Pinned Tweet
It seems vitally important to some to know whether a text was composed by a human or an AI.
I don't understand *why* this is so important (surely the text should stand on its own merits, or lack of them?) but I concede that such people exist. Okay, fine. There are people who claim that they can hear a difference if they replace the plastic knobs on their audio equipment with wooden ones. In itself, this isn't going to harm anything (other than their bank accounts) and provides some harmless amusement at their gullibility for the rest of us.
The AI v. human people are constantly searching for a "tell" β some simple-minded segmentation algorithm by which AI-written and human-written texts can be separated.
(OMG! I used an em dash! Some are going to stop reading here. π)
Most of these "tells" are are just plain dumb.
Let's say someone observes that architects use the word "fascia" far more often than non-architects (I haven't tested this, but I feel pretty confident that it's true).
Can you then conclude that any text containing the word "fascia" was written by an architect, or even was *likely* written by an architect? No, you cannot. Non-architects use the word all the time. Maybe the person was talking about the home repair he did last weekend. Maybe it's an internal memo at Lowes, talking about which items should be listed in this this week's flyer. Maybe it's someone with such terrible keyboard skills that his spell checker has confused "fascist" with "fascia", and he's blindly accepted the suggested change. There's absolutely no way to tell based on the presence of the word per se.
Now, if you had a large sample of texts written by the person over a long period of time (the time period is also important β our homeowner might write a lot of "fascia" texts while the repair is in progress, then never use the word again), which *as a whole* contained a disproportionate number of uses of "fascia", you could start to feel more confident that he is an architect (though maybe not even then... he might be a building contractor β or a surgeon).
From a single example? Sheer nonsense.
(Oops! I did it again! Twice! Definitely a bot!)
Why is this important? If it were just a matter of people making fools of themselves (like the wooden knob folks), it wouldn't be important at all. However, people are being given bad grades, perhaps even losing their jobs, based on one of these alleged "tells". That is completely unacceptable.
Even if such a "tell" did exist (which it doesn't), the likelihood that it would continue to work a year from now (or even six months from now) is very small. I follow this field fairly closely, yet I'm constantly surprised at how far each new generation has come.
Rather than wasting time looking for a "tell", teachers should concentrate on adapting themselves and their teaching methods to a world where AIs exist. They should learn about AIs themselves, then teach their students how to use them in a responsible manner. They're not going away.
One interesting idea I've seen: the students are assigned a topic, then given (say) a week to learn as much about the topic as they can. Every source is fair game. Books, Wikipedia, AIs, YouTube videos, you name it. Evaluating the credibility of the source is part of the task. Then they're put in a proctored exam room and asked to produce an outline of the topic. The students should be given different topics, of course. You don't want one student (or a small group) generating an outline for the others to cram.
There are many other things that could be done. Use your imagination! If you don't have one of those, perhaps you could ask an AI for help. π
pulpgrinder retweeted
1. LLMs are just stochastic parrots.
2. Fine, articulate parrots. Still just predicting the next word.
3. They canβt even multiply.
4. Fine, they can multiply. Try olympiad math.
5. Fine, they can compete. Try winning gold.
6. Fine, gold. Someone already knew the answers.
7. Fine, an open problem. Try a Millennium Prize Problem.
8. Fine, but it wasnβt Riemann.
9. Fine, Riemann. but did it understand the proof?
10β¦
pulpgrinder retweeted
1. LLMs can't even do multiplications.
2. Ok they can do multiplications but they can't do math competitions.
3. Ok they can do math competitions but they can't get gold at them.
4. Ok they can get gold but it's not like they can solve known unsolved problems.
5. Ok they can solve some known unsolved problems but not millennium prize problems.
6. Ok they can solve a millennium prize problem but it's not the Riemann Hypothesis.
7. Ok they can solve the Riemann Hypothesis but...
pulpgrinder retweeted
My grandfather has died of cancer.
On the morning the Riemann Hypothesis has been proven.
If you have advocated for a βpauseβ or βpacingβ or have a personal βp(doom)β you are my eternal enemy, and if given the opportunity, I will destroy you.
pulpgrinder retweeted
Everyone can spot AI writing now.
