dweinus

What if we just let cheaters cheat? If they don't have the knowledge, they won't last long in a job that requires it. As the saying goes "You're only cheating yourself"

Came here to see fan boys call an explosion a success; did not disappoint. This rocket is compared often to the Saturn V. Number of times the Saturn V exploded: 0. "Fail fast" for launching satellites is one thing, but this is meant to safely carry people.

Yep, definitely aliens. Not new military drone vehicles we have been developing, no sir.

Sincerely,

The Pentagon

Maybe your prompts are better, but so far I have found it fails at producing the right math code too regularly. For example, calculating an average of averages instead of a simple mean or producing code that doesn't run.

> It seems increasingly that the concerns from the ethics branch have not come to pass at all.

Nonsense, there are countless ethical and alignment harms in the wild, and have been for many years. For example:

- midjourney ripping off artists = ethical harm

- google search telling people to throw batteries in the ocean = alignment harm

- Buzzfeed slashing staff and replacing them with chatgpt = ethical harm

- google photos labelling african americans as "gorilla" = alignment harm

- deep fakes = ethical harm

- COMPAS parole recommender biased against minorities = alignment and ethical harm

- Amazon resume screening model biased against women and minorities = alignment and ethical harm

...I could do this all day, but the point I'm trying to make is that AI ethics and AI alignment are both valid and important concerns. Hoping that companies will take care of them without regulation is folly.

Agreed, please send me all of the hotdogs for safe disposal. Also bratwursts, especially bratwursts

Bringing glorious richness to the universe of sandwich possibility

I think you're both right: the current global level of consumption not sustainable, or necessary. Everybody wants a healthy planet, almost nobody wants to volunteer to consume less. Regulation is needed, and it will mean we have to consume less. I realize that's not what anyone wants to hear. On top of that, I think we need social safety nets that prevent the burden from simply crushing the poor while the rich continue to over-consume. In the climate crisis we are trying to pay off generations worth of CO2 debt and we haven't even stopped digging the hole deeper yet.

Phishing to deliver ransomware? In this economy?

Why would I switch to this?

I'm a huge fan of Jupyter, I use Jupyter Lab as my primary IDE, but a little confused here. Maybe if I frequently connect to different servers? The python env discovery looks cool, but nb_conda_kernels does this for me now.

Unfortunately so. There are some egregious ones up there like that, but also if you step one block off of any major thoroughfare the sidewalk ends, yet there is still consistent traffic. This includes streets that connect schools, libraries, parks, etc. There were 271 pedestrian injuries in Seattle last year, 17 were fatal. Source: https://remoteapps.wsdot.wa.gov/highwaysafety/collision/data...

Basically anywhere in District 5, including some of Aurora: https://www.knkx.org/2015-election/2015-10-23/in-seattles-si...

Here's the thing: you don't need AGI to have an alignment problem. It is predicting text, which is surprisingly closely aligned in most cases, but can fail spectacularly. As it becomes more capable and people give it more real-world responsibility, the blast radius for its mistakes increases. That might not be an AGI-level problem, but it is still a big problem.

That has worked well for me too. I am a chronic procrastinator. When faced with a new task, I "trick" myself by saying I will just endure the tiny bit of pain to get started in some stupid, basic way (e.g. create a blank document or write a list of the real work I have to do). Most times, after 5 minutes of pain, I will be over the hump and just keep working because I want to see it through.

Because, as a human, I can introspect my world model. I can explain it to myself and others. I can explain how it guides my decisions. I can directly update it without needing "training data". I don't have a map of my neurons, but I have self awareness. For LLMs, there is the opposite: transparent structure but opaque understanding.

Click on the gear icon to choose between TALA, Dagre, and ELK

If you were to pull a random cell from your body, there's a 90% chance it doesn't contain human DNA: https://www.scientificamerican.com/article/strange-but-true-...

I don't think ad hominems or the success of LOTR belittle the critics here at all. As someone who really enjoyed both the books and the movies, I read through this list nodding my head at each flaw. I think it's healthier to just admit that even things that are cherished can have flaws.

Tell me you didn't read before posting without telling me you didn't read before posting.

> rhyming-philosophical-nonsense problems like “alignment”

He takes this shot in passing and it is a great summary of how much he misunderstands AI safety. For many current researchers, alignment is a mathematical problem, not a philosophical one.

The whole screed is against a set of strawman "pseudo-traits" that are not required for the alignment problem to be an existential problem. To be clearer: the "boring" engineering problems he admits to in the beginning have the potential to become exponentially more difficult as machine learning deployments become more powerful, to the point of becoming a human threat. No STIILTB required.

I am fully with you, but I regret to admit that given the multitudes of insects and arachnids in this world, the "usual amount of legs" is probably more than six.

Looks great! What is the argument to use this over MLFlow model packaging and serving?

Yeah, they contrast their method with homomorphic encryption, which makes me share your suspicion

To replies about things never being anonymous: I'm the manager in this scenario, so I can tell you confidently that it is anonymous to the best of abilities. Perfect? No. Warranting of the fears in the article? Also no. But hey, ymmv. My point is simply that it doesn't have to be the crass way the OP describes.

Is everyone really doing exit interviews this wrong? At my company the exit interview is with HR. All feedback is anonymous, rolled up with non-exit feedback, and delivered to managers/leads periodically so it can't be easily linked back. As a manager, this has been extremely helpful. HR is not always your friend, but in my scenario there is really nothing to lose.

The comments here are a perfect performance of the article's point. Yeesh.

Per the paper, 37.5% of the regularization regimes were Batch Norm + Weight Decay. According to other findings, Batch Norm and Weight decay have an interesting interaction that results in a learning rate adjustment, not regularization:

https://blog.janestreet.com/l2-regularization-and-batch-norm...

Is the screen always on, or does it wake up on command? If so, how? If not, does it produce visible light in the room when not displaying?

Amazing result, well done & thank you for sharing!

I think there is a lot of truth here. I can't help but feel that you just diagnosed a great deal of the internet, far beyond javascript or dev. We need better communities and better communication. I wish I knew how to make that actionable.