@Yoshi This will not change the #enshittification, on the contrary, so I don't know why you abuse my post to place a promotion for an #AI corporation!

#GenerativeAI is a huge problem for #ecology and the #ClimateCrisis, independent if you run something in the cloud or your local machine.

These informations are about Google (back to my topic) but the ecological imprint is as bad for other companies:

▶️ https://www.theguardian.com/technology/article/2024/jul/02/google-ai-emissions
▶️ https://www.businessinsider.com/google-water-use-soaring-ai-make-it-worse-data-centers-2023-7

#water#drinkingWater#energyWaste

So, this got kind of big.I wrote a more detailed and thought out post about this topic on my blog. Comments are open and you're welcome to add your thoughts there.Also, if your kids' teachers do something like this, be kind. Take time to talk with them. Don't berate. We all work hard to make sure kids are taught well.https://blog.ohheybrian.com/2024/05/ai-citations/#AI#LLM#education#teaching

Despite what some may claim, "over-productivity" in science is not new:

"doubling time of 15-20 years had been maintained for centuries"

and

"Scientists have always felt themselves to be awash in a sea of scientific literature that augments in each decade as much as in all times before"

in:

"Absurdities, Ironies, Hyper-Production and Artificial Bogeymen"
https://zenodo.org/records/12746889

#ai #science

(...) the strategic use of AI's controversiality threatens the role of public controversy as a force of the #democratisation of research and innovation, and risks to undermine its capacity for inclusion and problematisation. (...) This situation not only has implications for our understanding of #AI controversies as occasions for the democratisation of knowledge; it also has consequences for the methodology of #controversy #analysis

https://journals.sagepub.com/doi/10.1177/20539517241255103

If you've received #NGIZero funding then you need to pay attention to this:

https://kolektiva.social/@hackstub/112762068812732277

@nlnet has funded many #fediverse apps and other awesome software for the commons.

What's disappointing is that Horizon has chosen to remove the cascade funding to put more funding to #AI. Which already has several AI funds.

This will in no way help small #FOSS development. I like @NGIZero because it funds some really unique ideas, that come from a diverse community. AI ain't that.

Next was a fantastic pair of talks at NBER by Moritz Lubczyk (gender bias and the misallocation of talent) and Robert Johannes (construction technology and medieval urban development). Johannes' talk was a mind blowing effort, combining archaeological records, geological data, and historical construction records to quantify the effect of the diffusion of brick technology in Germany (starting around 1150) on a whole variety of outcomes. Highly recommend https://www.youtube.com/watch?v=9OQQDvsO5sw (5/7) #economics
Next was an essential talk by @dhadfieldmenell on the challenges of specifying goals for algorithms at the Allen Institute for AI. Hadfield-Menell expertly illustrates why reward systems tend to be brittle due to difficulties in measuring/specifying everything using practical, theoretical, and empirical approaches. He concludes by describing the issue with aligning goals with those of a single individual, advocating strongly for inclusion. Highly recommend https://www.youtube.com/watch?v=ZmhmQzmK3OQ (6/7) #AI
Next was an interesting talk by Alessandro Lenci on why LLMs don't meaningfully understand language at UQAM https://www.youtube.com/watch?v=-ukVtAFQ83U (4/7) #AI #linguistics
First was an excellent talk by Kyle Mahowald on using language models for #linguistics at UQAM | Université du Québec à Montréal. LLMs are considered to be very good at syntax and grammar, but is this due to memorization, heuristics, or building up a decent model of language? Through a systematic review controlling training data on smaller models, Mahowald convincingly demonstrates that it's the latter, exploring implications for linguistics. Highly recommend https://www.youtube.com/watch?v=z8SwDlv0jx4 (2/7) #AI
@rachelcoldicutt @mia

> He asked DeepMind founder Demis Hassabis, “If you were a government today why would you not want to have all of your data in a cloud-based infrastructure so you can analyse it with an LLM?”, a question that even Hassabis had to swerve, no doubt mindful of the security and accuracy implications of such a move and the enormous amount of data cleaning, infrastructure management and optimisation that would be required.

#infrastructure#Policy#AI #security

@rachelcoldicutt @mia

Added this piece from Anna Dent to the reading list: https://www.computerweekly.com/opinion/Automating-Public-Services-A-careful-approach

> Technological progress happens in tandem with societal change; as my colleague Anna Dent set out yesterday in Computer Weekly, public trust is essential to deliver this well -- particularly for a government with such a wide and shallow mandate. 

#infrastructure#Healthcare #ai#Policy #trust

@rachelcoldicutt @mia

> He asked DeepMind founder Demis Hassabis, “If you were a government today why would you not want to have all of your data in a cloud-based infrastructure so you can analyse it with an LLM?”, a question that even Hassabis had to swerve, no doubt mindful of the security and accuracy implications of such a move and the enormous amount of data cleaning, infrastructure management and optimisation that would be required.

#infrastructure#Policy#AI #security