It's any sentence they disagree with
that has punctuation.
pulpgrinder retweeted
Researchers have used reinforcement learning to teach a 5-tonne John Deere skid-steer loader to perform complex acrobatic manoeuvres autonomously. Trained in simulation and transferred to a real machine, the system can balance on two wheels, navigate rough terrain, spin, and control its bucket before landing safely.
The research demonstrates the potential of AI-driven automation for heavy construction and industrial machinery.
Video Credit: @leggedrobotics / @ETH
Paper: wevlv.co/4kltYys
#engineering #technology
Letβs take this gentlemanβs argument at face value that anyone can throw money at rockets and get something like Starship, Falcon or Dragon.
Perhaps he can explain to everyone how SpaceX is able to catch giant starships from orbit with chopsticks and deliver 85% of all Earth payload to orbit with Falcon, while Blue Origin, which started before SpaceX and was funded by the immensely wealthy Jeff Bezos, is not yet able to launch reliably?
Also, why is it that Boeing and Lockheed received several times what SpaceX did from the government for development of their rockets and spacecraft, but the Boeing capsule is still not operational and ULA delivers less than 5% of SpaceX payload to orbit?
I am genuinely curious about these questions.
pulpgrinder retweeted
AI art isn't the absence of a human. It's a human with a new instrument.
Someone still chooses the light, the pose, the silence. Someone still decides when it's finished.
That someone is why I love it. β¨
pulpgrinder retweeted
If I was a billionaire I'd use my money to solve world hunger. π‘
Sent from a device purchased with money that could have been used to feed an Ethiopian family for a year, but wasn't because my hypothetical generosity is purely performative.
I have to give credit to PRC for a master play.
A guy as smart as Jensen Huang is gonna be instantly suspicious if a 22-year-old supermodel appears out of nowhere and starts throwing herself at him.... he might not think "spy", but "gold digger" will come to mind immediately.
The cute and smart 42-year-old CEO of a company that does a lot of business with his company? That's way more plausible.
71
pulpgrinder retweeted
βAI is for lazy people.β
Yes. Exactly.
Lazy people invented almost everything worth having.
Someone was too lazy to carry things, so we got the wheel. Too lazy to wash clothes by hand, so we got the washing machine. Too lazy to do math on paper, so we got the calculator. Too lazy to walk, so we got the car.
Every great invention in history started with someone saying βI really donβt want to do this anymore.β
Hard work was never the goal.
It was the price we paid because we didnβt have anything better.
Now we do.
Call me lazy all you want.
Iβm in very good company.
pulpgrinder retweeted
Maybe they could play competitive GTA for money. I'm totally serious. We already have professional gaming leagues.
We could build a machine that would bowl a strike essentially every time, even with older technology, but no one does that. That's because bowling is a sport. The point is not to bowl strikes per se, but to compete against another human being.
Sports of any kind are pretty much immune to AI.
There are other fields for which this is true, specifically anything for which "hand made" is a selling point. Tailoring. Cooking. Even (despite the fears) art. AI can generate images of pretty much anything you can imagine, but some of us have better imaginations than others. The stuff from people with better imaginations will still be valuable, even if the final image is generated using AI.
I also feel certain that there will be new jobs that we can't even comprehend, any more than a medieval serf could comprehend jobs like "software engineer" or "social media influencer".
We're already starting to see pretentious software projects describing themselves as "AI free". Almost invariably these are fringe projects with one committer and three users (compare and contrast the pro-AI positions of people like @esrtweet and Linus Torvalds).
I'm calling this "organic software", by analogy with "organic" vegetables: 5x the price and full of bugs.
48
Ask your favorite AI to make a picture of you using the information it has, and put you in the historical era it thinks appropriate.
π€ Made with AI
2
69
pulpgrinder retweeted
BREAKING: Linear algebra is now a national security threat. π
Matrix multiplication? ILLEGAL.
Gradient descent? Straight to jail.
Backpropagation? 25 to life.
And donβt even THINK about multiplying two tensors without a federal permit.
Apparently the next great technological revolution must be stopped before someone accidentally invents a better algorithm.
Somewhere in China, an engineer just read this and laughed his fucking ass off. π





















