John Humphreys @johnpharmd@qoto.org

'There is a health data “gold rush” underway, Elashoff observes, and a key goal of the big tech companies is to mine these data to produce marketable AI-related technologies—not just for the healthcare technology market but also for a myriad of other massively lucrative AI markets'

https://www.nature.com/articles/s41591-023-02290-y

"An Alzheimer's disease-specific pattern of cytokine secretion reduces expression of mitochondrial electron transport complexes and impairs mitochondrial respiration in healthy neurons."

#alzheimers #Preprint

https://www.biorxiv.org/content/10.1101/2023.04.07.536014v1

A market-leading garage door controller is so riddled with severe security and privacy vulnerabilities that the researcher who discovered them, Sam Sabetan, is advising anyone using one to immediately disconnect it until they are fixed.

Each $80 device, used to open and close garage doors and control home security alarms and smart power plugs, employs the same easy-to-find universal password to communicate with Nexx servers. The controllers also broadcast the unencrypted email address, device ID, first name, and last initial corresponding to each one, along with the message required to open or shut a door or turn on or off a smart plug or schedule such a command for a later time.

The result: Anyone with a moderate technical background can search Nexx servers for a given email address, device ID, or name and then issue commands to the associated controller. (Nexx controllers for home security alarms are susceptible to a similar class of vulnerabilities.) Commands allow a door to be opened, a device connected to a smart plug to be turned off, or an alarm to be disarmed. Worse still, over the past three months, personnel for Texas-based Nexx haven’t responded to multiple private messages warning of the vulnerabilities.

“Nexx has consistently ignored communication attempts from myself, the Department of Homeland Security, and the media,” Sabetan wrote in a post published on Tuesday. “Device owners should immediately unplug all Nexx devices and create support tickets with the company requesting them to remediate the issue.”

Sabetan estimates that more than 40,000 devices, located in residential and commercial properties, are impacted, and more than 20,000 individuals have active Nexx accounts.

https://arstechnica.com/information-technology/2023/04/open-garage-doors-anywhere-in-the-world-by-exploiting-this-smart-device/

Very very high recommendations for this brilliant podcast: Between Two Nerds.

Easily the best discussion between myself and @tomatospy released this week.

https://risky.biz/BTN30/

We examine TikTok from our unique and insightful perspective. Available now on all fine podcast apps.

Here's my latest weirdly specific GPT-4 enhanced project: we wanted to measure the temperature of a microwave Raku kiln (yes, that's a thing - talk to @natbat about it) over time without tediously watching the thermometer for hours... so instead we recorded a video of the thermometer then used ffmpeg and Google Cloud Vision to OCR readings from it into a database https://til.simonwillison.net/googlecloud/video-frame-ocr

Humans vs. machines: the fight to copyright AI Art.
@fnwdhs https://www.reuters.com/default/humans-vs-machines-fight-copyright-ai-art-2023-04-01/

One of the main challenges for new servers on here is discovering content to interact with.

Relays speed up the discovery process, and allow even a single user server to automatically see a large part of the Fediverse.

There's a new relay service by @astro­@c3d2.social called FediBuzz Relay:

➡️ https://relay.fedi.buzz

It allows servers to subscribe to custom relays based on tags or instances.

If you use this, can you let us know your experiences in the replies?

#FediAdmin #MastoAdmin

“If the data is wrong, the stories will be wrong, too. As Mitch Ratcliffe said, ‘A computer lets you make more mistakes faster than any invention in human history, with the possible exceptions of handguns and tequila.’”

@mikestucka explains the mad libs wizardry he did at USA TODAY

https://source.opennews.org/articles/choose-your-own-mad-libs-automated-data/

My thread from last night on the hypey "open letter" as a blog post:

https://medium.com/@emilymenonbender/policy-makers-please-dont-fall-for-the-distractions-of-aihype-e03fa80ddbf1

#LLM #AIhype #GPT4

@aoc is right. A TikTok ban wouldn’t protect Americans’ privacy because the same kind of data that TikTok collects can easily be purchased from data brokers. A comprehensive privacy law *would* protect Americans’ data—and wouldn’t require a ban. https://www.tiktok.com/@aocinthehouse/video/7214318917135830318?_r=1&_t=8awLpxdw8ZG

I have been asked a few times how I map my #CrossBorderRail routes

I have now written it up 👇
https://crossborderrail.trainsforeurope.eu/mapping-my-crossborderrail-projects/

In short:
- uMap to combine it all
- Pierre Beyssac's tool for the rail routes
- Komoot for the bike legs
- Google Maps for the road transport
- OpenRailwayMap for information
(and some geojson and GPX exports to make it all work)

If my grandfather had managed to get his story "Barnwoggler's Invention" published in 1926, he would have set the scene for ChatGPT and the rest a century later. But in the story it all actually works: Barnwoggler builds a machine that can generate publishable academic prose on any subject. https://www.wallandbinkley.com/rcb/2021/12_31_barnwoggler-redux-a-1920s-vision-of-library-and-research-automation.html#barnwogglers-invention #RobertCBinkley #AI #DigitalHumanities

The inner homunculus

"Penfield found those visceral signals harder to decipher than the brain's map of the body's surface. Brain regions responsible for different internal sensations seemed to overlap. Sensory regions were hard to distinguish from those that sent motor instructions such as telling the intestines to contract. Penfield once asked participants to swallow an electrode to detect changes in gut contractions while he stimulated their brains. But his map of the inner organs was blurry and ambiguous—and stayed that way for most of the next century."

https://www.science.org/content/article/newly-detailed-nerve-links-between-brain-and-other-organs-shape-thoughts-memories-and?utm_campaign=SciMag&utm_medium=ownedSocial&utm_source=LinkedIn

Anthropic introduces Claude, a “more steerable” AI competitor to ChatGPT https://arstechnica.com/?p=1924161

"There was a very strong influence of the computational biology community that was already used to sharing code and the concept of open source code was not at all novel."
#EMBOPodcast

@tlemberger on the creation of the first #OpenAccess at EMBO, @MolSystBiol

https://www.embo.org/podcasts/a-steppingstone-to-an-open-science-future-embo-press-moves-to-full-open-access/

Early thoughts on regulating generative AI like ChatGPT

https://www.brookings.edu/blog/techtank/2023/02/21/early-thoughts-on-regulating-generative-ai-like-chatgpt/

Since I wrote that on Saturday the LLaMA large language model has been shown running on a 4GB RaspberryPi, and this morning on a Pixel 6 phone! Added a "It's happening" section to the post here: https://simonwillison.net/2023/Mar/11/llama/#its-happening

The shocking profiteering of Elsevier...

The Cost of Elsevier
https://wp.me/pko9D-eVx

The academic humblebrag:

"Yay, we got a grant! Thanks to the funder and all the fantastic people in our lab that made this possible! We will be recruiting soon for this awesome project"

What it really means:

"Apologies to all the brilliant ECRs out there with their own novel research ideas, but due to intramural underfunding, we need to force you to work on my own mediocre ideas I had to waste precious weeks on selling to the reviewers of extramural funders."