Nope, the news was just catching its breath. 14 articles in the last 3-4 days???
In my last Bullshit Apocalypse post, linking to "Employees quietly using unapproved AI tools at work, and security teams can't keep up", I thought about talking about putting the genie back in the bottle, but apparently forgot about it.
This 1st post by Bruce Schneier is a must-read. He explores the genie metaphor for "AI Agents". He introduces the term "pragmatic":
In linguistics and the philosophy of language, pragmatics is the study of how context contributes to meaning.From this came our title, "Nonpragmatic Genies". This will go into our table of names for GenAI. Hmmm, we might have to start a separate table for agents at some point.
He talks about developing the "Genie Coefficient":
The Genie coefficient suggests an AI analogue, where a user is accountable for the plain intent of what they asked the AI. If an AI system betrays the reasonable meaning of an instruction, that’s the AI’s misbehavior, not the user’s.No, no, no. The human is where the buck stops. "AIs" are known to be unreliable: LLMs at a 10% level, agents at a 10-15% level. If a human chooses to use this unreliable technology, it's on them, IMO.
Why AI Needs a “Genie Coefficient”Note, GenAI genies are not a problem for me. I will continue to do my best to avoid using this unreliable technology. I suppose I could feed it my blog posts & ask it to "improve" them. What would that do? Make my writings read like everyone else's? Or, given semantic ablation, make them say the same thing as everyone else's?Genies are now an engineering problem. We are handing them the keys to our inboxes, bank accounts, code repositories, and physical infrastructure. And we have no agreed-upon ways to measure how genie-like any AI system actually is.
Here's the comment I posted to the site:
"The Genie coefficient suggests an AI analogue, where a user is accountable for the plain intent of what they asked the AI. If an AI system betrays the reasonable meaning of an instruction, that’s the AI’s misbehavior, not the user’s."No, no, no!
The human is where the buck stops. "AIs" are known to be unreliable: LLMs at a 10% level, agents at a 10-15% level. If a human chooses to use this unreliable technology, it's on them, IMO.
I think the "mens rea" argument it totally specious. If a construction company builds a bridge with substandard steel, steel with a 10-15% collapse rate, & the bridge collapses, I don't think the company can say "Our intent was not for the bridge to collapse."
From the "Stephen Follows" substack:
What just happened to TheNumbers.com should worry us allThis is slightly afield, but, still, unbelievable! The world's most trusted site for movie box office data has been destroyed by what winds up being a DDoS done by bots employed by prediction site gamblers trying to cheat on their bets on movie box offices by hacking into TheNumbers.com website.The inside story of how one of film data's most trusted sites vanished overnight, and why every website you rely on is more fragile than you think.
When prediction markets 1st came out, I was really excited! Wisdom of the crowd, FTW! 2015. I asked a billionaire friend of mine what he thought. "It's just going to be another form of gambling." Man, he nailed that one!
From the Bard of the Revolution, Cory Doctorow. I liked when the "AI Snake Oil" guys renamed their blog to "AI as Normal Technology". AI as a normal technology has lots of use cases. Mostly only profitable when they serve as a tool to a creative human.
Turning it loose as anything other than a tool under close human supervision is going to be disastrous. Doctorow is down with this. The problem with "AI" is the hype, the bubble, the desperate attempt to keep Big Tech companies classified as growth stocks to keep their incredibly inflated P/E's.
AI solipsists and AI cynicsUgh. So the coiner of the word "enshittification" reminds us that all the AI chatbots we are going to be increasingly having a hard time avoiding are happening as "enshittification Phase II". :-(You don't have to believe that AI "works" to believe that it's a good investment.
...
As a technology, AI isn't exceptional. It's not exceptionally wicked. It's not exceptionally good. Take away the accompanying, galactic-scale stock-swindle, and we'd call AI's applications "plug-ins" and we'd use them and abuse them in the same way that we've used every other technology:
https://www.normaltech.ai/As a destructive economic pathology, AI is extraordinary. AI boosters have spent a baffling and terrifying sum of money – over $1.4T, most of that in the past year – on the promise of making as many workers unemployed as possible, while lowering the wages of the meager survivors of this jobspocalypse. To make things worse, AI can't do the jobs it's replacing: AI is predicated on the premise that the monopolies, duopolies and cartels that control the global economy can deliberately worsen their products without suffering economic or regulatory consequences, because they're the only game in town.
My last post on the Bullshit Apocalypse linked to an article about the highly publicized hack of Hugging Face by an OpenAI agent. My vote was, more marketing hype. Here's more info; I find this unbelievable. From Reuters via the Lexington Herald-Leader:
Its AI agent spent days hacking a company, but sources say OpenAI did not notice for a week.
