directevolve

The practice of effective altruism, as distinct from the EA movement, is good for our culture. If you have a lot of money or talent, please think critically about how to leverage it efficiently to make the world a better place.

Doing that doesn’t buy you personal virtue. It doesn’t excuse heinous acts. But within the bounds of ordinary standards of good behavior, try to do the most good you can with the talents and resources at your disposal.

There are explanations for the equal happiness stats other than the validity of IQ tests. People’s top goal might be something other than happiness. Happiness and other goals might trade off. Happiness reports may be relative to an expected baseline that’s higher for smart people.

We should factor medical tasks into a larger number of specialized roles akin to phlebotomists and dental techs.

These roles should perform highly in-demand, relatively straightforward and repetitive tasks that don’t require complex medical decision-making, where training can be efficiently scaled up.

An example that currently doesn’t exist would be a specialist who can prescribe short term courses of drugs like methodone for opioid addiction as a bridge to longer term care by a doctor. This would enable us to have bridges to treatment readily available all over cities whenever an addict walks in ready, perhaps only for that moment in time, to start treatment.

I make 1-2 flash cards on the topic and main advance of papers I read. It’s a nice sweet spot, and I find that I retain much more than what’s on the cards. Without them, I forget the whole thing.

Lack of understanding would mean you haven’t decided meaning from the words on the page. He gives the example of memorizing a German textbook without speaking German.

Memorization without learning means that you don’t know the relationships between individual facts that give them meaning and relative importance that lets you make wise decisions about what to memorize and what not to for your purposes.

So: make sure you understand what the words and phrases you’re reading mean (understanding). Look up terms and definitions. Identify the main points of the sentence, paragraph, page and chapter and why they matter (learning). Then memorize those main points, starting with the most important basics.

Does insincerity, cruelty, unfriendliness, and impatience make a community less likely to be a cult?

It depends. If I was also trying to invent a lightbulb, or maybe develop new materials for you to try as filament, I might be very interested to know what you’ve already tried.

Updates as marketing? Makes sense, but not well-executed by OP.

The update titles are good, but try focusing on the feature, not the “toil.” If I’m using your product and want the feature, you’re telling me what I already know. Keep text to a minimum.

Revised first example (duplicate button):

You can now create repeating events by clicking “Options > Repeat.” Sync with company holiday schedule to skip repeat events on holidays.

Conservatives don’t seem happier than liberals.

I’m quite confident the conservatives I know are fairly unhappy and unstable people, so is there a population of well-adjusted conservatives I never meet, or is this self-report bias?

I’m waiting until I get some concrete examples of what this happiness gap looks like in practice before I take these results too seriously.

What HN has for the rationalist movement isn’t just annoyance - it’s deep contempt and hatred.

True, but then terrorists need to have access to AI bombs, drones and AI targeting systems.

Counterpoints and questions:

- Non-wired drones can be jammed. It’s early days for building defenses against these attacks.

- Non-state actors have far less access to the sophisticated intelligence needed to strike hard targets or secure against counter strikes.

- Setting up hidden bombs for remote detonation on soft targets, like the freeway, has been possible, no need for drones. What other factors have been preventing these types of attacks? How do drones change those factors?

If America was hunting Osama bin Laden today, I bet we’d have used a drone to kill him rather than sending in special forces. Likewise, if I was a cartel in the jungle or rebel force in the mountains, I’d be damned scared of the military or police coming after me with an endless wave of drones.

For me, the keys are:

Break down tasks into micro-tasks. Doesn’t have to be written down.

Get my body involved. Take handwritten notes when reading.

Create micro-deadlines. Have short meetings with colleagues to share progress.

Inject mini-rewards. Do one 20-minute task. Then watch one music video on YouTube.

I could see room for a productivity app that smoothed out this workflow.

The article’s not titled precisely - it’s specifically about LLM disasters, not AI disasters in general.

One option is to put your thoughts down in great detail on paper, then feed it into AI and ask it to design you a slideshow with text and suggestions for visuals.

The reason we use animal models is that we don’t know enough about animal biology to be confident we can simulate it accurately.

However, computational tools that help biologists “replace, reduce, and refine” animal experiments are helpful. That catchphrase is a good one to Google to look for previous efforts.

Building computational tools to help scientists improve the design of necessary animal experiments is certainly worth working on.

It’s not “group blame” - it’s accurately naming the specific wealthy people who put this kleptocrat in office for a tax break.

I did this for a summer in a beautiful spot in rural Oregon as I was contemplating a career change, living out of my Prius compact. It was a special time, though not something I would do long term.

