It is likely that many airlines accommodate people who are being deported as regular passengers on outbound international flights.
For movement of humans at industrial scale, though, there are only a few operators with ICE contracts. Avelo, GlobalX, Eastern, Key Lime Air, and Omni come immediately to mind. For international flights, there's at least one Learjet operator that flies a bunch for them, too.
True. Applications as these go back a few decades. From the news buzz when it was launched, ASCI White was similarly used in early '00s to understand nuclear explosions and shockwave propagation (instead of relying on live tests) - classic CFD problems[1] Successor supercomputer clusters were also used to do weapon design & nuclear physics simulations. One supercomputer IIRC even simulated a tornado genesis.[2]
I can only imagine the classified applications must have grown ten-fold in complexity in the interim.
Yeah I was thinking they probably picked a public use case that looks a lot like their classified workloads...
The article mentions the simulation ran quickly, the time spent was debugging. Suggests to me the real classified system will be much more capable.
Who decides what behaviors we should be nudged toward? Is it us, or someone else?
To me, one of the greatest dangers of the present moment is that we can't tell whether the LLMs are being asked to give subtly biased answers (or product-placement) on some questions. One cannot readily tell from the output.
And I agree with you, but an even bigger problem then is "how do you even make a verifiably trustable LLM?"
The training compute footprints are enormous, several orders magnitude beyond what the average person has access to. Even if a company came out and said "here's our completely open-source model. All the training data, the training procedure, and here's the final-product model."
Maybe you could hire an auditing company? But how long would it take to audit? Would the state of the art advance drastically in the time between?
And people like to keep downvoting my "Make Classwarfare MAD Again" but like I'll wager 90% of people on HN are on the losing side of the war.
For as long as Backblaze has been doing this and at this level of quality, I have no doubt that these reports are good for business.
(As an anecdotal example -- I first heard about Backblaze from these reports many years ago and have relied on them to an extent in selecting new drives. I'm now a Backblaze customer.)
If not now, soon, the bottleneck will be responsibility. Where errors in code have real-world impacts, "the agentic system wrote a bug" won't cut it for those with damages.
As these tools make it possible for a single person to do more, it will become increasingly likely that society will be exposed to greater risks than that single person's (or small company's) assets can cover.
These tools already accelerate development enough that those people who direct the tools can no longer state with credibility that they've personally reviewed the code/behavior with reasonable coverage.
It'll take over-extensions of the capability of these tools, of course, before society really notices, but it remains my belief that until the tools themselves can be held liable for the quality of their output, responsibility will become the ultimate bottleneck for their development.
I agree. My speed at reviewing tokens <<<< LLM's token's. Perhaps an output -> compile -> test loop will slow things down, but will we ever get to a "no review needed" point?
https://www.youtube.com/watch?v=9O34BnFu8Kk