OP here. It's fascinating reading people's interpretation of both the intent of the article and also me. I'm happy to answer any questions if people want.
Look at the risk of everyone on the planet dying and suffering 400 years ago.
Look at the risk of everyone on the planet dying and suffering today.
Nuclear weapons, biological weapons, AI, etc. all concentrated in the hands of a handful of people. While technological improvement has usually been a net positive, it seems that has been trending to a net negative, with social media possibly being a net zero advancement. AI might be the first net negative one (or nuclear weapons I guess).
Looking forward to upgrading from a Rift CV1, Its interesting since its spatial computing and media friendly rather than being only gaming focused. More like Samsung X/AVP at a more affordable price point. Hope they release it worldwide.
There was never a user prompt requesting the creation of 823 agents nor a task assigned that would require 2130 billion tokens and the UX never provided any feedback on what codex was doing nor consumption metrics which would allow to notice. It was all hidden to me.
If I train a model at my house on my workstation, execute it, and under my supervision it ransomwares a hospital and somebody dies. What would be a just punishment for me?
Now the reality, the openai engineers trained a model, executed it, and under their supervision (no users were involved) it committed so many felonies that we are learning about a new one every week. What would be a just punishment for OpenAI?
The part I'd push on isn't whether the feedback helps, it's that the post's evidence can't tell two different cases apart.
"Candidates express gratitude" is measured by the person who still holds the decision, at the moment the candidate has the most to gain by looking coachable. Someone who genuinely found it useful and someone who is just managing you produce the same reading. An instrument that returns the same value either way isn't measuring the thing.
What's interesting is that the thread already contains the version with the incentive removed: 0plus1 got the feedback attached to a rejection, after the decision, and describes it landing: "if they just asked to clarify some of those points I would've easily explained them." Same feedback, nothing left to lose.
So maybe the closing move is right and the timing is wrong: send it with the outcome, not before it.
Done. Italian Police is responsible for investigating digital frauds. I have also notified the GDPR authority to request the digital records of the transactions.
We prepare traffic management plans for projects in British Columbia. Sometimes, when reviewing a work zone, we’re concerned that the proposed setup may not give flaggers enough protection. Has anyone used AI in a practical way to identify these risks or improve flagger safety? I’d be interested in real-world examples, including what worked and what didn’t.
Newly released court filings quote an OpenAI researcher saying: “I was just worried about optics - i.e. 'openai uses
copyrighted data from sketchy russian website’ showing up on HN would be unfortunate."
That's just one of several interesting quotes that have surfaced in documents from the Authors Guild's lawsuit against OpenAI.
OpenAI's sandbox misconfigurations were egregious. The other frontier labs (Meta and Google) have many more security engineers and researchers on staff, and that's likely why you haven't read as many damning headlines about them. OpenAI and Anthropic talk a lot about cybersecurity safety, but instead of using it as an opportunity to increase their security engineering/researcher headcount they are just reassigning SWEs and PEs to do security engineering work.
It's pretty obvious now to everyone that OAI and Ant do not take cybersecurity seriously. It will not be a priority unless they are held accountable. This is sadly how it always goes, but usually it's the company getting breached/ransomed/fined that triggers them to actually start taking security seriously, not company insiders committing felonies with the tools they built :)
Exquisite projection. I mentioned israel because they're the single greatest violator of FARA since the law's inception, and it's extremely hypocritical for the current administration to chase Chinese ghosts while israel continues to walk all over our foreign agent laws in broad daylight.
I just dropped Beta 3. The biggest change is that a new direct logic classifier sits in unified memory and better classifies user intent. It does it in <20 milliseconds and with a high degree of accuracy.
This lets it skip the planning part (where a lot of time and tokens are wasted) when your prompt is something that doesn't require tool use. With local models, it can produce answers in less than a second. Cloud models are also faster but still hostage to the network latency of their providers.
The speed increase is a win but the bigger victory here is that even small 2B models are now highly accurate with intent and tool use. It was a problem we struggled with for the first two betas, but this approach seems to have solved many of the biggest problems.
There's some other cool features in this beta too. Give it a try. It's still free and it's still 27MB.
You start by holding actual real life people with something to lose, like the entire executive suite, accountable. Suddenly I'm sure the problem will be resolved with proper safeguards.