Why do you think there’s an ethical problem with this type of blood transfusion?

They can’t be fired, but if they stop bringing in grants, they can face steep pay cuts and lab closure.

Interesting, what type of dev work do you do? Performance does vary widely across languages and domains.

We’re already heading toward the sigmoid plateau. The GPT 3 to 4 shift was massive. Nothing since had touched that. I could easily go back to the models I was using 1-2 years ago with little impact on my work.

I don’t use RAG, and have no doubt the infrastructure for integrating AI into a large codebase has improved. But the base model powering the whole operation seems stuck.

You definitely don’t have the intelligence of the entire human race on Claude. You have a generative model of most of the published, digitized text the human race has produced. I’m happy to call it intelligent, and, as a daily user of the most recent pro LLM models, extremely useful for tasks like coding and, to some extent, search.

But as soon as I turn from implementing and debugging code to conceptual thinking or building a mental mode of a problem domain, LLMs currently function for me mainly as Grammarly on steroids. There’s no one tool in this space that’s supreme - Zotero, DuckDuckGo, Google Scholar, pen and paper, draw.io, and Google Docs all play a role. I’m a PhD student, YMMV, the LLMs do improve over time, etc.

For fast, physically intuitive, flow-of-thought expression, there’s nothing like pen and paper, for me.

I throw away almost all my handwritten notes, because of the downsides: illegibility, unsearchability, physical bulk.

But the qualitative difference of thinking or reading +/- handwriting is huge. I don’t get caught up in editing. I can draw diagrams easily. It slows me down, makes me dwell on an idea while I’m writing it, and thereby physically calibrates my thought process.

I live in Portland. Walk around certain neighborhoods, and you’ll see items out on the street - mostly garbage. Broken couches, broken tables, broken appliances, that sort of thing. Typically, items with remaining utility are exchanged via Goodwill or an online tool like Buy Nothing.

My guess is that the college does not want to allow students to dump their mostly-garbage on the neighbors.

At the same time, wealthy students who will throw out items of value may feel they have better things to do than separate out items to donate. That’s especially true if they wait until the last minute.

Overall, separating items to donate from one’s trash is a way to avoid paying to dispose of trash at the dump and a way to perhaps be charitable or environmentally conscious. I’m not convinced it’s all that effective as an altruistic move. If the college sets up a place to dump for free, I’m not at all surprised many people take advantage of it without another thought.

> I expect the system will only catch a small fraction of the cheating that occurs.

The main kind of cheating we need them to prevent is effective cheating - the kind that can meaningfully improve the cheater's score.

Requiring cheaters to put their belongings in a locker, using proctor-provided resources, and being monitored in a proctor-provided room puts substantial limits on effective cheating. That's pretty much the minimum that any proctor does.

It may not stop 100% of effective cheating 100% of the time, but it would make a tremendous impact in eliminating LLM-based cheating.

If you're worried about corrupt proctors, that's another matter. National brands that are both self- and externally-policed and depend on a good reputation to drive business from universities would help.

With this system, I expect that it would not take much to avoid almost all the important cheating that now occurs.

You can try to pull them out, but they’ll hop back in. At least in my family, having been severely abused as a child appears to have been the major risk factor for growing up into a conspiratorial adult. And no wonder.

However, asking questions is genuinely a good way to interact with these types of people when they are making a scene. It’s not to try and change their mind, though. It’s to restore an air of calm and decency that gives them a chance to reconnect socially with others in the room. But you can’t be a pushover, either, and it’s OK to just walk away, ignore it, etc.

The one thing you shouldn’t do, IMO, is try to change their mind. Your need to do that will be palpable and will fuel the fire.

If a kid achieves a great 8th grade test score at age 18, is that a success or a failure of the system?

What we care about is the level of achievement by a given age. To determine that, we need to be comparing states using standardized tests given to age groups, not grade levels. It is fine to hold students back, if we think that will do them more good than advancing them. But they still need to be tested the same way as their age group if we want to do a meaningful comparison between states.

Doctors were using Google to diagnose patients before. The thing is, it's still the doctor delivering the diagnosis, the doctor writing the prescription, and the doctor billing insurance. Unless and until patients or hospitals are willing and legally able to use ChatGPT as a replacement for a doctor (unwise), ChatGPT is not about to eat any doctor's lunch.

> As a result, there is no risk to any sexual or urinary functionalities.

You can say "based on [X], we believe the risk of [Y] is insignificant." Saying there is "no risk" is far too sweeping a claim given the level of evidence it sounds like you've gathered.