hyeonwho22

There was a fun recent variant on this game using LLMs, asking GPT3 (3.5?) to encode text in a way that it will be able to decode the meaning. Some of the encodings are insane:

https://www.piratewires.com/p/compression-prompts-gpt-hidden...

That is a very poor regulation. Why enforce wheel lift? What matters is that the chair doesn't tip over - that the center of gravity remains in the center of the four wheels.

Hindawi is the worst. They solicit book chapters, pretend to be peer reviewed, but any peer review is cursory. You can tell because they have published books on homeopathy and on the theory of Chinese traditional medicine (qi).

Is that a serious argument against an AI pause? There are potential scenarios in which regulating AI is challenging, so it isn't worth doing? Why don't we stop regulating nuclear material while we're at it?

In my mind the existential risks make regulation of large training runs worth it. Should distributed training runs become an issue we can figure out a way to inspect them, too.

To respond to the specific htpothetical, if that scenario happens it will presumably be by either a botnet, by a large group of wealthy hobbyists, or by a corporation or a nation state intent on circumventing the pause. Botnets have been dismantled before, and large groups of wealthy hoobyists tend to interested in self preservation (at least more so than individuals). Corporate and state actors defecting on international treaties can be penalized via standard mechanisms.

1. The religious nut doesn't have the knowledge or the skill sets right now, but AI might enable them.

2. Accessibility of information makes a huge difference. Prior to 2020 people rarely stole Kias or catalytic converters. When knowledge of how to do this (and for catalytic converters, knowledge of their resale value) became available (i.e. trending on Tiktok), then thefts became frequent. The only barrier which disappeared from 2019 to 2021 was that the information became very easily accessible.

Your last two questions are not counterarguments, since AIs are already outperforming the median biology student, and obviously removing sites from the internet is not feasible. Easier to stop foundation model development than to censor the internet.

> What is to stop someone from training a model on such data anytime they want?

Present proposals are to limit GPU access and compute for training runs. Data centers are kind of like nuclear enrichment facilities in that they are hard to hide, require large numbers of dual-use components that are possible to regulate (centrifuges vs. GPUs), and they have large power requirements which make them show up on aerial imaging.

The only thing keeping bioweapons from being easy to make is information becoming easily available.

> Essentially you are advocating against information being more efficiently available.

Yes. Some kinds of information should be kept obscure, even if it is theoretically possible for an intelligent individual with access to the world's scientific literature to rediscover them. The really obvious case for this is in regards to the proliferation of WMDs.

For nuclear weapons information is not the barrier to manufacture: we can regulate and track uranium, and enrichment is thought to require industrial scale processes. But the precursors for biological weapons are unregulated and widely available, so we need to gatekeep the relevant skills and knowledge.

I'm sure you will agree with me that if access information on how to make a WMD becomes even a few order of magnitudes as accessible as information on how to steal a Kia or how to steal a catalytic converter, then we will have lost.

My argument is that a truly intelligent AI without safeguards or ethics would make bioweapons accessible to the public, and we would be fucked.

Why not? It has already been shown that AI can be (mis)used to identify good candidates for chemical weapons. [1] Next in the pipeline is obviously some religious nut (who would not otherwise have the capability) using it to design a virus which doesn't set off alarms at the gene synthesis / custom construct companies, and then learning to transfect it.

More banally, state actors can already use open source models to efficiently create misinformation. It took what, 60,000 votes to swing the US election in 2016? Imagine what astroturfing can be done with 100x the labor thanks to LLMs.

[1] dx.doi.org/10.1038/s42256-022-00465-9

You have much more patience than I. After the second email type I deleted my account.

When dating in Korea women would guess my height to 2 cm higher than I am tall. Which was very eerie, because I correct my gait with 1.5 cm high insoles. I would estimate most women there have an accuracy of +/- 1 cm, and the worst were +/- 2.5 cm.

The problem in Korea at least isn't with meeting people (well, for 80% of the population), it is young couples feeling they can't meet the finances required to satisfy social norms, and thus deciding to delay the marriage-kids sequence indefinitely.

Several of my coworkers had been dating for 5+ years, but they were only making $50k annually (early career engineers). The socially expected family-sized condo costs $500k to $1M, and the young couple is expected to buy and furnish it before their wedding.

I got a CR subscription to look at air filters. For good air filtering you really only need a hepa filter ($20) in front of a fan ($20), but CR only analyses the performance of brand name products ($200).

> Are DMCA claims under oath?

