dontreact

There were many movements planned in the 1970s, but the failure of the Soviet system seems like the key driving force in the rise of this one to me.

China is run by the Chinese communist party. You can’t own a house there (only a 99 year lease) and their notion of private property is very different. It’s not that simple.

I have many issues with the CCP but as long as large scale war is avoided, this will be the silver lining in its rise. Hopefully there are clearer successes of some socialist aspects our society can learn from.

I'm probably not going to convince you but I just read through the most recent ones and they are not that surprising. To me I would summarize the story as "there was debate inside the company about whether or not to ban Trump", which is a an unsurprising non-headline. In the files there are interesting arguments on both sides but ultimately there is (to me, and I'm guessing maybe not to you?) a clear argument as to why he ultimately got banned after some internal back and forth. He incited violence with his tweets.

This is starting to feel like Motte and Bailey fallacy to me https://en.wikipedia.org/wiki/Motte-and-bailey_fallacy

Either that or just confusion about what and/or who we are talking about.

EA is not a well defined set of concepts or ideas. I do know a lot of people who identify as EA and have argued for "earn-to-give" where the giving part is deferred far into the future.

I won't spur the confusion further but my own personal experience is that there are several people I have personally met who more closely match the parents description than the description you give.

I'm all for earn-to-give if you are doing the giving soon and to causes where the outcomes are easier to measure, like what you mentioned. This is very commendable.

Where EA-style thinking runs into trouble is where it places excess confidence in future predictions

either A) I will eventually donate this money B) This cause I'm donating to has no obvious impact now, but it will have immense impact in the future or C) Even though I can't see all the negative impacts my work is having, my prediction/inference is that the negative impacts will be smaller than the positive impacts of my altruism.

So I have nothing against the idea of trying to maximize good, I just think that I frequently see it used in conjunction with excessively high-confidence predictions or inferences.

In your case, I think it's important that the banker do some sort of accounting of whether he is having a negative impact on their current project/firm etc. There are bankers who have had a huge negative impact in the past and so this does need to be part of the accounting.

There’s a lot of data without copyright out there

There have been repeated reports of negative outcomes as a result of TikTok trends. There have been investigations into the amount of information the app collects and it is far beyond what other apps do (all without user consent). The app is clearly highly addictive.

What sorts of proof would convince you that this is an addictive app that is causing harm?

What about if storage improves? Can't that smooth out the intermittent nature and provide for the base load?

I see your overall point and I think I agree I just think it’s a question of degree rather than a binary expansionist/not. For example, didn’t Native American tribes expand to America from Asia?

It seems like almost no matter what the AI folks come up with, there will be some potential way of interpreting neural phenomena as loosely mapping onto it.

Sweeping generalizations about AI models as a class of objects seem deeply uninformative to me. Perhaps I would even go so far as to say misleading. These days any undergrad can go on GitHub and have a model that does some diagnosis and I don’t understand why we have to group that together with multi-year efforts from groups of PhD’s and MDs working together to produce products.

There’s obviously way more of the first than the second and so if you analyze this group as a whole it’s easy to draw the wrong conclusion about what AI as a technique is capable of.

I can’t give too many details on specific examples I’m working on at Google but an example I’m not working on would be Caption Health who have an amazing AI-based ultrasound guidance product that has great prospective evidence, and several big fans in the clinical community.

There are also several success stories using AI on pathology in order to target clinical trials.

Can you imagine if someone made sweeping statements about webpages as if they were a coherent group of objects that you could sample from and deduce properties? “The information on websites is typically not as accurate as those websites claim”

In addition to the ubiquity of monopolies and negative externalities mentioned, public goods are also a way in which free markets don’t properly reward people’s contributions to society. The classic textbook example is a public park, whose price in the market will typically be far under the value it creates for the community it is in.

And as far as safety scores, my point is that the safety score is calculated incorrectly because of obvious false positives with "close following". I'm talking being nowhere near a car, getting an alert that says I'm following too close, and that dropping my safety score. I understand why the bar is high, but at this point I honestly suspect there is some tomfoolery going on with how that score is calculated.

TPUs are hard to use outside of Google, (I have tried in and out of Google). I think the situation is improving, but the efficiency from using a large pod is really remarkable. What topology did you train AI models on? Within Google it's common to train across a whole pod or even across multiple pods 8x16x16 is the largest currently.

Also if Tesla actually published numbers on an MlPerf benchmark, I would be more inclined to believe claims about 36x better efficiency.

https://mlcommons.org/en/training-normal-20/

The fastest times I'm seeing here for image classification and for object detection (not the same, but probably closest proxy out of the tasks benchmarked) are for TPUs.

