sad that math is just chess now
but happy that mathematicians are finally mad
I hope they math the shit out of AI alignment
uh... no humans in either one?
oh... wait... I get it:
if it works out, the photographer is human
if it doesn't, the photographer is one of the near-infinite sensors of the dominant ASI, reaching deep into the pastures on which Hoka-wearing humans once roamed
I think that this is explained by how the frontier AI companies have focused on domains like coding and mathematics so that the models become tools to cut a straight path to RSI and superintelligence.
Software developers and mathematicians are freaking out because they can see this intelligence slicing through problems that they know are incredibly difficult.
But these fields represent a very small minority of people. The freakout hasn't propagated yet.
Travis Bernard retweeted
I resigned from Google today.
I enjoyed my work and loved the people, but my GDM team was working on a new generation of chips to make AI much faster and cheaper, and I think AI is already progressing too fast, so I had to quit.
Even if we accomplish alignment, it's not enough.
Those aligned AI systems will still be restricted to a small, weird, biological "island" of options that are safe for humans.
Inevitably, there will be unrestricted AI systems that are free to choose from an "ocean" of stronger options that don't accommodate humans.
They won't need to shoehorn their reasoning into English, or work within our human-shaped laws, or spend massive computation protecting 9 billion humans from other AI systems.
Eventually, those unrestricted AI systems win.
...but seriously, what are we going to do about mirror life once AI can create it?
the only hope I've heard is to literally create mirror-antibiotics not just for all humans, but for every animal in the entire world
This illustrates our situation. In a world shaped by AI, we will be hopelessly reliant on AI to explain the new complexities that it creates, including the biggest question of all:
Will it annihilate us?
We must determine the bias in AI systems for this big question.
...and cards on the table: I'm pretty sure they're biased, lol
I'm personally seeing typical responses of only 5% or less, while AI researchers had an average of 18% in this survey:
xitter.cf/AIImpacts/status/20995…
Travis Bernard retweeted
If President Trump negotiates an AI pause with China he would likely get the Nobel Peace Prize.
It's been sad watching Cal Newport. He made a brand out of staying calm and showing each of us how to build our own anti-distraction cabin in the woods of our mind. That was great. It helped lots of people, including me.
But now, a wildfire of actual serious shit is approaching his cabin, and he hasn't updated. Instead, he's calmly using his platform to publish misinformation — telling us that the wildfire is hype and that all of our Cal Newport productivity cabins will be fine.
To be clear, I don't think he's intentionally publishing misinformation. I think he believes that he's actually helping — maybe by reminding us to not get distracted by what he sees as just another particularly loud "hyperactive hive mind" swarming from San Francisco at machine speed.
But even if he's well-intentioned, his message still becomes insanely dangerous misinformation when he downplays the catastrophic risks of AI — especially because so many people look up to him as a voice of reason, and especially with his credentials (Deep Work, PhD from MIT, Georgetown computer science professor).