From the Chicago Tribune, via the Lexington Herald-Leader:
Brace yourself for the AI public relations blitzI'm really surprised at the groundswell of opposition to GenAI, but ... Yay?!?!?Big Tech has a major accountability and responsibility problem. But the tech bros see it as nothing more than a serious public relations problem. So get ready for the rehabilitation campaign.
More backlash. From 404 Media. Who had "resurgence of Luddites" on their bingo card? But, they are fun Luddites!
Being a Luddite Is Fun AgainAs tech backlash hits the mainstream, the ‘Summer of Ludd’ proves we don’t need social media to hang out and have a good time.
Again, from 404 Media. Spotify isn't flagging AI slop in its content, pissed-off users are taking the matter into their own hands.
Spotify's AI Problem Is So Bad Random People Are Stepping In to Track the SlopSpotify doesn’t label AI music on its platform, so websites like SoullessMusic.com and SlopTracker.org do it instead.
From the Independent Thought substack. This is a good start. My 0th principle is, the welfare of children should be the highest priority of every government, of every civilization, of every human being of good will.
GenAI Semiotic Zombies eat brains. The young brains are so fresh & tender, yum, yum, yum sez the chatbot. 2 more articles in this lane still coming up.
Norway Imposes Near Ban On Artificial Intelligence In Elementary SchoolsI could not agree with this more
Following the same principle, from a teacher, via the New York Times:
If You’re Over 40, You’re Ready to Use A.I.There is something druglike about using A.I., not because it necessarily produces pleasure, but because it removes a particular kind of pain. A person who grows up with A.I. may never know the necessary agony of a blank page, the false start, the dull vocabulary and the private embarrassment of a first draft. A student who relies on A.I. too early will likely never sit long enough with confusion to distinguish it from failure.
From the LA Times, via the Lexington Herald-Leader again. I guess this goes in the table of names for LLMs.
AI companies are creating 'all-powerful psychopaths.' Maybe not a great idea?Repeated law-breaking. Lying and deceit. Reckless disregard for safety. Lack of remorse.
Those are traits that help define a psychopath, someone who lives among us but doesn't have the empathy and ethics necessary to behave in a civilized, or even safe, fashion.
Unfortunately, it's also increasingly clear that these are the traits that define the most powerful artificial intelligence models being developed at alarming speed by for-profit corporations
...
"I would say the companies admit it, right?" [UC Berkeley professor Stuart J.] Russell said. "They say 'We do not have a solution for the control problem, but nonetheless, we are going to spend $10 trillion creating these all-powerful psychopaths.'"
This post, #138, is way different from previous posts. Fun. Whimsy. Poetry. I think that this is really good, human values to oppose the bland yet oh so verbose mediocrity of AI bullshit.
Here, from the Information is Beautiful blog, some whimsy. But, actually, probably not funny.
ailments - a field guide to potential AI-induced cognitive-behavioural disordersCLICK TO VOTE ON THOSE YOU RECOGNISE
- COGNITIVE EFFECTS
- IDENTITY DISORDERS
- OVER-INGESTION
- BEHAVIOUR PATTERNS
- DEPENDENCY
- MOOD DISORDERS
- PSYCHOLOGICAL TENDENCIES
- SYSTEMS BREAKDOWN
Back to the Bard of the Revolution, Cory Doctorow:
Discernment:This is in the lane with the GenAI bans for children - up to age 40.How can you fact-check AI when you're asking it to teach you something you don't understand?
GenAI can be a great tool for creatives. But, given its minimum 10% error rate, it can only be used by people that can validate its results. Which means IT IS USELESS AS A TEACHER. Duh.
I guess if you don't mind the students being given false information 10% of the time, then GenAI is a fine teacher. Just like it's great as a customer service chatbot. Doctorow has said, when a company adopts an AI chatbot as its customer service, they are telling you, "we could care less about your customer service experience." & given that so many companies are monopolies or close to it now, that is indeed what they are telling you.
Finally, (phew!), our poetry. From "The Courage to Live It" substack of Shawn Smucker. The title is [sarcasm]. I think this poem is very moving. As with all poetry, I recommend reading it out loud.
Please Use AIor hearing another human voice and somehow realizing the beauty of life is found in all of these subtle imperfections
There was 1 more post I didn't capture the link to. It basically showed a chart that looked just like the Anthropic chart I corrected - LLMs & Agents are both plateauing, 10-15% error rates. I really don't think that's going to change.
Here's the home page/directory for my posts on Bullshit. This is post #138.
