Nick @internic@social.cleverlibre.org

NFTs died a slow, painful death in 2023 as most are now worthless https://www.newscientist.com/article/2406198-nfts-died-a-slow-painful-death-in-2023-as-most-are-now-worthless/

It's the time for twinkly lights, wrapped presents, and the beheaded corpse of an esoteric half-goat demon. Let's talk about religious freedom and stuff! https://skepchick.org/2023/12/deck-the-skulls-of-half-goat-demons-fa-la-la-la-la/

Practically everyone in the lower 48 will be experiencing temperatures warmer than average in the days leading up to Christmas. Anomalies in the Midwest will be nearly 25 to 30°F or more

Time to put on some cheesy holiday movies and day dream about snowy landscapes I guess!

My article "Bias Optimizers"— about the workings, dangers, and potential remedies for large language models and GPT-type systems— is available at proquest; meaning there are all kinds of free login options:
https://www.proquest.com/openview/6030bf85f07428eefea4ebe55cce98f4/

Example: I had absolutely no optimism that anyone implemented contact discovery without, uh, storing all my contacts. And here Signal has been doing it all along?

This should be the norm. An OS-level feature, even. Tremendous dedication by the Signal team.

For 32 years, arXiv.org has brought scientific & academic papers to the world — they're a free distribution service and an open-access archive for nearly 2.4 million scholarly articles.

Today, we're helping you access their papers faster and safer, and we're helping #arXiv deal with the tidal wave of scrappers brought on by this #AI age.

Welcome to #FastForward ⏩

https://blog.arxiv.org/2023/12/18/faster-arxiv-with-fastly/

I am very sorry to report that this is the final edition of the free Cyber 202 newsletter from the Washington Post. It fell victim to a cost cutting initiative that has also seen the buyouts of major talents across the newsroom, including 202 authors Tim Starks and David DiMolfetta. Some cyber news will continue in the Technology 202, and subscriptions to the paper and website, which feature deep coverage by me and others, remain dirt cheap. https://www.washingtonpost.com/politics/2023/12/21/year-cyber/

@fuzzyface @makershour
I suppose you start small, share a project in first and second parts before a working together type of project.

I started the #MakersRelay on Twitter. I made a random thing, sent it on to another maker who added to it and sent it on, etc.

We had a list of makers but I think it died when Twitter died and I don't know where in the world it has got to now.

Want to see something cool? My team at @ProPublica took all of the Supreme Court justices' disclosures, cleaned them up, connected them, and put them in an easy-to-use database for you to search: http://projects.propublica.org/supreme-connections/

Curation/aggregation remains an amazingly vast -- and under-appreciated -- opportunity on the Internet .

Decentralized networks are giving smart folks a new opportunity to take advantage of the opportunity -- and we are all the beneficiaries.

Flipboard is doing some of the most interesting, and important, work in this genre.

Read this from @mike to understand this excellent new trend.

https://medium.com/@mmccue/flipboard-begins-to-federate-4a80d6bdc209

Morel mushrooms can be deadly https://www.scientificamerican.com/article/morel-mushrooms-can-be-deadly-food-poisoning-cases-show/

After questions from ProPublica, the #NYPD has agreed to end its practice of withholding body-cam footage of #police shootings from civilian investigators. Under the deal, police will hand over requested footage within 90 days.

It's essentially a good faith agreement that does not have the force of law or lay out penalties.

https://www.propublica.org/article/nypd-release-body-camera-footage-civilian-investigators?utm_medium=social&utm_source=mastodon&utm_campaign=mastodon-post

Face recognition technology “can turn into a way for the government or businesses to track us as we move about the world,” EFF’s @Adam_D_Schwartz warns in the @washingtonpost. https://www.washingtonpost.com/travel/2023/12/20/clear-facial-recognition-technology-airport-security/

@tinyrabbit @gerrymcgovern Available at fine bookstores everywhere! https://www.acityonmars.com/

So, the LAION data set, used by pretty much every major diffusion model on the planet, has been found to contain tons of images of child sex abuse.

https://fortune.com/2023/12/20/top-ai-image-generators-trained-child-sexual-abuse-stanford-internet-observatory/

It's previously been found to contain images of pornography, rape, and violence https://arxiv.org/abs/2110.01963 and of private medical data https://arstechnica.com/information-technology/2022/09/artist-finds-private-medical-record-photos-in-popular-ai-training-data-set/

Maybe this is the third strike that gets people to stop using this data set and the models built using it? Right? This must be where people in tech draw the line?

No?

"12 climate change books to give friends and family over the holidays" by Michael Svoboda for Yale Climate Connections: https://yaleclimateconnections.org/2023/12/12-climate-change-books-to-give-friends-and-family-over-the-holidays/

Need a last minute Christmas thing for a bookish kid? My favorite review for Bea Wolf was the kids who (I'm told) refused to sleep ever again unless their Dad kept reading it to them. Second-best is this review by an adult: https://afuse8production.slj.com/2023/09/27/review-of-the-day-bea-wolf-by-zach-weinersmith-ill-boulet/

I was challenged by my boss to try and find some valid use cases despite being very skeptical of the hype. There are some value-adds I've discovered, I don't doubt that the costs outweigh the benefits, but in situations where there is ground truth AIML can help out.

So far I've built a prototype code optimizer, a human writes readable and easy to maintain code, then the LLM is used to optimize it for performance in conjunction with symbolic execution to ensure functional equivalence.

Retrevial-augmented generation can be a useful way to make use of a database of true data. Take a bunch of trouble tickets for example, and then use the LLM to find relevant ones that might help in a new situation or query.

From Damien Desfontaines on LI: Scientists at the U.S. Census Bureau just published a 75-page paper giving an extremely detailed description of the reconstruction and reidentification attack they ran on the 2010 Census data https://www.linkedin.com/posts/desfontaines_scientists-at-the-us-census-bureau-just-activity-7142942238375911424-0Ahb

New: Rite Aid secretly scanned shoppers with facial recognition for years, leading to false shoplifting accusations, "embarrassment, harassment and other harm," the FTC says. The chain banned workers from talking about it, and errors were common; one 11-year-old girl was distraught after being searched due to a false match. “Every black man is not [a] thief nor should they be made to feel like one"

https://www.washingtonpost.com/technology/2023/12/19/ftc-rite-aid-facial-recognition/