CriticalAIhttps://read.dukeupress.edu/critical-ai/
@CriticalAI
Read our new issue on LLMs, gen AI, and rise of chatbots! Follow us on Mastodon + Bluesky where our editor now posts.
We'll do a more full-throated introduction to our just-published issue in due course. But for now: here 'tis. Congrats to editors and all contributors. Please disseminate widely as free access disappears after the first few weeks! read.dukeupress.edu/critical-ai/is…
Happy to repost this wonderful grouping. We hope that they-and you-are also reading peer-reviewed interdisciplinary scholarship in #criticalAIstudies. read.dukeupress.edu/critical-ai/is…
This looks potentially interesting (though none of us has as yet read it).
The paper discusses the growing challenges of the "memory wall" problem, where the speed of memory and communication is not keeping up with the exponential growth in compute power. This is becoming a major bottleneck for training and serving large AI models, particularly in the…
Sneak preview of an article about Eryk’s, Caro’s and my work — to be published soon in @CriticalAI criticalai.org/2025/01/17/sne…
The UK's Intellectual Property Office has tweeted about the government's proposals, which would allow AI companies to train on copyrighted work without a licence. They've asked for your views, so I encourage you to reply to their tweet.
The UK government AI and Copyright Consultation opens today. We want to hear from a wide range of voices. Respond to the consultation to share your views. [Link]: gov.uk/government/con…
This development illustrates that the ethical issues around the building of AI models designed to produce derivative text &/or images isn’t limited to the issue of consent. We need to ask ourselves broader questions about AI production & the impact on creators & society. @Knibbs
Harvard says it's releasing a high-quality dataset of ~1M public-domain books, created with funding from Microsoft and OpenAI, to help train LLMs and AI tools (@knibbs / Wired) wired.com/story/harvard-… techmeme.com/241212/p1#a241…
Check out Zachary Zimmer's preprint review of Fei-Fei Li's memoir on #CriticalAI's blog! criticalai.org/2024/11/08/sne…
Essential reading. @ednewtonrex forcefully & thoughtfully sets out why societies should adopt (or maintain) legislative frameworks that require AI companies to affirmatively obtain consent prior to using creative works to train AI models. 🙏🏼⚖️ ed.newtonrex.com/optouts
I will humbly note that I argued back in April that the AI promises were — in the truest sense of the term — smoke and mirrors bloodinthemachine.com/p/ai-really-is…
Multiple sources are now reporting LLMs aren't scaling as hoped—larger datasets and more compute aren't improving AI systems as fast. The companies, naturally, are pressing on, calling for billions more in investment. One way to read this: The hope is to make AI too big to fail.
Not directly AI-related but relevant to its ecosystem. MSM bake in narratives before adding all the ingredients--like the most populous state in the country! thenation.com/article/politi…
Just learned for the first time about this Princeton response to the infamously crummy MSFT puffpiece, "Sparks of AGI." Haven't read it yet but the title alone is well worth the price of admission. Embers indeed! arxiv.org/abs/2309.13638
As OpenAI and Meta introduce LLM-driven searchbots, I'd like to once again remind people that neither LLMs nor chatbots are good technology for information access. A thread, with links: >>
The election debacle cannot be entirely blamed on the failures of our media system, but such long-standing systemic pathologies certainly played a key role. Likewise, rebuilding a vibrant information infrastructure must be a central piece of a broader re-democratization project.
The other big loser Tuesday was the mainstream media, which has all but imploded in public trust and as a force in American democracy. Trump won with the most disinformed electorate ever The battle to create a world for truth starts now. My new column🎁share.inquirer.com/bpjtJ3
Very happy to share this worthy enterprise from @RosenzweigJane. Looking forward to reading! A good opportunity to say that in the future #criticalAI will be spending less time here and more on the blue space. Please find us there...
great to be part of this important conversation on AI & democracy at the Council of Europe in Strasbourg & tomorrow I'll address the Council on pervasive data harvesting practice & its insidious impact on freedom of speech/movement using IDF's "Lavander" as an extreme example
"We all want #AI to benefit society but there is a massive gap between wanting that and what happens on the ground [...] impact of technologies is not new, but what we tend to forget is that [...] these technologies harm the most #DisadvantagedMinorities" - @Abebab. #CoE_WFD
AI policy is, at present, way too dependent on a cooperating executive branch. Part of this trend was pragmatic (ie agencies hold the tech expertise, legislation is slow & difficult, etc), but it's exhausting to reset & re-strategize at the start of every new administration.
The Biden-Harris Administration didn't rescind Trump's OMB guidance that Adam mentions here, nor did it rescind Trump's Feb 2019 or Dec 2020 executive orders on AI So there's a real tradition of bipartisanship on AI that Trump could build on if he wanted to
I'll go further and say that @Adobe (which I use primarily for marking up pdfs) is in full-on enshittified. It's aggressive preference for its own cloud platform is invasive (I struggle to save work to my hard drive.) And the AI button gets in the way of EVERYTHING. 🦹
Personal rant but I’m still bummed that Adobe decided to turn away their backs on the artists and creatives it relied upon to go full throttle on GenAi, a tech that is intent on replacing us. It’s also wild to me Adobe is not even offering a version without any GenAi at all.
Pleased to be named a ✨Global AI & Market Power Fellow ✨ by the European AI & Society Fund, with $68k to support my work on corporate capture in AI, alongside a fantastically brilliant cohort of other fellows: europeanaifund.org/newspublicatio…
Are we still pretending this is "news", at this point?
None of us as read this one yet, but we really enjoyed Chayka's work on ChatGPT4o which was cited in the introduction to the latest issue. read.dukeupress.edu/critical-ai/ar…
"The main people benefitting from the launch of A.I. tools so far are not everyday Internet users (...) but those who are producing the cheap, attention-grabbing A.I.-generated content that is monetizable on social platforms." #AIEthics newyorker.com/culture/infini…
If you're taking away from last week's searchGPT hype that Google's dominance is on it's way out, AI Now's @Kabakab in @techpolicypress refocuses our attention to the entrenched (and interlinked) structural advantages Google holds across AI and search markets
Does generative AI mean Google’s search dominance on its way out? Pressure from AI companies would have us think so - but @kabakab argues in @techpolicypress that in the DOJ case Google’s advantages in search & AI remain the elephant in the room: techpolicy.press/the-elephant-i…