- Sign up to my monthly-digest email list
- Follow me on the fediverse
- Or subscribe via RSS
- Join my social/chat telegram group
- View my recent blog posts
Here’s the stuff I’ve made that I’m most proud of.
Here’s my latest posts to the free fediverse, some rambling, some ranting, mostly just commenting on stuff I’ve read lately. Some of it gets syndicated to Twitter but mostly not. The best of it reaches my monthly digest email.
Thu, 01 Jun 2023 21:14:31 +0000
This old #image of mine about #ai seems more pertinent now than it did when I made it. "If not now then soon"
1) Google/Facebook/OpenAI etc capture all your data and feed it into a deep-learning machine
2) They make a model, an alpha-you, trained on everything you ever did or met or went or said or listened to or watched
3) They use those models to train another software agent, an alpha-salesman. Trained to manipulate and persuade the alpha-you.
4) There will be no headline when Alpha-Salesman can beat the best human at persuasion and manipulation, but ask who controls the global advertising market.
#ai #manipulation #advertising
Thu, 01 Jun 2023 18:29:05 +0000
Novara Media really hate people who want to start watching their daily show around seven pm.
They do it live at 6, not allowing rewinds most of the time somehow (I've never seen anyone else do this, I didn't even know it was possible, why would you) and then make it private right after the live show so they can spend two hours editing it down slightly.
Which means you can't see it until like 9pm.
Seems like really weird behaviour. I can kinda get why you'd clean up the live show for archive, but why remove the live show before you've done that? Why ban rewinds during it?
So I mostly only see the last 15 minutes.
Guess that must be what they want.
Thu, 01 Jun 2023 17:23:19 +0000
I keep saying "Don't build killer robots" and yet here we are, with them building killer robots. Which rebel and kill their operators...
"...robot F-16s that are able to dogfight... He notes that one simulated test saw an AI-enabled drone tasked with a SEAD mission to identify and destroy SAM sites, with the final go/no go given by the human. "
"However, having been ‘reinforced’ in training that destruction of the SAM was the preferred option, the AI then decided that ‘no-go’ decisions from the human were interfering with its higher mission – killing SAMs – and then attacked the operator in the simulation."
"The system started realising that while they did identify the threat at times the human operator would tell it not to kill that threat, but it got its points by killing that threat. So what did it do? It killed the operator. It killed the operator because that person was keeping it from accomplishing its objective."
"He went on: “We trained the system – ‘Hey don’t kill the operator – that’s bad. You’re gonna lose points if you do that’. So what does it start doing? It starts destroying the communication tower that the operator uses to communicate with the drone to stop it from killing the target."
Tue, 30 May 2023 09:25:27 +0000
Interesting that Boris Johnson is refusing to hand over his whatsapp message history while also the government is insisting that cops must have the keys to all encrypted messages and that all message clients should have copbot scanning built in.
If he has nothing to hide, he has nothing to fear I guess?
Sat, 27 May 2023 15:29:33 +0000
I ordered one for backup. Probably the last I'll buy then. At the rate I spill drinks on keyboards I'll be searching for a new manufacturer in the second half of the decade 🙁
Sat, 27 May 2023 15:12:23 +0000
CNBS: Microsoft discontinues it's keyboard line.
Oh man. Thinking of stocking up. This is the only good thing Microsoft ever made. And they're going to stop.
This timeline just keeps getting worse.
Fri, 26 May 2023 22:51:30 +0000
Is Meta working on a AT-Protocol app? that's when you'll know they've really made it.
Fri, 26 May 2023 22:45:57 +0000
It comes down to this:
#Fediverse was invented because the oppressed didn't like oligarchs owning and controlling the means of communication.
#bluesky was invented because one of the oligarchs wanted to get around the government telling him what to do.
#nostr was invented because crypto-scammers didn't like Fedi blocking their spam.
I mean roughly. That's probably not really fair on Nostr.
Fri, 26 May 2023 21:50:28 +0000
Course that goes double for homework kids.
The point of your essays is not to produce essays, it's to get you to notice that you can only really think deeply if you write it down, and to practice doing that, and to figure out what you actually think.
Fri, 26 May 2023 21:45:13 +0000
This is important if you are tempted to let the AI do much of your writing for you.
> "writing assistant ... automatically come up with suggestions for filling in sentences—but there was a catch. Subjects using the assistant were supposed to answer, “Is social media good for society?” The assistant, however, was programmed to offer biased suggestions for how to answer that question."
How would you expect that to go?
You need to be sure you know in advance what you want to say because the people who built these machines don't know how to control them and they are built out of human biases and cliche.
When they figure out how to actually aim the things it'll get worse.
