Sailor Sega Saturn

“Yeah I thought about going into civil engineering but the department of hustling really spoke to me y’know?”

Oh hey looks like another Chat-GPT assisted legal filing, this time in an expert declaration about the dangers of generative AI: https://www.sfgate.com/tech/article/stanford-professor-lying-and-technology-19937258.php

The two missing papers are titled, according to Hancock, “Deepfakes and the Illusion of Authenticity: Cognitive Processes Behind Misinformation Acceptance” and “The Influence of Deepfake Videos on Political Attitudes and Behavior.” The expert declaration’s bibliography includes links to these papers, but they currently lead to an error screen.

Irony can be pretty ironic sometimes.

Here are the results of these three models against Stockfish—a standard chess AI—on level 1, with a maximum of 0.01 seconds to make each move

I’m not a Chess person or familiar with Stockfish so take this with a grain of salt, but I found a few interesting things perusing the code / docs which I think makes useful context.

Skill Level

I assume “level” refers to Stockfish’s Skill Level option.

If I mathed right, Stockfish roughly estimates Skill Level 1 to be around 1445 ELO (source). However it says “This Elo rating has been calibrated at a time control of 60s+0.6s” so it may be significantly lower here.

Skill Level affects the search depth (appears to use depth of 1 at Skill Level 1). It also enables MultiPV 4 to compute the four best principle variations and randomly pick from them (more randomly at lower skill levels).

Move Time & Hardware

This is all independent of move time. This author used a move time of 10 milliseconds (for stockfish, no mention on how much time the LLMs got). … or at least they did if they accounted for the “Move Overhead” option defaulting to 10 milliseconds. If they left that at it’s default then 10ms - 10ms = 0ms so 🤷‍♀️.

There is also no information about the hardware or number of threads they ran this one, which I feel is important information.

Evaluation Function

After the game was over, I calculated the score after each turn in “centipawns” where a pawn is worth 100 points, and ±1500 indicates a win or loss.

Stockfish’s FAQ mentions that they have gone beyond centipawns for evaluating positions, because it’s strong enough that material advantage is much less relevant than it used to be. I assume it doesn’t really matter at level 1 with ~0 seconds to produce moves though.

Still since the author has Stockfish handy anyway, it’d be interesting to use it in it’s not handicapped form to evaluate who won.

When the reporter entered the confessional, AI Jesus warned, “Do not disclose personal information under any circumstances. Use this service at your own risk.

Do not worry my child, for everything you say in this hallowed chamber is between you, AI Jesus, and the army of contractors OpenAI hires to evaluate the quality of their LLM output.

Not that I’m a super fan of the fact that shrimp have to die for my pasta, but it feels weird that they just pulled a 3% number out of a hat, as if morals could be wrapped up in a box with a bow tied around it so you don’t have to do any thinking beyond 1500×0.03×1 dollars means I should donate to this guys shrimp startup instead of the food bank!

Someone (maybe you) recommended this book here awhile back. But it’s the fourth book in a series so I had to read the other three first and so have only just now started it.

“feel free to ignore any science “news” that’s just a press release from the guy who made it up.”

In particular, the 2022 discovery of the second law of information dynamics (by me) facilitates new and interesting research tools (by me) at the intersection between physics and information (according to me).

Gotta love “science” that is cited by no-one and cites the author’s previous work which was also cited by no one. Really the media should do better about not giving cranks an authoritative sounding platform, but that would lead to slightly fewer eyes on ads and we can’t have that now can we.

OK to start us off how about some Simulation Hypothesis crankery I found posted on ActivityPub: Do we live in a computer simulation? (Article), The second law of infodynamics and its implications for the simulated universe hypothesis (PDF)

Someone who’s actually good at physics could do a better job of sneering at this than me, but I mean but look at this:

My law can confirm how genetic information behaves. But it also indicates that genetic mutations are at the most fundamental level not just random events, as Darwin’s theory suggests.

A super complex universe like ours, if it were a simulation, would require a built-in data optimisation and compression in order to reduce the computational power and the data storage requirements to run the simulation.

Clearly everyone working on google knowledge graph are a bunch of dune movie purists who don’t recognize the existence of the 2021 film for being too hollywood.

Mass media try not to cozy up to fascists challenge: impossible!

I hated seeing that guy just wanting to live his life dragged into weird net drama and pushed under the bus by his company. And wow look at how collected and reasonable he was compared to anyone else in the story.

