Cédric

Karma: 52

Cédric 6 Oct 2023 7:37 UTC
1 point
−26
on: Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
This is alignment’s Attention Is All You Need moment

Cédric 21 Nov 2022 4:58 UTC
2 points
0
on: On Morality, Ethics, and all that Jazz
You could get your framework by adapting existing frameworks to fit your meta-agent utility function. Examples:
1. The utilitarianism framework which seeks to maximize the sum of utility over all agents.
2. The Rawlsian maximin framework which seeks to maximize the utility of the worst-off agent.
3. The Nozickian entitlement framework which seeks to give each agent the maximal entitlement they could have, given the constraints of the system.
4. The Nussbaumian capability approach which seeks to give each agent the maximal capability they could have, given the constraints of the system.
I think in the end you would get stuck on the unsolved problem of balancing the needs of individuals and the collective.

Cédric 19 Aug 2022 1:47 UTC
9 points
8
in reply to: DirectedEvolution’s comment on: The Parable of the Boy Who Cried 5% Chance of Wolf
Explaining more to align understanding.

Just translate this:

“I didn’t say there was a wolf!” cried the boy. “I was estimating the probability at low, but high enough. A false alarm is much less costly than a missed detection when it comes to dying! The expected value is good!”

Into normie language and should be fine.

However it’s very hard to communicate nuance at scale. I have no idea how to solve that.

Cédric 9 Jun 2022 4:49 UTC
7 points
in reply to: Cédric’s comment on: Theses on Sleep
Follow up:

I didn’t try the experiment for very long but here are my observations when sleeping fewer hours:
- Lower self-control (more Uber Eats, more checking socials during work)
- When I was able to exert enough control to concentrate on something, my concentration was deeper
- Felt more tense/stressed? I guess it must be due to more adrenaline and cortisol
- More creative (increased frequency of random ideas while doing chores/mundane tasks, more lost in interesting thoughts—when a thought was particularly interesting, I’d deeply and automatically follow the thoughts for multiple minutes)
So basically I felt tired, energetic, more disinhibited, creative, unable to stay concentrated but when concentrated, the concentration was deeper. It was a very oxymoronic experience with the tired + energised and contradictory effects on concentration.

Overall, pros and cons but if I can overcome the self-control problem it will be a net positive productivity wise. However, due to the stress that’s another point deducted from the healthiness of sleeping fewer hours (not to mention the other non stress related potential issues).

The main thing that this has unlocked for me is that it has freed me from the fear of missing out on sleep. I sleep my 8 hours but don’t sweat it if I don’t sleep well or if I stay up late.

I also think it’s a good idea to try random things, even if they fly in the face of conventional wisdom. Obvious caveat is you might cause irreparable damage to yourself but as long as you’re careful not to take self-experimentation too far there might be some benefit to walking off the beaten track.

Cédric 8 Jun 2022 3:35 UTC
2 points
1
in reply to: JBlack’s comment on: AGI Ruin: A List of Lethalities
Well then let’s use hyperbolic discounting to our advantage. If we make paddling sufficiently taboo, the social punishment of paddling will outweigh the rewards of potentially building AGI in the minds of the selfish researchers.

Cédric 8 Jun 2022 3:28 UTC
5 points
0
in reply to: lc’s comment on: AGI Ruin: A List of Lethalities
What I’m doing is trying to help with the wings by throwing some money at MIRI. I am also helping with the stopping/slowing of paddling by sharing my very simple reasoning about why that’s the most sensible course of action. Hopefully the simple idea will spread and have some influence.

To be honest, I am not willing to invest that much into this as I have other things I am working on (sounds so insane to type that I am not willing to invest much into preventing the doom of everyone and everything). Anyway, there are many like me who are willing to help but only if the cost is low so if you have any ideas of what people like me could do to shift the probabilities a bit, let me know.

Cédric 8 Jun 2022 2:34 UTC
8 points
2
on: AGI Ruin: A List of Lethalities
Imagine we’re all in a paddleboat paddling towards a waterfall. Inside the paddleboat is everyone but only a relatively small number of them are doing the paddling. Of those paddling, most are aware of the waterfall ahead but for reasons beyond my comprehension, decide to paddle on anyway. A smaller group of paddlers have realised their predicament and have decided to stop paddling and start building wings onto the paddleboat so that when the paddleboat inevitably hurtles off the waterfall, it might fly.

