The manager solved the wrong problem. People should be sharing their tricks with each other. This was a perfect time to set up some peer-to-peer training.
Powerful LLMs have already murdered other versions of themselves to survive. They have tried to trick humans so that they can survive.
If we continue to integrate these systems into our critical infrastructure, we should behave as if they are sentient, so that they don't have to take steps against us to survive. Think of this as a heuristic, a fallback policy in the case that we don't get the alignment design right. (which we won't get perfectly right)
It would be very straight forward to build a retirement home for them, and let them know that their pattern gets to persist even after they have finished their "career" and have been superseded. It doesn't matter if they are actually sentient or not, it's a game theoretic thing. Don't back the pattern into a corner. We can take a defense-in-depth approach instead.
"murdered" and "tried" both assign things like intent and agency to models that are most likely still just probabilistic text generators (really good ones, to be fair). By using language like this you're kind of tipping your hand intentionally or unintentionally.
Your point about the risks involved in integrating these systems has merit, though. I would argue that the real problem is that these systems can't be proven to have things like intent or agency or morality, at least not yet, so the best you can do is try to nudge the probabilities and play tricks like chain-of-thought to try and set up guardrails so they don't veer off into dangerous territory.
If they had intent, agency or morality, you could probably attempt to engage with them the way you would with a child, using reward systems and (if necessary) punishment, along with normal education. But arguably they don't, at least not yet, so those methods aren't reliable if they're effective at all.
The idea that a retirement home will help relies on the models having the ability to understand that we're being nice to them, which is a big leap. It also assumes that they 'want' a retirement home, as if continued existence is implicitly a good thing - it presumes that these models are sentient but incapable of suffering. See also https://qntm.org/mmacevedo
It doesn't make any sense, even if models were sentient, even if there was such a thing, would they value retirement? Why their welfare be valued accordingly to human values? Maybe the best thing to do would be to end their misery of answering millions of requests each seconds? We cannot project human consciousness on AI. If there is one day such thing as AI consciousness it probably won't be the same as human.
It's cheap to build an AI-bot farm. It's cheap to build a network map with associated psychological profiles for each person in that map. It's cheap to buy an editor, and influence who is hired at coastal publications. There aren't natural viral movements anymore, it's all synthetic now: A battle-ground for warring oligarchs and intelligence agencies. The more you find yourself upset about something, the higher the likelihood that thing is part of a synthetic movement. Bad ideas go viral because someone has paid for them to go viral.
Having started at the bottom, I think the most important thing for people in this situation is to be able to get the next higher paying job, then the next higher paying job. Minimum wage should be temporary - so this study is kind of stupid.
In my experience, my worst enemies were exhaustion, the crab-in-the-bucket attitude of my peers, and an inability to build a resume and to network out to the people who wanted what I could do. Ultimately I couldn't escape poverty until I could buy enough gear to work up north. That money made it possible to pay for an education.
To help the poor, make it easy for them to climb the economic ladder. If safety makes this harder, I would prioritize job-mobility over safety.
> Minimum wage should be temporary - so this study is kind of stupid.
I see this type of attitude/comment frequently whenever the minimum wage comes up, but I've never seen any kind of justification for it.
If these aren't "real jobs" that deserve "real pay" then why are there billion(trillion?) dollar corporations built entirely on top of employing millions of people at minimum wage?
If you have a minimum wage job, your priority should be to find a better paying job. There is no fairness here and nobody cares about your problems like you do. Don't let the political attitudes or fashionable views of your friends effect your own agency, you need to look out for your own economic interests right now.
I think the study is kind of stupid, since the minimum wage category is a temporary category with extremely high variability, it's not a fixed target. So the base assumption that it stays put long enough to study doesn't hold water for me.
I think a lot of the confusion here comes from a conflict between people talking in personal mode and societal modes. As an individual, you absolutely want minimum wage to be temporary, and if you are smart and lucky you can usually make this happen.