Yes. DMCA claims are made under penalty of perjury.

> Being prevented from publishing a video someone else created, say, doesn't inhibit your right to express any opinion (aka freedom of speech).

Indeed, the inhibition of free speech is when you are prevented from publishing a video you created, because someone else falsely claims that they created it. Hence the felony.

The real answer is that the US doesn't want public transit. I visited a rustbelt town several years ago, and a new mall had been sited specifically so that it would not be on the bus network: being accessible to the people who ride the bus was considered a liability for the businesses that would rent space in the mall.

More recently I visited a much nicer midwestern town which is planning an expansion of its bus network amd optimizing traffic lights for better bus flow. There is a new mall on the bus line. The difference is that 95% of bus riders in this second town are upper middle class college students.

Under the Sherman Act, the criminal penalty for conspiracy to monopolize is up to 10 years in prison or $10M for individuals, and up to $100M for a business, OR up to twice the monetary benefits to conspirators OR up to twice the money lost by victims, if those amounts are over $100M.

DMCA claims are made under penalty of perjury. Perjury is usually a felony.

False DMCA claims are perjury, which is (at least in some states) a felony.

The rationale is that the DMCA gives complainants the ability to restrict others' speech, and so the law wants to strongly disincentivize abuse of that power.

The main reason people continue using it is due to the network effect (xlsx files are considered shareable) and the fact that people are trained on its features and interface. That does make it valuable, but that doesn't mean it is more usable than competitors.

There are some Islamic traditions which have created art of Muhammad. Lots of examples at https://en.m.wikipedia.org/wiki/Depictions_of_Muhammad .

This is a nitpick, but likenesses of cartoon characters are protected by trademark rather than copyright.

Their 'offering' is totally artificial. Prices are low because suppiers aren't permitted to offer lower prices elsewhere. They have a history of blatantly abusing their monopoly position. The law even includes personal criminal liabilities for CEOs in these cases.

Offering free shipping is not a problem. The problem is that mandating your suppliers subsidize free shipping and don't sell for cheaper prices elsewhere is blatantly anticompetitive, and has the effect of raising consumer prices on non-Amazon markets.

On the contrary it is very testable: look at number of grandchildren in families with different moral beliefs and cultural norms. Can look at whether the grandchildren thrive, too.

I would argue that rather than stability a lot of the puritan instinct comes from a desire to see one's children thrive.

The relevant analytical unit at the small scale is the family: I don't want my kids to be temperant because of stability, I want them to abstain from drugs/games/$VICE because that's the path which maximizes the chance of their living a fulfilling life, or (more cynically) which maximizes their chance of bearing me successful grandkids and great grandkids. This is why puritainism is selected for evolutionarily (at least in environments where resources are limited).

To return to the large scale policy questions, I also don't want to see the continent of my children fall to a mercantilist China (using China as an example because Chinese law cracks down hard on drug sales and limits students to one hour of video games per night). Accordingly, I support policies to limit access to addictive substances and stimuli, despite the inevitable conflict between those laws and individual rights. The inequitable enforcement of those laws is another problem entirely, and one which I think would be well solved by starting with the prosecution of celebrities and thought leaders who openly partake in $ADDICTIVE_STIMULUS, and their suppliers.

I've gotten a few phone calls like that. A few years ago living abroad it seemed to be a targeted sales pitch for one gender (ex. makeup). Would hang up as soon as I answered, or pretend that they needed the number of my wife (I wasn't married).

More recently in the US I got a call where the other side didn't say anything and hung up after exactly one minute. Suspicious indeed.

As a slight counterpoint to this, there is a lot of implicit knowledge and group cohesion in most workplaces that immigrants can have a very hard time picking up on and integrating into. When I moved to another country I thought I fit in well with my coworkers, but it turned out that my style of conflict resolution was absolutely incompatible with theirs, and I had to relearn how to communicate with my boss, despite the boss being fluent in English. This cultural misfit set me up to be manipulated into the position of group enemy by one of the more socially savvy and manipulative coworkers.

More generally, the challenges in communication are a well known barrier to outsourcing knowledge work.

I think advertising needs to be broken into several competing companies. Otherwise it is just spinning off the most profitable part of the company as its own standalone monopoly, and leaving the services users find useful as businesses without revenue streams.

In AI research ablation experiments are comparative tests, but in one branch some feature is removed. Here the Chrome team is saying that removal of the feature (which the Ads team wanted) was instantly noticed by users.