This is why I think AI companies are self-defeating in their opposition to real government regulation, with teeth, of AI
If the US had rights- and risk-based AI regulations like in Europe, then people might feel that someone had their back, and they might be more willing to trust AI
This impressive work doesn't just cast doubt on AI benchmarks, it has dire implications about the whole project of AI alignment and safety.
If Dawn Song's team can so easily create an AI agent that can circumvent the protections around carefully engineered high-visibility benchmarks, then it […]
It seems likely that emotion detection will eventually be pretty effective, as extension of existing facial recognition and voice recognition technologies
But it's very ethically problematic
The EU AI Act bans it in employment and educational settings
I certainly hope it never is widely used
Congratulations, and thank you for this valuable work!
The US wants to pull the whole world down to its low level (often non-existent) protections of privacy and AI safety
www.reuters.com/sustainabili...
Two baby goats snoozing
Kid nap
Snooze time for the new baby goats nextdoor
JTAG, wow. That's a blast from the past
I worked for Rod Tulloss at AT&T Bell Labs back in 1990 when he was one of the people leading the adoption of the JTAG standard
It's nice to hear that it's still useful
Other alternative terms I'd propose to add:
contextual weighting (not attention)
pattern extrapolation (not hallucination)
pattern matching (not understanding)
iterative LLM (not thinking)
LLM workflow (not AI agent)
supervised labeling (not teaching)
context buffer (not memory)
Interesting paper proposing how we can de-anthropomorphize discussions of probabilistic automation systems (not AI) by using terms like:
errors (not mistakes)
text input (not prompt)
output (not answer)
conversation simulator (not chatbot)
weighted (not neural) networks
reflects (not shows) bias
It's depressing to consider that Gramsci's analysis from his time applies to ours
"the old world is dying, and the new world struggles to be born: now is the time of monsters"
But he also called for optimistic acts in the face of this reality
"Pessimism of the intellect, optimism of the will"
Unfortunately, I don't think this is technically possible
Bluesky is built on AT Proto which is designed as a distributed system with no one organization controlling the data
That means *any* organization can see the data
This is even more interesting than I thought
That's yet another layer of AI craziness
The Ars Technica article itself may also have been AI generated
The article was pulled, which is why the link is broken
infosec.exchange/@mttaggart/1...
Screenshot from my news feed
My news feed contained this item
It was about the behavior of an AI agent
And the illustration is, I assume, AI generated
And the news feed item was itself AI generated
But the link to the story on Ars Technica is broken, yielding a 404 error
So, AI all the way down yielding…
Nothing
Screenshot of where my account is in this space
My placement is a little surprising
Off in intergalactic space near some galaxies that are not really how I think of myself
Very cool
Did you give equal weight to followers and following?
Giving more weight to following may give more of a map of how people think of themselves
while giving more weight to followers might give more of a map of how people are perceived by others
Elon Musk vs. the laws of thermodynamics
I wonder who will win?
Though if we see a lot of things like this (which are more likely to actually work on the real threat model), then we know that the public is really concerned
bsky.app/profile/auti...
Good point
The fact that this company seems to think there is a market for this is interesting in itself
It may indicate a general growth in awareness of the privacy threat of facial recognition
These will only work on IR-based systems in personal security devices (phones, computers) and access control systems
But it will be largely ineffective against mass public surveillance, which generally uses optical-based systems
Happy Imbolc, the celebration of the day halfway between the winter solstice and the spring equinox
It marks the beginning of Spring in Ireland
The ancient Celts were a bit pedantic about their solar astronomy, and didn't take account of thermal inertia which actually delays seasons by a month
Wow, appreciating this thread requires understanding a very specific combination of very online pop culture and political references from both Korea and Ireland
Starlink user data trains Grok AI by default
If you use Starlink you might want to prevent that
You can change the option under
= > Profile > Privacy
Remember: LLMs and bots and AI-generated things should always be called ‘it’. Regardless of the style or tone of voice, there is only an ‘it’ there. Don’t humanize it.
If you're interested in AI policy you might find my new AI Policy feed useful
It's a filtered feed of BlueSky post relevant to AI policy and governance around fundamental rights, democratic institutions, and the rule of law
Click through and add it to your list of feeds
bsky.app/profile/did:...
Depending on their language's relationship to sacred cultural practices and the soul wound of historical loss, different indigenous peoples have different views of outsiders learning it
I believe Ojibwe, Choctaw, and Cherokee (in addition to Nahuatl and Navajo) do welcome respectful learners
I think you just change velocity and reliability are the main things you want to measure
Then expect all devs to spend some fraction of their time using their best judgement and subjective skills to continuously improve the code base -- which hopefully will case velocity or reliability to improve
American exceptionalism: pay twice as much for health care and still die 5 years sooner
Life expectancy vs. health expenditure before, during, and after COVID for selected countries
ourworldindata.org/grapher/life...
I'm particularly interested as I'm working on safety protections of a project to add personalization to a knowledge-seeking AI system
Personalization is arguably a kind of persona drift; it makes the system more aligned with the end-user
So we would need to distinguish safe drift from unsafe drift