It seems to me like the most sensible course of action is to stop paddling until the wings are built and we know for sure they’re going to work. So why isn’t the main strategy definitively proving that we’re heading towards the waterfall and raising awareness until the culture has shifted enough that paddling is taboo? With this strategy, even if the paddling doesn’t stop, at least it buys time for the wings to be constructed. Trying to get people to stop paddling seems like a higher probability of success than wing building + increases the probability of success of wing building as it buys time.

I suspect that part of the reason for just focusing on the wings is the desire to reap the rewards of aligned AGI within our lifetimes. The clout of being the ones who did the final work. The immortality. The benefits that we can’t yet imagine etc etc. Maybe infinite rewards justifies infinite risk but it does not apply in this case because we can still get the infinite rewards without so much risk if we just wait until the risks are eliminated.

Cédric 17 Mar 2022 22:40 UTC
3 points
in reply to: Dagon’s comment on: I read Einstein’s biography. Here are 15 quotes that reveal his philosophy on life.
Well, I hope you’re right because I’d feel bad if someone tried to write something useful for us and it was so bad the comments are just speculation about whether the person is a spammer.

I’ll keep on assuming people are actually trying though and try to provide constructive feedback and encouragement because the demand for LW posts outstrips supply. Even if the person is a spammer, perhaps being more encouraging and constructive will make others who are hesitant to post more comfortable. And if the person is not a spammer, they can use the feedback to improve on their next post and hopefully iterate until their posts become really good.

Cédric 17 Mar 2022 10:28 UTC
1 point
in reply to: rsaarelm’s comment on: I read Einstein’s biography. Here are 15 quotes that reveal his philosophy on life.
What would be the purpose of doing such a thing? There is no link in the writeup which would indicate farming backlinks for SEO

Cédric 17 Mar 2022 3:46 UTC
2 points
on: I read Einstein’s biography. Here are 15 quotes that reveal his philosophy on life.
You’re getting downvoted without feedback so I’ll try to provide some.

The post does not provide any particular insight, it’s a disparate collection of quotes. The same and more can be gotten by googling ‘Einstein quotes’. Some ideas about how you can make the post more insightful:
- Connect the quotes to uncover Einstein’s worldview and approach to his work and life
- Give your personal thoughts about the quotes, its caveats and nuance
- Explain what Einstein means a little more as some quotes don’t really mean much without context/explanation
Also, Einstein’s brilliance was in his physics, none of those quotes really touch on that. In fact, his worldview outside of physics is not that sophisticated, especially relative to his work in physics which is as sophisticated as it gets. Some of those quotes are also quite obviously wrong e.g. the one about having no special talents is wrong even if it wasn’t self-contradictory. Perhaps focusing on his work and how he did it would provide more value as that’s what he was best at.

Cédric 3 Mar 2022 23:27 UTC
1 point
on: Recognizing and Dealing with Negative Automatic Thoughts
A failure mode for this is that when someone is faced with strong negative emotion, they are unable to think about the problem rationally. Their brain gets hijacked by negative emotion with no capacity to actually go through the correct steps.

A potential solution to that is putting things into the broadest perspective by asking yourself whether this is even a big deal at all (will it matter in a week? two weeks? a year?). Most problems don’t. But it can be hard to do even that. So you must change your entire mindset to recognise most problems as being not a big deal. You could go through past situations when faced with problems which you had a strong negative reaction to and recognise with the benefit of hindsight that they weren’t such a big deal after all. Then, when faced with a new problem, you can wedge in the broadest perspective trick before spiralling into negative emotion and that perspective will have strong prior evidence of being true which increases the likelihood of it working.

I think the broader point I am trying to make is that whatever rational technique you use for dealing with negative emotion is going to be out of reach when actually faced with strong negative emotion. So one has to, by whatever means, eliminate/attenuate the negative emotions enough to get the neocortex back online such that one can then go on to use the right techniques. The best elimination/attenuation techniques ideally would engender a mindset change such that the frequency and severity of strong negative emotions is reduced.

