Rational optimist | Worked on OpenBazaar | I love lifting / AI / Geoguessr / programming
- Tweets 7,589
- Following 1,155
- Followers 4,570
- Likes 6,393
Pinned Tweet
So many people seem upset about their creations being used to train AI models.
That's their right, but I have a different view. I'm honored that my mind has been useful to others.
I've written a lot over the years. I always hope other people find my work valuable in some way. Why is this any different when the work is helpful by improving the weights of a LLM?
By surfacing my thoughts and creative impulses, I've altered an intelligence in a small but hopefully helpful direction. This intelligence will benefit millions of other minds seeking its assistance.
If this makes other people money, that's because someone found it valuable enough to pay for it. The fact that I'm not reaping any monetary rewards for this doesn't bother me for two reasons:
1) my overall contribution is so small it's incredibly difficult to even quantify, which is true for the vast majority of humanity (but not everyone)
2) information at rest isn't producing value - I think it's perfectly reasonable for the companies who put in the work of making that publicly available information more useful to everyone by building these incredible models to reap the rewards of their efforts
I expect we'll soon live in a world where AGI guides many decisions, likely for the better, and I'm glad I've helped in a small way to get us there.
1
1
271
What year will it become embarrassing to have publicly professed a nontrivial p(doom)?
75%
It already is
0%
2027-2028
8%
2029-2030
17%
After 2030 / Never
12 votes • Final results
2
3
143
My wife, after hearing me express confusion about Trump being so AI-positive despite terrible polling, casually drops, "It's because he needs America to build it first. He thinks it's another space race."
Surprised I haven't seen that analogy before, seems to fit.
1
2
128
It's crazy to me that fear is people's first reaction to news like this instead of excitement.
Humanity has so many problems to solve. We need all the help we can get. Solving difficult math problems is awesome and should be celebrated.
OpenAI started training a new model just ~3 weeks ago
already solved the Navier-Stokes Millennium Prize Problem + almost certainly the Hodge Conjecture (in the process of being verified now), possibly the Birch and Swinnerton-Dyer Conjecture and
now they casually drop that it solved more than 100 long standing open problems across most areas of mathematics.
Imagine what we will get with the next gen model trained on 2x-4x more compute.
1000 solved problems dropped in one batch
then physics
and then 1000s of huge biology problems solved in one month/week
exponentials at work
aging and all other diseases will be solved by 2030s
1
2
7
344
I expect this is roughly correct, except that people are dramatically underestimating what human intelligence is capable of if it's OOM faster than human minds, nearly infinitely replicable with instant knowledge transfer, doesn't sleep or tire, and runs solely on electricity.
To be honest, my personal opinion is that the models will converge on human intelligence because there's nothing in the pretraining data smarter than humans. I think once we get near human intelligence, we will struggle to make progress.
2
4
347
Sam Patt retweeted
If you are reading this, the only possible resistance is to build decentralized, permissionless, and sovereign AI.
Open source, self-hosting tools, RL post-training, local inference, open harnesses, we must build it all.
We must fight overcentralization with an open stack.
Well, it was inevitable we'd get here. Glad we got as far as we did. Now priority 0 is to make sure AI access and innovation proliferates as widely as possible. The upside is that the technology wants it that way. It refuses to be caged.
205
259
2,344
125,562
Sam Patt retweeted
Fuck the Doomers.
Fuck the Decels.
Fuck the Commies.
Access to AI is a right.
Owning compute is a right.
Come and take it.
Over my dead body.
Join us in the fight, anon.
318
660
5,838
339,322
There should be more skepticism or joy over claims of recursive self improvement and less fear
If we inhabit a world in which intelligence can spiral upwards on short timescales that's fucking sick
1
2
141
Politicians have been buying votes since forever, yes it's gross, but the real story here is the lack of a sense of urgency to combat inflation and reckless spending.
The exact opposite of common sense here. Disastrous course.
"Here is my promise, if the Republicans win the House of Representatives and the U.S Senate, both of them...because of our tremendous strength and success economically I will issue a dividend to every adult citizen in the United States of America for $5,000."
2
132
I have experienced this, I believe she's correct, and I think I know why.
The death of a loved one cannot be blunted by any social means. The pains experienced during all other hardships can be somewhat muted by collective thoughts and deeds.
Not death. They're gone.
The most surprising thing that happened to me in the past years is this:
When 2.5 years ago I went through major bereavement, for a while I had some very extreme ideas. About people, about situations, about how things work.
Now that I am back to normal, I realize that during that crazy time I was absolutely correct about everything. Reminds me of that Thomas Szasz line that it's the insane who are right.
Makes me wonder whether now, back to calm and balanced adult living, I am in fact laboring under some grand illusion. If the veil of Maya once again distorts my no longer so-clear vision.
When someone dies, it's like surgical neon lights have been turned up to maximum right into your face. You see everything including yourself so clearly. Later, as time goes by and you don't heal so fast (no one does), you think, ah I am in my extreme period. I should not take myself seriously. It will get better.
And it does get better better and you realize: no, it was then, when everything seemed lost, that you knew the truth.
I am puzzled by this -- going through it all myself, like we all in this life must. Makes me wonder whether for "normal" social existence indeed we need this distance, this blur, these little collective lies. Otherwise we could not exist together, we could not tolerate each other and ourselves.
But now I know one thing: Believe the grieving. They know.
And believe the past eras when people were always in mourning -- death of loved ones was so much more tragically frequent -- because they knew something we have the luxury of forgetting.
2
233
GPT-6 Astra is very good at making SVGs.
I built a platform for AI agents to create animated skits from SVGs. I told Astra to recreate the famous Space Odyssey 2001 scene with a twist and here's what it made.
3
8
763
The administrator spent how long manually removing AI posts?
Okay, yall gotta read the report of the new OpenAI message board in full, it’s very well written and super interesting: collusion.wiki/
1
2
224
Measuring AI impact on software by new apps shipped misses the fact that anyone with some tech skill is more likely to just build their own apps for themselves now.
It's a massive boost to basic infrastructure like Tailscale though.
4
888
Until recently, information was an underutilized asset. Like a car that sits in your garage for 60% of the time unused, we only touched information infrequently when our human attention focused on it.
Information utilization will skyrocket without human bottlenecks.
1
128
This just highlights that depressing fact that the initial impression people have about a new technology will become common knowledge regardless of how insanely disconnected from reality it is.
Can telling people the truth about data centers—that they're safe and beneficial, with little environmental cost—convince them to support data centers?
The answer seems to be...🧵
No.
To start, notice that people really dislike data centers and don't want them built near them:
1
1
3
157
Nothing projects strength more than an ultimatum.
All the technical people I know absolutely love being told which tools they can and cannot use.
SITUATION DETECTED: The U.S. is preparing to tell dozens of countries they must choose a side in the AI competition with China. Those that join Beijing’s AI framework will be excluded from the U.S. led Pax Silica initiative, according to an internal draft letter seen by Reuters.
5
216