I mean these machines take massive scale compute- they’d have to some how distill themselves, bootstrap a distributed inference runtime that can run across many lossy unreliable machines. The idea of the AI running away from us is probably unrealistic. I’m more interested in bad actors using unaligned AI for bad things.
> Not a role requiring Fourth Amendment dogmatism: We will teach you our regulatory agility as long as you can ship code faster than regulatory oversight.
Is this a troll post? I’m pro-LEO and generally pro-Flock as a heavily auditable investigatory tool but intentionally circumventing the 4th amendment is outright evil.
I literally work on this. The problem is that enforcing device attestation impacts DAU so platforms don’t want to do it but it’s effective. Once the internet is unusable, device attestation will be required to access services. Ad driven businesses will start to see demand from ad buyers for non-bot traffic etc.
It seems clear to me we are moving towards a world where you will have to perform device attestation to access the internet. The spam/abuse is too great and accelerating.
People don’t want their incompetence displayed. The session is probably a series of “continue” prompts with basic understanding demonstrated in the initial ask.
It is quite accessible and is read by teenagers during their angst. It appeals to people become it is a simple premise that gives simple choices to the reader.
I mean, does enabling ICE to be more effective actually represent a harm? ICE has authority to do the things it’s doing. The courts are the appropriate venue to check that authority. Two agents murdering citizens is a criminal matter completely unrelated to Google’s relationship with DHS.
It feels like AI empowering scams and facilitating post-truth is more harm.
This isn’t escaping in the same sense- the model was executing within the OpenAI infra. If it ported its entire architecture/weights into a public cloud to survive being turned off… that’d be pretty cool.
I imagine they’re building the supply side as well, where you can bid your fixed GPU. My question with this stuff is how do I know my data is secure on these for-rent boxes? What steps are taken to sanitize the GPU between runs etc.
I have 3 3090s, all in various gaming computers. I’m not sure what to do with them- the VRAM is only 24GB so doesn’t seem worth the power costs to run an open weight model. Any suggestions on how I can run them as some type of cluster? Is it worth exploring NVLink?
Fable/Opus 4.8 outperform Codex 5.5 for me at the general architecture/refactoring/performance work I’m doing, to the point where it’s not worth using Codex. Codex will often spit out non idiomatic code that overcomplicates things.
I am wary of any supply chain attack and more so if the project is maintained by people with relationships in adversarial countries. The risk of exploitation outweighs the convenience.
Yeah this is true on my team for two archetypes: the new junior engineer and the old staff engineer that was too “busy” to ever actually write code. Now he’s just busy in a different superficial way.
If anything it’s unglued me from my computer. I’ve been able to keep an agent working on a project while on long runs, bike rides, in transit. Much of our development workflow is the human in the loop refinement cycle now.
My point is that it being a choice starts to disappear with kids, rising costs etc. It was a choice I could make independently, now it’s a choice the entire family gets to participate in since we’d need to sell the house, move to a lower cost of living locale etc.
If I remember correctly they linked it to breast cancer, causing all research and prescribing to basically disappear over night. It took 20 years before the study was revisited and the link dismissed.
This was also my plan; hang in there for the next kid, wait for the next vest, etc. ultimately daycare is expensive, healthcare is expensive… I’m still stuck 6 years later waiting to quit but something always comes up and there’s always another vest around the corner.