What if people are required to try a tiny bit of each before deciding which one they want a full helping of? In that setup I'm still extremely confident the ice cream wins - it would be crazy imo for anyone to update meaningfully due to this survey.

Of course, as we dive more into the nitty gritty of the setup it becomes relevant how the product/stunt is positioned etc. - but the fact that there are all these considerations that need to be made serves to further reinforce the point that it's hard to predict human preferences!

Ironically - it's hard know whether to take your survey results at face value given the complexity of human motivations (if it wasn't for the meme, would people's reaction be different?).

If we ran an experiement where there were 2 stalls set up in public (ideally not in SF, thus almost nobody is in on the joke), and each offers free food - one serving the best ice cream available and one serving the bear fat + salt + honey mixture - which do you think will draw longer lines?

(If anyone is persuaded by this post that the bear fat would win and is willing to wager, perhaps we can run it for real!)

I think an important consideration being overlooked is how comptetntly a centralised project would actually be managed.

In one of your charts, you suggest worlds where there is a single project will make progress faster due to "speedup from compute almagamation". This is not necessarily true. It's very possible that different teams would be able to make progress at very different rates even if both given identical compute resources.

At a boots-on-the-ground level, the speed of progress an AI project makes will be influenced by thosands of tiny decisions about how to:
 

• Manage people
• Collect training data
• Prioritize research direcitons
• Debug training runs
• Decide who to hire
• Assess people's perfomance and decide to should be promoted to more influential positions
• Manage code quality/technical debt
• Design+run evals
• Transfer knowledge between teams
• Retain key personnel
• Document findings
• Decide what internal tools to use/build
• Handle data pipeline bottlenecks
• Coordinate between engineers/researchers/infrastructure teams
• Make sure operations run smoothly
   

The list goes on!

Even seemingly minor decisions like coding standards, meeting structures and reporting processes might compound over time to create massive differences in research velocity. A poorly run organization with 10x the budget might make substantially less progress than a well-run one.

If there was only one major AI project underway it would probably be managed less well than the overall best-run project selected from a diverse set of competing companies.

Unlike the Manhattan project - there's already sufficently strong commercial incentives for private companies to focus on the problem, it's not already clear exactly how the first AGI system will work, and capital markets today are more mature and capable of funding projects at much larger scales. My gut feeling is if AI was fully consolidated tomorrow - this is more likely to slow things down than speed them up.

Yes that's true that beginning almost any form of exercise will deliver most of the benefits compared to what an optimal rountine would! But your post is all about trying to be as time efficient as possible (e.g you also discussed the potential to use drop sets to go even beyond failure!).

For vast majority of people reading this post - if their goal is to get the greatest possible benefit from resistance training in shortest amount of time - the biggest mistake they're making right now is not having their sets be difficult enough.  You're right that you don't need to go to failure to get most of the benefits, but if time efficiency is the goal, spending that extra 15 seconds to add those two final pre-failure reps to a set is the first thing I'd reccomend.

(This comment is directed more at the rest of the audience who I think are likely to take the wrong lessons from this post, rather than OP who I don't doubt knows what they are talking about and has found success following their own advice)

[For supersets] the main downside being that the overall greater work packed in to a smaller amount of time being somewhat more motivationally taxing.

At least for me personally, this is an overwhelmingly important consideration! There's no way in hell I could go to failure on both ends of a superset at once without throwing up.

 (I just go 1 or two reps shy of failure and don't worry too much about perfect weight selection to hit the same rep ranges all the time). Studies have found even 2 sets to failure once a week to be sufficient for maintenance if you are super strapped for time. I normally do this twice a week.

This might work for OP, but I don't think it's good advice for most readers (80% of whom I'd bet have never actually pushed a weight lifting set to failure and will be mistaken about where their failure point actually is).

If you aren't tracking your lifts exactly, reccording them each time, and forcing yourself to do more than the previous time every time you return to the gym, it's very hard to know for sure that you were actually 2 reps shy of failure.

We can experience a lot of discomfort/panic before the point at which we actually fail - (for most weight exercises if you're not involuntarily making loud yelling noises during the last couple reps, you're probably not anywhere close to true failure).

If you're reading this, and you don't already have a book/app where you write down exactly how much weight you lifted and are progressively increasing that, do this first!

 

I hear you that teenagers spending hours computing hundreds of analytic derivatives or writing a bunch of essays is a pretty sad time waste... But if incentives shifted so instead that time got spent perfecting a starcraft build order against the AI a hundred times or grinding for years to pull off carpetless star in SM64, this might be one of the few ways to make that time spent even more pointless... (And for most people the latter is no more fun than the former)

You might be right that the concept only applies to specific subcultures (in my case, educated relatively well-off Australians).

Maybe another test could be - can you think of someone you've met in the past who a critic might describe as "rude/loud/obnoxious" but despite this, they seem to draw in lots of friends and you have a lot of fun whenever you hang out with them?

Maybe an analogy which seems closer to the "real world" situation - let's say you and someone like Sam Altman both tried to start new companies. How much more time and starting capital do you think you'd need to have a better shot of success than him?

Out of interest - if you had total control over OpenAI - what would you want them to do?

I think OP is correct about cultural learning being the most important factor in explaining the large difference in intelligence between homo sapiens and other animals.

In early chapters of Secrets of Our Success, the book examines studies comparing performance of young humans and young chimps on various congnitive tasks. The book argues that across a broad array of cognitive tests, 4 year old humans do not perform singificantly better than 4 year old chimps on average, except in cases where the task can be solved by immitating others (human children crushed the chimps when this was the case).

The book makes a very compelling argument that our species is uniquely prone to immitating others (even in the absense of causal models about why the behaviour we're immitating is useful), and even very young humnans have inate instincts for picking up on signals of prestige/compotence in others and preferentially immitating those high prestige poeple. Imo the arguments put forward in this book make cultral learning look like a very strong theory better in comparison to Machieavellian intelligence hypothesis,  (although what actually happend at a lower level abstraction probably includes aspects of both).