️

"The greatest risk is that large language models act as a form of ‘shock doctrine’, where the sense of world-changing urgency that accompanies them is used to transform social systems without democratic debate". Read more of my evidence to Parliament at https://committees.parliament.uk/writtenevidence/124038/pdf/

source: https://twitter.com/jamestitcomb/status/1718271967617331466

A review of 'Resisting AI' https://blogs.lse.ac.uk/lsereviewofbooks/2023/10/17/book-review-resisting-ai-an-anti-fascist-approach-to-artificial-intelligence-dan-mcquilllan/ - "He might be right on this one" 😂

'ChatGPT use shows that the grant-application system is broken' https://www.nature.com/articles/d41586-023-03238-5 - I'd go further; that LLMs can fool bureaucracies with bullshit shows the role of bureaucracy has never been to properly order and regulate, but simply to obfuscate the location of real power

It's fair enough to demand safety standards, when self-driving cars are blocking ambulances and pinning down a crash victim. But it misses the point that AI is a priori unsuitable for this kind of application and should just be stopped. https://www.nytimes.com/2023/10/11/opinion/driverless-cars-san-francisco.html?unlocked_article_code=OnUSTOUYq_8-wUcaUzXC7DWIN3iM0eN3VFWt1xH4rjqIb_cMiR3Sq8TZ6uH1EBNdw7qOLzp0hGzNTKHdeyRnCmAj_8hEYg6odEMCcOU7RZXnQW5OjsoNGwURKIFtpym7aW4p8UHoakSE5YHl4M4p4i00h1laJHJ4iAt_dwjbXU3N6JYspnEKFC2K11Lq0qE0StaXNx-eUp5Ut8mme--gOmz2PQpiW4yqXpm7RTSEjRlUhqij3CkmGWRgxZFetgm9H7D0ysAGZOQHsNX_h75mGCkTBeROiuquYaCuZs2MF55d5Ao7cejiyjwKhEJCddVzBQ100AHOmel8uD3Qd8dMC1n9gBu3gQ&smid=url-share
#selfdrivingcars #autonomousvehicles

Not bullshit, but superbullshit: 'Introducing Superalignment' by
OpenAI https://openai.com/blog/introducing-superalignment - even the AGI fantasies are starting to feed on themselves...

WGA board member: "Right now workers all over the world are being replaced by AI... So I do hope that people are looking at this and realising what the power of solidarity looks like" #WGAStrike
Could have come straight out of Ch.6 of 'Resisting AI' https://bristoluniversitypress.co.uk/resisting-ai ✊

"The greatest risk is that large language models act as a form of ‘shock doctrine’, where the sense of world-changing urgency that accompanies them is used to transform social systems without democratic debate."
My evidence to the House of Lords Communications and Digital Select Committee inquiry on Large Language Models https://committees.parliament.uk/writtenevidence/124038/pdf/ #ai #uk #regulation #shockdoctrine #llm #parliament

"Rather than being an apocalyptic technology, AI is more aptly characterized as a form of supercharged bureaucracy that ramps up everyday cruelties" 'Resisting AI' p4 https://bristoluniversitypress.co.uk/resisting-ai
#chatbots #LLMs #resistingAI #bureaucracy

The abstractions of scale aren’t simply a model architecture or a server infrastructure, but a homogenising force that eliminates context; a technological terra nullius that legitimates expropriation.

Letting the grass grow over the data centres...

"with GPT-4, the errors are often much more subtle and require careful attention for an instructor or grader to recognize" https://www.oneusefulthing.org/p/the-homework-apocalypse
- when 'improvement' means more nuanced bullshit, gained at untold computing and human (clickworker) cost
#llms #generativeai #gpt4 #bullshit

The way the algorithms of AI operate internally make it a form of automatised conspiracy theory. See also Chapter 3 of 'Resisting AI' https://bristoluniversitypress.co.uk/resisting-ai

the nested climate harms of AI: 1) direct harms from intensified data infrastructure via water resource, emissions 2) use of AI to legitimate green tech solutionism 3) reinforcement of 'one world' modernist worldview over relational ontologies
#AI climatecrisis

new interview in computer weekly https://www.computerweekly.com/news/366537843/AI-interview-Dan-McQuillan-critical-computing-expert

seeing like a state machine

"UK government ‘hackathon’ to search for ways to use AI to cut asylum backlog" https://theguardian.com/uk-news/2023/apr/29/government-hackathon-to-search-for-ways-to-use-ai-to-cut-asylum-backlog
- condenses in one go pretty much all the AI-driven necropolitics that I warned about it 'Resisting AI' 🤖💀☠️ #AI #asylum #refugees #ResistingAI

Big thanks to Tim @MLStreetTalk for a wide ranging and philosophical discussion about the implications of AI. The full interview will be out later today
?? ?
https://www.youtube.com/shorts/-OBNUDCuvgQ

new blog post: 'We come to bury ChatGPT, not to praise it' https://www.danmcquillan.org/chatgpt.html #ChatGPT #ResistingAI #AI

the idea of a chatgpt consultancy is recursive; one mode of bullshit generation selling you another mode of bullshit generation