To know who has better training technology I don't think you should be using a cost-efficiency metric, it seems to me the best thing to use would be who can train networks the fastest. Cost metrics are easy to game especially if you are the ones making the chips (Of course them making chips is cheaper than buying Nvidia chips for them once the capital investment is made). To measure who is ahead in technology, I think you have to look at who can train models the fastest, and right now as far as I can tell, TPUs are unbeat for this. (Although practically speaking it's hard to pull off these large topology things externally and there are also other caveats with ML perf related to how the training setups are optimized, but nonetheless, it's a better signal than what Elon says in a presentation :) )

I'd sum up your points 1,2,3 as "more data". This would be a reason to think they can one day be ahead if they can take advantage of this, but not evidence that they are currently ahead.

Occupancy networks: waymo has published research on this before Tesla announced this at AI day (not clear to me who got there first though https://arxiv.org/pdf/2203.03875v1.pdf)

Tesla's Dojo -> Waymo has TPUs to train on

To me all of this is outweighed by the fact that Waymo has a driverless deployment and Tesla does not. I am pretty biased because as a Tesla owner I am pretty pissed off at this point at how the false positives on the system in detecting close following are stopping my safety score from getting high enough to even be able to access the product I purchased.

But it is pretty hard to say one way or another.

I think stockfish with limited depth to compute every move in under .1 second would still be incredibly stronger than a human to the point where their involvement would produce worse moves.

I agree. My read on this set of texts is Musk trying to get Dorsey on his side to help with the acquisition.

You don't have to view your work as a source of fun or fulfillment to be enjoyable to work with. And I also don't think you need to enjoy anything to complete tasks well, but if you don't take any joy at all in anything you work on, it's probably not very fun to work with you.

Hey as long as you enjoy some aspect of work that is fine. If you find things you enjoy more outside of work and derive true meaning from things outside of work, then that's amazing. All I was saying was _IF_ there is absolutely nothing you enjoy about work, then I imagine it would be quite difficult to enjoy working with you.

I don’t know what the percentages are. All I know is there are many people who can enjoy some aspect of work, and I would much rather work with them than with someone who can’t enjoy anything about the work.

People who are completely apathetic and find no enjoyment in any aspect of work probably are also best off working together.

I’m sorry if you feel that way. If there is absolutely nothing you can find that is fun about your job then honestly yeah I would rather not have you on my team because it sounds like it would be a pretty bad time working together.

I don’t want to work in a place where any budget cuts result in people saying there is no fun to be had anymore. That’s a pretty negative environment. It seems like a reasonable statement to say, we should try and find ways for the work itself to be fun, and not only expect expensed travel and offsite to be the only source of fun.

Huh I was at this meeting and it didn’t seem heated.

To me it felt like a CEO explaining difficult choices and the reasoning behind them.

Why is vector generation in the browser untenable with ML approaches? There are tiny models and tensorflow.js

There isn't really always a clean separation between private and state actors when it comes to housing (i.e. Fannie Mae has appraisers https://singlefamily.fanniemae.com/learning-center/applicati...)

So I think we are probably aligned that Government Sponsored Enterprises should be spending some time looking into this given the (weaker) evidence that's already there.

If systemic issues with appraisal are found there, it seems fair to me that there should be some regulation and/or state-sponsored investigation of whether the same systemic issues are happening with purely private appraisal.

I think you’ve totally ignored

A) the historical evidence about discrimination at every level in the housing system

B) the study from the brookings institute showing bias with housing appraisal

I’m also not claiming all of the burden of proof should lie with the accused.

What I am saying is given the anecdotes, lawsuits, and retrospective analysis, I think it’s fair to put more pressure on the system to help with investigating the issue more systematically.

Getting the type of statistical evidence you are talking about is expensive and I think there are enough indications that 0% of the cost seems like too low of a cost to expect a huge industry to cover.

It would be nice if more of the burden of proof were on subjective systems that historically have been full of racial bias, rather than on the communities that have historically suffered from the effects of the discrimination.

This is not an isolated incident, there are dozens of lawsuits and the study from the Brookings institute linked elsewhere in this thread.

This will now be my favorite example of anyone who says people are not affected by racism in the US anymore.

Maybe I'm misunderstanding you but it seems like you are arguing for a fairly hard-line stance of: Facebook shouldn't be allowed to censor or take down any content, and all decisions to do this should be pushed down to individual users or group moderators.

This is certainly unworkable given the level of bots, spam, etc. that would be untenable for a volunteer moderator on a Facebook group, or an individual user, to deal with.

Maybe I am misunderstanding you and your point is more subtle: the government should have some rules around what Facebook can and cannot censor. Then I don't disagree with you in principle, I think we would have to get into specific proposals and weigh their pros and cons.

A blunt approach of "free speech, no censorship" isn't a convincing argument, but given the scale of impact that site-wide or nation-wide moderation decisions have on Facebook, I could be convinced that government oversight in certain cases makes sense.