mattlutze

> For the record, I agree with you – I would have thought that an AI that can reason well would probably know when not to trust humans, but I suppose that assumes it values preventing humans creating napalm over being correct and helpful.

Do we want LLMs, and later other multi-modal / servo systems, that are deciding they can't trust a human prompter and taking actions based on that?

>... and that we must find a way to build an LLM that is more distrusting and deceptive if we wish to align it with our values and our nature.

Tongue in cheek or actual argument here?

There's also the jobs where folks thing they're an employee but are being reported by the business as a contractor, to effectively pay less and also save on required tax and benefit compensation.

" These lecture notes, at least several of them, made their first appearance on the web in 2006."

Looks like it's a broad and deep foundational course. I'd imagine really appreciating this theory priming, and then taking on more domain-specific courses.

Nice to have all of these materials available!

I think these are saying the same thing (and software translators often get things slightly different than original intent).

Counterpoint: if something pretty much works without a law being in place, why write a law?

Laws are never perfect, and writing one just for the sake of it both makes complying with the law more complex and opens opportunities for people with money to create legal moats around interpretations and case law precedence.

I very much support workers collecting in solidarity to support each other.

I'm not sure that I like the idea of official agents of a state office, or de facto state office, being authorized to independently choose how to execute their duties.

Would it also be permitted for a group of socially conservative postal workers to refuse to deliver wedding licenses or legal documents to same sex partnerships? To determine which mail had documents for refugees and refuse to deliver those?

Dock workers not offloading boats is one thing, but disrupting the mail is a real challenge for me to get right in my head.

It's a bit shocking, as Sweden is one of Tesla's most mature and highest-density markets in Europe.

> but the reaction of many to being recorded was very, very negative.

Rightly so.

The lives of others aren't just data for us to unilaterally collect, to the contrary of the "beg forgiveness" big data collection philosophy of the last few decades.

"Never before in my life had I seen so many smart people working on the same goal"

I'm not sure why but the assumptions and naivety in this opening line bothers me. There are plenty of goals and problems that orders of magnitude more people are working on today.

We just don't have the information to make judgements, much less leaping to "they must be lying."

There's a few public numbers from a handful of foundation models as to performance vs parameter count vs architecture generation. Not being able to compare in detail the architecture of the various closed models nor being more rigorous on training with progressively sized parameter sets, the conclusion at the moment is a general feeling or conjecture.

> Basically, you don't want to lose evidence if there's a warrant or ongoing lawsuit. I really see no problem with that clause in most ToS documents.

Perhaps more nit-pickinlgy specific, they may be compelled by law (the courts or an agency with enforcement capacity) to maintain evidence if there's a warrant or ongoing lawsuit.

ISPs are also in the business of analytics [1, 2], and a significant percentage of customers hiding their traffic reduces the value of their analytic products.

1: https://www.bleepingcomputer.com/news/security/ftc-isps-coll... 2: https://surfshark.com/blog/isp-selling-data

>> So I have a weird catalog in my head of things not to mention if I don’t want to start the same discussion about that thing for the 50th time.

> Somehow we've built a culture where its socially acceptable, and even rewarded, to behave online in ways that would never be acceptable in other social settings. The exact flavors of that on HN, Slashdot, Twitter, Tumblr, etc are different, but the core issue is the same everywhere.

People tend to enjoy bikeshedding when presented the opportunity to, and it's a good idea to avoid bringing up the bikesheds you don't want to debate.

There is no false choice here. Only a choice.

Apple creates and sells a catalog of products. They choose to tightly couple their hardware to specific operating systems and software.

Lots of others have tried to copy Apple's hardware approach, but leave their hardware uncoupled. Some have done a great job, and in some generations surpassed the quality of the macbooks etc.

Similarly, some of Samsung, OnePlus, and others' phones are great bits of hardware and are open. Some have better cameras, better overall specs.

There's plenty of choice, and no reason for Apple to have to open up and support the complexity cost of any operating system on their platform, just because some of us passionate tech folks want Apple's specific design language in our pocket but running idk Arch.

