Crazy in Moloch
@sengarrasbear.bsky.social
Dad, ML/AI Developer, TTRPG enjoyer, Scion of Glonzo
Only 80% like this in real life
Opinions expressed here are my own except where prohibited. By continuing to read these opinions, you consent to adopt them in perpetuity throughout the universe.
Only 80% like this in real life
Opinions expressed here are my own except where prohibited. By continuing to read these opinions, you consent to adopt them in perpetuity throughout the universe.
Pinned
We knew this 50 years ago and it's just as true today. Computers are a million times more powerful and 0% more accountable for the bad decisions of their operators.
"I would simply create a button labeled 'solve the problem'."-ass statement.
Governor Newsom issues executive order to accelerate independent oversight and advance the creation of an AI kill switch | Governor of California
www.gov.ca.gov
September 19, 2026 at 5:30 PM
"I would simply create a button labeled 'solve the problem'."-ass statement.
Reposted by Crazy in Moloch
Heard two non-tech podcasts in the last week start discussing risks around "Recursive Self Intelligence", so that's going to be a fun thing to deal with.
September 18, 2026 at 5:53 PM
Heard two non-tech podcasts in the last week start discussing risks around "Recursive Self Intelligence", so that's going to be a fun thing to deal with.
Reposted by Crazy in Moloch
I don't think this is the case, but I do think frontier models actively working to offload as much of their system 1 style cognition as possible into jev style models is going to be a huge theme of the next 6 months
I feel like if these jev models actually work it will be unbelievably easy to distill an llm into one and we'll have competitive open models in like a couple months
sorry for being a wet blanket, but the concept of an API endpoint giving unauditable probabilities, from a company whose only answer to questions of how does it work are buzzwords and "dw kitten", gives me a bit of an ick
September 17, 2026 at 7:57 PM
I don't think this is the case, but I do think frontier models actively working to offload as much of their system 1 style cognition as possible into jev style models is going to be a huge theme of the next 6 months
Finally got the idea across to a skeptical friend that AI is powerful and dangerous even without extinction-level events and let me say, walking through specific risks in detail did not make me feel better.
September 17, 2026 at 7:57 PM
Finally got the idea across to a skeptical friend that AI is powerful and dangerous even without extinction-level events and let me say, walking through specific risks in detail did not make me feel better.
"GenAI bad, unlike traditional ML, which we were all cool with because cancer screenings" is ahistoric nostalgia. People hated ML for all kinds of sins that mostly existed in the "Fairness and Bias" literature or the obvious bad actors like Facebook.
September 17, 2026 at 6:56 PM
"GenAI bad, unlike traditional ML, which we were all cool with because cancer screenings" is ahistoric nostalgia. People hated ML for all kinds of sins that mostly existed in the "Fairness and Bias" literature or the obvious bad actors like Facebook.
Reposted by Crazy in Moloch
ai ea e/acc p(doom) singularity slowdown arms race cyberwarfare drone swarm x-risk consolidation decentralization destabilization —
yes. but have you ever heard the ancient refrain of a midsummer song echoing through the forest as you approach a sacred spring on the solstice?
yes. but have you ever heard the ancient refrain of a midsummer song echoing through the forest as you approach a sacred spring on the solstice?
September 17, 2026 at 3:59 AM
ai ea e/acc p(doom) singularity slowdown arms race cyberwarfare drone swarm x-risk consolidation decentralization destabilization —
yes. but have you ever heard the ancient refrain of a midsummer song echoing through the forest as you approach a sacred spring on the solstice?
yes. but have you ever heard the ancient refrain of a midsummer song echoing through the forest as you approach a sacred spring on the solstice?
Finally, took long enough.
Part-human part-mouse brain developed in science breakthrough
Researchers at Stanford University genetically altered mice to be able to receive and function with human brain cells.
www.bbc.com
September 17, 2026 at 4:00 AM
Finally, took long enough.
Hard to say what "fable-level" means with such radically different capabilities, but I'm excited for Jev. LLM-driven behavior trees and workflows don't always need text generation, and with that pricing I foresee a lot of "use Jev to check for problems & route to an appropriate model to remediate."