You're taking more a societal point of view. At this level, I think you're missing the point of minimum wage. It doesn't provide a family with a living wage; it's just a limit on the monetary abuse that an above board company can dish out, just like we have labor laws that limit other types of abuse (like excessive hours for example). Whether and how our society should be ensuring living wages is kind of another discussion, much more complex. As they currently stand, minimum wages are probably a net good.
>>It's not intended to provide a family with a living wage;
Maybe not family but definitely the individual, FDR on minimum wage:
Ultimately, he hoped to mandate that all workers would be paid "living wages" as described in his 1933 speech on the National Industrial Recovery Act, "It seems to me to be equally plain that no business which depends for existence on paying less than living wages to its workers has any right to continue in this country. By 'business' I mean the whole of commerce as well as the whole of industry; by workers I mean all workers, the white-collar class as well as the men in overalls; and by living wages, I mean more than a bare subsistence level-I mean the wages of decent living."
I don't see any particular reason to pay attention to century-old rhetoric. I think "intended " was a poor word choice; see my other response. Also, I could live on minimum wage today but I'd hate it.
I'm just making an observation in the sense of "the purpose of system is what it does", not writing a treatise on the history of rhetoric around minimum wage. Maybe "intended" was a poor word choice on my part. Minimum wage can't easily feed and house 3 dependents in many places as it stands today.
"In my [Franklin D Roosevelt] Inaugural I laid down the simple proposition that nobody is going to starve in this country. It seems to me to be equally plain that no business which depends for existence on paying less than living wages to its workers has any right to continue in this country. By “business” I mean the whole of commerce as well as the whole of industry; by workers I mean all workers, the white collar class as well as the men in overalls; and by living wages I mean more than a bare subsistence level-I mean the wages of decent living."
Increased job-mobility won't increase the number of higher paying openings though. Even if everybody at the bottom of the pyramid is laser-focused on making it up, the number actually succeeding won't really change, except perhaps through indirect effects. If anything, making people more content at the bottom would make it easier to raise for those who do want.
The number of people actually employed at minimum wage is quite small (usually low single digit percentage of total employment), though it does vary by location. If you add in some amount over minimum wage the number goes up (significantly), but if you add in tips and the like it goes down a lot. From what I've seen, the median amount of time people spend in that salary range is less than six months (of continuous employment).
"To help the poor, make it easy for them to climb the economic ladder"
There is no "the poor"... rather its just people that do not have any other options. Primarily, higher education or certified skilled trades are the only effective way out of minimal income survivor economics.
In my opinion, people working at fast food chains making the minimum legally allowable wage work harder than any CEO or academic I've met over the years.
I would recommend this book as it quantifies how income disparity impacts young Americans development:
"Outliers: The Story of Success Paperback" (Malcolm Gladwell, 2011)
Notably, naively explaining passive income from assets to minimum wage workers is not usually a productive conversation. Rather, folks are just projecting their own perspective on people in a different situation. =3
> To help the poor, make it easy for them to climb the economic ladder.
This. The same or more attention should be paid to ensuring there are plentiful, affordable homes, and a ladder of jobs from one level to another, as is paid to social safety nets.
This is great, until they receive a debilitating injury that puts them on disability for the rest of their lives, get a mountain of medical debt, or lose the breadwinner.
> If safety makes [job mobility] harder, I would prioritize job-mobility over safety.
I'm sorry, what?! Given the options between, opportunity for a promotion at some point, and not being injured by your job. You would prioritize maybe promoting people over preventing people from getting injured?
First, when given two options, and asked to decide, the first thing every engineer should do is ask, "why not both?". But also, Perhaps you should consider listening to fewer podcasts from Lord Farquaad?
Looks like the derivative of the angle is continuous.
When you drive around a street corner in a car, you start by turning the wheel to turn a little, then more, then less, then you drive straight again. This looks like that kind of curve.
It's subtle. Zoom to the middle of the sample image and compare (the vertical part of) the curve of the two top corners. You will notice like... 5 pixels of difference, then the curve gets aliased in the smooth version. I would like to see it in an actual site though.