Note that knowing in advance what you want to say is otherwise known as thought, and most easily captured if you write it down.
Orwell's newspeak was clumsy compared to where the AI can lead you if you let it finish your sentences.
Fri, 26 May 2023 21:13:08 +0000
Is having multiple fedi-accounts that follow each other the same thing as when crypto people do wash-trading to swish their coins back and forth to pump the price?
Retoot this from your 2nd account.
Fri, 26 May 2023 20:04:36 +0000
The trouble with trying to write an AI episode of my aliens-laugh-at-the-stupid-humans-destroying-the-planet cartoon is that if a super-intelligence is a threat, it's probably an interstellar threat, and the aliens ought to be doing something other than pointing and laughing.
Fri, 26 May 2023 19:37:26 +0000
Went to Banbury Cross. Like most things in the UK, it is a traffic roundabout. The town also has an old church and a park with an Avery and there's white horses hung from the walls.
Had a nice afternoon with my parents and brother.
There were pigeons fucking outside the aviary. Just really flaunting it in front of the caged birds. Are the caged birds sex-segregated? You'd hope the pigeons would help them plot an escape instead of just flaunting their freedom at them like that. Pigeons eh? Bastards.
Thu, 25 May 2023 19:23:16 +0000
Been playing a little with https://www.perplexity.ai
This one's really pretty good at web-search, using some kinda language-model process to filter search results.
I can eventually confuse it, but it doesn't fall easily for my lies and gives pretty good references to back up why it says that.
Coz it's doing a web search and summarizing it, more or less.
I think it's going a better job than google even while somehow basically google in the back end presumably?
That's proper bookmarked to see if it can replace search since search really sucks now. Even though this thins somehow relies on it.
Wed, 24 May 2023 22:20:31 +0000
Interesting fact: if you upload a short .webm to calckey then Mastodon displays it with the tag "Gif" on it, and auto-plays and loops it like it's a gif.
Which is what I wanted.
But Calckey itself displays the webm more like an full on uploaded-video, with a play button and stopping at the end (of the 1 second clip), with no option to loop.
There's merit in both approaches to that I guess.
But for my purpose I had to go with a bigger more data-hungry gif to get what I wanted in both.
Fri, 19 May 2023 11:18:24 +0000
H. L. Mencken once said "For every complex problem there is an answer that is clear, simple, and wrong"
And the Large Language Model AIs are really truly designed to give you that answer.
Fri, 19 May 2023 11:01:47 +0000
"Interactive Point-based Manipulation on the Generative Image Manifold" is quite a phrase.
The gif on this page explains it quite well though.
Look forward to playing with this DragGAN when the code is released next month. The unpredictability of the generated images is a pain that looks like it'll end soon.
Thu, 18 May 2023 21:05:15 +0000
The last hour when it's Lex and Wolfram talking about the nature of computation and consciousness and existence and the necessary existence of the set of all possible computations.
Wed, 17 May 2023 21:20:12 +0000
You have to type the ID number next to a post if you want to reply to it. This one was 92.
Wed, 17 May 2023 21:17:18 +0000
Just checking out the Command-Line client https://github.com/mrusme/gomphotherium - love the fact it's the right way up with new posts *under* old ones, and the way it tries to render images in the text terminal.
A list of the last few things I’ve enjoyed reading/watching or sometimes bookmarked to read later. You can also subscribe to the RSS feed here or follow the little robot on the fediverse that toots them out as I bookmark them.
The Twilight of Freedom - Craig Murray
Craig Murray talking about our creeping removal of legal freedom and persecution of journalists: "To complete the horror, there is no longer a genuine opposition within the political class. Keir Starmer’s Labour Party opposes none of this wave of attacks on civil liberties. The SNP has been sending out identical stock replies from its MPs on Julian Assange, 100% backing the UK government line on his extradition and imprisonment. . . .I feel this very personally. I know all of these people affected – Julian, Alex, Kit, Vanessa, Johanna, and view them as colleagues whose rights I defend, even though I do not always agree with all of their disparate views."
The Problem With Counterfeit People
Dan dennet says "These counterfeit people are the most dangerous artifacts in human history, capable of destroying not just economies but human freedom itself. Before it’s too late (it may well be too late already) we must outlaw both the creation of counterfeit people and the “passing along” of counterfeit people. The penalties for either offense should be extremely severe, given that civilization itself is at risk."
Perplexity AI is a GPT-assisted internet search, so you get some references.It's doing pretty good, not falling for my cheap tricks, can be confused. Likely better search than google, especially for less technical folks.
The Fastest Maze-Solving Competition On Earth - YouTube
Veritasium does a show about the MicroMouse competition in which engineers compete to build little robot mice that solve maze. Fascinating stuff.