Cédric 16 Feb 2022 6:15 UTC
2 points
on: How do you think about mildly technical people trying to advance science and technology?
That’s pretty much me. I spent 2021 learning fullstack web dev and I’m spending 2022 bootstrapping a vertical SAAS. I’m doing this on the side while I work as a dev for not for profits (civic tech).
The long-term goal is to keep building those niche SAAS products and open source all the code while donating a yet undecided % of profits to effective charities.
I’m also learning how to write well with the goal of sharing whatever knowledge I acquire along the way through my blog.
I am also working on a moonshot solo research project in the field of complexity science. This will likely take years (decades?) and has a close to zero chance of success but if it works out it’ll be a pretty decent contribution to the field.
The general principles I’m following are:
- Work on tech products as they’re highly scalable and the market will ensure you’re solving real problems
- Maximise positive externalities (donate % profits, open source, share learnings)
- Do science on the side on something that interests you and which won’t be a waste if it works out
Getting a PhD isn’t likely to pass the opportunity cost test for impact unless you’re a genius.

What I’m doing is really low risk, I’m not taking venture capital and I’m still working at my job but that’s obviously limiting. If you really want to have a huge impact and don’t mind huge risk you could try a startup that solves a huge problem you really care about having solved.

Cédric 13 Feb 2022 11:13 UTC
3 points
on: Theses on Sleep
Unlike the other commentors, I am sold enough on the conclusion to give it a try.
I’ve been waking up at the same time every day (~95% of the time) and sleeping at roughly the same time every night (~70% of the time). What I’ll do is keep waking up at the same time but just go to bed when I’m tired instead of ~8 hours before my waking time.
I don’t have any issues with tiredness/low-energy. Ideally that would continue and I would be able to reclaim an hour or two from each day.

Cédric 13 Feb 2022 11:08 UTC
3 points
on: Theses on Sleep
I used to use a Fitbit to track my sleep. I noticed that after sleeping very few hours, I’d have more % deep and REM sleep the following night (and less % light sleep / restlessness) often enough for me to notice a correlation. I checked my friend’s Garmin sleep stats and it’s roughly the same.
I wouldn’t give too much weight to the accuracy of this observation (confounding, device accuracy etc). However it just about makes the cut to share it in a comment imo.

Cédric 9 Jan 2022 1:31 UTC
3 points
on: Cédric’s Shortform
A conversation with a friend of mine in the car about him learning React.
This conversation might be interesting to others as rationality techniques were successfully used to persuade someone to act more rationally for their benefit in an everyday life type context.
K: ‘When I get home I’m going to keep working on my portfolio website and use that to learn React’
Me: ‘How are you going about that?’
K: ‘I’m following a YouTube tutorial’
Me: ‘I think you should just learn the concepts yourself then apply them to making the website without following a tutorial. When I was following tutorials I could not make projects independently afterwards using what I had just learned in the tutorials’
K: ‘Yeah, but I already started, I’m just going to finish this...’
Me: ‘That’s the lost cause fal...’
K: ‘Yes, I know it’s a fallacy, but when I start something, I want to finish it’.
Me: ‘But your goal is to learn React, your approach is not effective. I mean, you can do whatever you want but I think you’re going about it the wrong way’
K: thinking expression for many seconds
Me: ‘It’s not just for me that the tutorial approach didn’t work, I routinely see posts about how people fail to apply anything they learn in tutorials after they complete them. Or posts about people doing tutorials for months and not being able to actually work on anything themselves’
K: More thinking
Me: ‘Whacha thinking about?’
K: ‘I’m thinking about how I best learned at uni’
Conversation then shifts to what methods of learning we found most effective. Friend has done the rational thing and moved away from the ineffective technique.
If anyone has ideas about how I could have been even more effective, I’d love to hear them!

Cédric 7 Jan 2022 10:47 UTC
14 points
on: AMA: I was unschooled for most of my childhood, then voluntarily chose to leave to go to a large public high school.
What’s the best part of unschooling that’s missing in high school and vice versa?

Cédric 7 Jan 2022 3:09 UTC
3 points
on: How To Raise Others’ Aspirations in 17 Easy Steps
These statements might spur some action in the moment but are unlikely to have lasting effects. People have a certain level of trait conscientiousness and certain behaviour/thinking patterns. I think it’ll take more to actually meaningfully push someone to aim higher and act on those higher aims.
One could probably use those consistently on a friend over time and actually see a change (provided the friend actually wants to change/is receptive).