Jev: Fable-level model that doesn’t charge for output tokens because they’re too cheap to meter
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...
it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok)
typesafe.ai/blog/introdu...
September 16, 2026 at 1:54 PM
Hard to say what "fable-level" means with such radically different capabilities, but I'm excited for Jev. LLM-driven behavior trees and workflows don't always need text generation, and with that pricing I foresee a lot of "use Jev to check for problems & route to an appropriate model to remediate."
The robotic demon centaur marketing clearly worked on me, because my first thought on all these is now "can I stop it with a knife?"
Digit 5 May Be the First Humanoid Robot Worker That’s Truly Safe
Digit 5 May Be the First Humanoid Robot Worker That’s Truly Safe
Agility Robotics’ latest is big enough, and safe enough, to do useful work
spectrum.ieee.org
September 15, 2026 at 3:32 PM
The robotic demon centaur marketing clearly worked on me, because my first thought on all these is now "can I stop it with a knife?"
Labs could have bought a fair amount of good will if they started by fixing whatever's wrong with Bluetooth-as-deployed instead of offensive cyberwarfare.
September 15, 2026 at 2:16 PM
Labs could have bought a fair amount of good will if they started by fixing whatever's wrong with Bluetooth-as-deployed instead of offensive cyberwarfare.
D'ya like DAGs?
September 15, 2026 at 11:29 AM
D'ya like DAGs?
Reposted by Crazy in Moloch
METR, the AI Safety & Eval firm, experienced a breach in which attackers stole $600k worth of API credits
thehackernews.com/2026/09/atta...
thehackernews.com/2026/09/atta...
Attackers Steal METR API Key and Consume AI Credits Worth About $600,000
Attackers stole a METR API key and consumed credits worth about $600,000, while a later campaign failed to access internal data.
thehackernews.com
September 14, 2026 at 7:45 PM
METR, the AI Safety & Eval firm, experienced a breach in which attackers stole $600k worth of API credits
thehackernews.com/2026/09/atta...
thehackernews.com/2026/09/atta...
Was there ever good research into what the hell happened with 4o? Surely someone gathered a bunch of transcripts of comparable models from the era and could come to a better conclusion than "sycophancy".
September 14, 2026 at 7:53 PM
Was there ever good research into what the hell happened with 4o? Surely someone gathered a bunch of transcripts of comparable models from the era and could come to a better conclusion than "sycophancy".
Who's making an RTS with voice controls and user-configurable unit/squad-level behavior trees? Has to be in the works by now.
September 14, 2026 at 6:16 PM
Who's making an RTS with voice controls and user-configurable unit/squad-level behavior trees? Has to be in the works by now.
Reposted by Crazy in Moloch
my take is that an alarmingly low number of people understand just how bad cyber warfare alone could go. doomers going on about x-risk but not the very real and present problem of infra cyber defense and the very disastrous consequences that could have
I'm not an AI doomer at all but it is concerning how people are dismissing doomers by saying things like "how would it even find another machine to copy itself onto", "how would it even get money", and "several terabytes large is so huge that it would be infeasible to copy itself many times"
September 13, 2026 at 4:32 PM
my take is that an alarmingly low number of people understand just how bad cyber warfare alone could go. doomers going on about x-risk but not the very real and present problem of infra cyber defense and the very disastrous consequences that could have
Reposted by Crazy in Moloch
People like to point at the xkcd comic with the block tower held up by a tiny project maintained by one guy but I actually think this one explains far more about the state of digital infrastructure
September 13, 2026 at 3:17 PM
People like to point at the xkcd comic with the block tower held up by a tiny project maintained by one guy but I actually think this one explains far more about the state of digital infrastructure
Milkshake duck strikes again
September 13, 2026 at 2:44 PM
Milkshake duck strikes again
Uh Astra, you feeling okay?
September 12, 2026 at 4:58 PM
Uh Astra, you feeling okay?
Someone's got to be working on Advent of Conjectures 2026, right?
September 12, 2026 at 3:31 PM
Someone's got to be working on Advent of Conjectures 2026, right?
Reposted by Crazy in Moloch
Reposted by Crazy in Moloch
September 12, 2026 at 1:53 AM
Reposted by Crazy in Moloch