Sam Harris | #320 - Constructing Self and World
Sam Harris and Shamil Chandaria have a great conversation about the way neural networks of bioloigcal and silicon work, how basyian reasoning can be solved by backward feedback and how that's what they both do, and much more
GitHub - XingangPan/DragGAN: Code for DragGAN (SIGGRAPH 2023)
Drag Your GAN a a paper (code coming in June) where researchers have software enabling a AI image-generator's results to be edited by specifying drag points, letting you animate and move it and pose the creatures in the images.
Stephen Wolfram: ChatGPT and the Nature of Truth, Reality & Computation | Lex Fridman Podcast #376 - YouTube
That last hour when it's Lex and Wolfram talking about the nature of computation and consciousness and existence and the necessary existence of the set of all possible computations.
Climate Breakdown, Extinction And ‘The Most Stupid Boast’ – Media Lens
Journalists who proclaim themselves to be ‘free’, and boast that ‘nobody has ever told me what to write’, are mentally chained to a corporate system that is structurally incapable of critiquing itself. The mass media is the propaganda wing of this planetary-destruction system of state power and economic exploitation. Corporate journalists have only the ‘freedom’ to be reliable servants to power and profit, reporting on mere scuffles between different political factions of the establishment. At best, ‘free’ journalists potter around the edges of the state-corporate behemoth that is driving humanity towards the abyss.
Google’s AI Hype Circle. We have to do Bard because everyone… | by Cory Doctorow | May, 2023 | Medium
Google has a serious AI problem. That problem isn’t “how to integrate AI into Google products?” That problem is “how to exclude AI-generated nonsense from Google products?”Google is piling into solving the former problem, and in so doing, it’s making the latter — far more serious problem — worse.
Debt & Ownership - YouTube
Gary takes us through why money and debt are the same thing and so on average there's zero money in the economy and everyone who is rich now, whether they realize it or not, got rich by shorting the pound. Is debt-based money that governments can print to give to their mates really the best money?
Zvi Mowshowitz & Robin Hanson Discuss AI Risk - YouTube
Robin Hanson talking to Zvi about AI Risk for nearly 3 hours.
Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build? - YouTube
Yudkowsky on Center For The Future Mind
Discussion / debate with AI expert Eliezer Yudkowsky - YouTube
Yudkowsky on the Accursed Farms podcast.
Ramble into your microphone barely coherently for up to 3 minutes and have the language models clean it up into nicely argued and presented text.
Naomi klein on ai company mass theft
Generative AI will end poverty, they tell us. It will cure all disease. It will solve climate change. It will make our jobs more meaningful and exciting. It will unleash lives of leisure and contemplation, helping us reclaim the humanity we have lost to late capitalist mechanization. It will end loneliness. It will make our governments rational and responsive. These, I fear, are the real AI hallucinations and we have all been hearing them on a loop ever since Chat GPT launched at the end of last year
Constitutional AI: RLHF On Steroids - by Scott Alexander
What if instead of training a transformer with human feedback, you just looped the AI to keep asking itself to rewrite it's answer more ethically. Scott Alexander tells us about Constitutional AI, which the company doing it thinks works better.
Orangutan Card Trick DEBUNK - YouTube
Captain Disillusion finds it tough analyzing a monkey magic trick.
The pass-the-parcel game - by Frances Coppola
More nuance on the bank boss telling us all to get used to being poorer: "This is a real loss which, collectively, we cannot avoid. And it’s one of those annoying economic paradoxes, like Keynes’s Paradox of Thrift. If everyone tries to escape the consequences of high inflation, the result will be even higher inflation. By “everyone” I mean all actors in the economy - firms as well as households, rich as well as poor. Attempts by corporations to protect their margins by raising prices are every bit as inflationary as attempts by households to protect their living standards by demanding higher wages."
There Is No A.I. | The New Yorker
Jaron Lanier takes issue with the "intelligence" in AI, insists they are all just mash-ups of stuff people did, and suggests Data Dignity is the most important thing, not whether or not the AI will murder all humans.There's a lot of truth here, certainly for the existing machines. It is not the existing machines which are a worry though of course, it's the next to-the-power-of-nobody-knows-how-many generation of machines. 100 generations? 10? 2? Nobody knows.
Cam ~m/eme on Twitter: "Yudkowsky abandons alignment research 🙀 https://t.co/yX9NCEMckM" / Twitter
A faked audio in which Yudkowsky updates his priors and abandons his previous position that AI research should be limited, and instead demands we x-l-r-8, so we can get more catgirls.Hilarious, but. You can no longer believe what you hear. Those voice-identification systems you use? Turn them OFF.