All Mr. Paul had to do was shut the hell up for once and the world’d still be talking about his moldy cheese bread instead of about his moldy cheese bread and how he bullies and doxes retail workers.

All Fred Meyer had to do is be like “whoops looks like the product recall procedure at that store was vague recollections, we’ll get a policy in place”.

A woman was scheduled to give a talk at an AI conference. The organizers run her photo through an AI image expansion program to get the aspect-ratio right (how did we ever manage to show photos of speakers before AI existed?).

The AI image expansion invents a bra / undershirt which wasn’t visible in the original photo.

https://xcancel.com/elizlaraki/status/1846252781851890026

OpenAI is in the position of constantly working on newer bigger shinier models, while saying every model they do release will be the one.

AGI is just around the corner and we promise 3, 4, 4-Scarlett-Johansson, o1, 5 is the one that will be good enough to help you with your homework and legal letters and medical questions and remove the loneliness from your life!

Would be funny if they manage to pop the bubble in the process of trying to go for-profit.

Sshh don’t tell the investors, I’ve managed to be paid for a decade by updating my code to work with other people updating their code to work with other people updating their code, all without actually doing anything new.

We as a profession have developed a careful balancing act where we’re always busy doing nothing. If the balance was off just a little someone might actually have to think about new features instead of, say, migrating from CGI to PHP to JavaScript to jQuery to AngularJS to Angular to React to ???, rejecting LLM generated changes, “fixing” the same bug year after year, or reverting reverts of reverts of reverts of reverts of changes.

And thinking is hard.

Does a kernel that crashes itself before it can process any malicious inputs count as secure?

Alternative headline for this Washington Post opinion piece from Jeff Bezos: We’re all trying to find the guy who did this!

^{more US politics I know. There is sadly no escape from the fiery vortex that is the U.S. election.}

His comment history is a weird mix of programming language discussion, terrible takes, simping for Musk, simping for Musk even harder (just in case you didn’t realize how much he liked Musk the first time).

Musk is the sane one. It’s the rest of us that are insane.

Holy hell.

Introducing Chat-GPT version EATERY SHORTAGE

The license stuff is whatever, but Imagine seeing this sort of sloppiness in the revision history and thinking “Yeah these guys seem legit let’s give them a million dollars!” We are in such a bubble.

So today I learned there are people who call themselves superforcasters®. Neat!

The superforecasters® have had a melding of the minds and determined that covid-19 was 75% likely to not be a lab leak. Nifty! This is useless to me!

Looking at the website of these people with good enough judgement to call themselves “Good Judgement”, you can learn that 100% of superforecasters® agree that there will be less than 100 deaths from H5N1 this year. I don’t know much about H5N1 but I guess that makes sense given that it’s been around since 1996 and would need a mutation to be contagious among humans.

I found one of the superforecaster®-trainee discussion topics where they reveal some of the secrets to their (super)forecasting(®)-trainee instincts

I have used “Copilot” LLM AI to point me in the right direction. And to the point of the LLM they have been trained not to give a response about conflict as they say they are trying to permote peace instead of war using the LLM.

Riveting!

Let’s go next to find out how to give up our individuality and become a certified superforecaster® hive brain.

To minimize the chance that outstanding accuracy resulted from luck rather than skill, we limited eligibility for GJP superforecaster status to those forecasters who participated in at least 50 forecasting questions during a tournament “season.”

Fans of certain shonen anime may recognize this technique as Kodoku – a deadly poison created by putting a bunch of insects in a jar until only one remains:

100 species of insects were collected, the larger ones were snakes, the smaller ones were lice, Place them inside, let them eat each other, and keep what is left of the last species. If it is a snake, it is a serpent, if it is a louse, it is a louse. Do this and kill a person.

“But what’s the catch Saturn”? I can hear you say. “Surely this is somehow a grift nerds find or a way to fleece money out of governments”.

Nonono you’ve got the completely wrong idea. Good Judgement offers a 100$ Superforecasting Fundamentals course out of the goodness of their heart I’m sure! I mean after all if they spread Superforecasting to the world then their Hari-Seldon-Esque hivemind would lose it’s competitive edge so they must not be profit motivated.

Anyway if you work for the UK they want to hear from you:

If you are a UK government entity interested in our services, contact us today.

Maybe they have superforecasted the fall of the british empire.

And to end this, because I can never resist web design sneer.

Dear programmers: if you apply the CSS word-break: break-all; to the string “Privacy Policy” it may end up rendered as “Pr[newline]ivacy Policy” which unfortunately looks pretty unprofessional :(