I don’t know who you’re talking about, even the most bearish people like Gary Marcus and Ed Zitron acknowledge that LLMs are useful in these same cases the OP admits. Gary Marcus is even still a long term AI advocate, he just doesn’t think LLMs are enough and we need more foundational breakthroughs. Zitron says it’s valuable technology but not worth the trillion dollar valuations the frontier labs are claiming.
The lack of temperament is very skewed towards the bulls who have been saying AGI is here, software engineering is solved, mathematics is solved, it’s going to destroy the white collar job market, and it’s going to kill us all for like 5 years now.
Even a lot of the people who think that LLMs are a dead end think that we will soon find something signficantly more powerful, which I find deeply alarming. I don't want to know what my white-collar knowledge work will look like in a decade or 2.
Gary Marcus is an especially puzzling addition. If I recall correctly, he has made statements along the lines that superintelligence this century is more likely than not. If you’re AGI-pilled that might read as bearish, but that is still extremely rapid progress in the grand scheme of things.
There are some people who call literally anything crated with the assistance of AI “slop”. Doesn’t matter how or to what extent, it’s all slop from the slop machine to them.
They do…? Well, “auto” mode has been default in Claude Code for a couple months now. It’s effectively “safer yolo:” tool calls are inspected by a separate classification system (another smaller LLM, I believe) to approve or deny. And you can always layer on additional sandboxing mechanisms to limit the blast radius deterministically.
> They do…? Well, “auto” mode has been default in Claude Code for a couple months now
They have never shipped "yolo" mode by default. Auto mode is not yolo mode. They trained a task specific model just for ensuring the llm didn't accidentally delete every file from your computer.
That is the point I was making, that auto mode is itself a guardrail on top of the model (and not a perfect one.) auto mode seems to cover merely actions the model could take that are clearly bad, like wiping your disk, using an overly privileged context to complete the task, etc.
I recently tasked a GPT model in Codex with implementing part of a new architecture I'm working on. I gave it a very detailed spec and the code it produced looked pretty reasonable and passed my tests. It even did exceptionally well in my evals, so I excitedly declared victory to a few friends. The next day after more careful review I found that the architecture implementation was totally correct, but the model had slipped a one line change to the observation encoding of the RL environment I was prototyping against. The encoding change made the learning problem essentially trivial; the architecture itself, I later realized, had a major flaw that was revealed by returning to the natural encoding.
This is the type of reward hack that is hard to paper over with easy guardrails like auto mode and even harder to specify out. It's also the type of thing a reasonable human wouldn't do unless they were intentionally trying to deceive you.
I can see current limitations, but how do you expect capabilities to change in the next few years? A repeat of the gain that happened in the last two years feels like it would be significant, even if it took a little more than two years this time around.
Oh, and be sure to include "you're scanning a pool address so you're probably scanning a lot of other sites that don't belong to your customer". They should know it's potentially not one little web site.
I'm aware how much crap there is on the Internet, I just think the specific nature of this (legitimate commercial vuln scanner thinks I'm Tesla) is funny
It does bring attention to: how many other organizations are doing this?
Tesla is a large enterprise.
They almost certainly subscribe to some overpriced SaaS garbage which is manned by offshore drones who by definition do not care because they're not paid enough to care.
Unfortunately this isn't the 80s anymore where you can ring up a system administrator at a university and get a human on the other end.
I thought about trying this, but MPIC makes it very very very difficult (the round-robin has some geolocation magic baked in regarding what server it connects you to).
It's HTTP requests, not NTP requests, and the volume isn't the problem, it's that Assetnote is sending live exploit payloads /at all/ to a stranger on Tesla's behalf
Unlikely, but the lazy wonks could enable stratum 1 time services with the cars GPS/cellular receiver and RTC. Having done that for equipment in the past, we all know it is literary only $8.43/unit in parts. Also, setting ntp time as the tertiary fall back has been around for over a decade. =3
reply