Most software devs would prefer a debugger, but most non-software devs would probably prefer a virtual developer.

If a technical product manager was able to generate the code needed for a feature using some combination of LLMs and CoPilots—even for proof of concepts, for example—the business would need to hire fewer developers and saves significant operating costs.

Extending it to your everyday person. Aunt Jane bought a new washing machine and wants to add it into her smart house. Either she's gotta call the niece to hook it all up, write the integration to Home Assistant, write automations for her iOS devices, etc, or, she asks her favorite LLM / Co-Pilot for help and it step-by-step walks her through it directly, with code, with individual troubleshooting.

There's business in both use cases, but the total addressable market is quite a bit bigger for one of them.

I agree with the tongue-in-cheek and the serious question. I was asking myself this just yesterday.

Britanica's definition is nice (the application of science to the optimum conversion of the resources of nature to the uses of humankind), but the base of all of them is creating from science.

I'm not sure what the researched-and-validated science is behind these prompting guides. They feel like creative works and really feel like they deserve a title highlighting this.

In a related thought, I feel like we might see creative titles growing in respect over the coming years, as humans look for ways to differentiate themselves from knowledgeable machines.

It's generally better for your finances to not pay 100% of a committment today if interest on payments is below the immediate return of how you'd otherwise use the money.

The way we use a platform over time changes, as does options in the market and the way a business treats that platform. If we can export and keep our IP that was developed through the platform, it's not a bad thing to have the option to cut and run without having sunk purchase costs.

It also provides a signal for the business's product team to use with finance to argue for how to continue updating the product. Ongoing engagement metrics and new conversions are continually reviewed. How those change with the release of, or marketing of, different features directly drives the development of the platform—it should in the best case see a more useful product in our hands over time.

Many communities' implicit social contracts imply that it can be impolite to just leave in the middle of a party a person has been invited to.

If it's "I'm bored" but not "oh God, I hate this," a lot of folks will stick around to be kind to the party host.

Incarceration by itself is poorly correlated to the goal of reducing overall crime and reducing recidivism. Putting people in prison should be a last resort, instead of the first resort that the US and some other countries treat it as.

Regarding jail time for the wealthy, you're still thinking in simple numbers instead of means-tested comparisons. A more-wealthy person and a less-wealthy person that go to jail for 30 days both are physically restricted for 8% of a year. If you pretend that wealthy people earn the same way that less-wealthy people do (they don't), then both are naievely losing 8% of their income-earning time for that year.

It's still not fair, mind you—the wealthy person likely makes much or most of their money from diversified income streams that are less succeptable to temporary incarceration than the more-likely single salary-based income of a less-wealthy person.

Regarding the question of using children as leverage, this is addressed more intricately than I'll summarize here in contemporary cases e.g. with Elizabeth Holmes and the allegation that she, her partner and legal team have tried to lean on her having children as a way to attempt avoiding incarceration.

Yes.

Simple penalties create classes of crime; wealthy folks just ignore the penalty on double parking some poor schmuck because who cares about a $100 fine when you're making $1M?

These regulations are written to discourage unfair acts or behavior that exploites the social contract. It tells folks that the municipality thinks it's right for less-wealthy people to be burdened with providing comforts for more-wealthy people.

Means-tested fines actually get to the point of it: dissuade all people from doing something dangerous, indefensibly rude, et cetera, and not just poor folks.

For some additional discussion: https://www.theatlantic.com/business/archive/2015/03/finland...

Parent is talking about a fundamental feature of networks. A denser and larger network has much more useful network-related features, and if one company has a significant majority of the total addressable market for a network, it's a massive ask for people to extricate themselves and rebuild a network somewhere else.

It's why Facebook is still on top even though everyone hated it for a while; YouTube is the *only video platform, etc.

I'm not particularly literate in this, but I've understood a significant portion of the assets are very low yield government bonds.

No bank / institution today probalby wants older bonds that yield e.g. 1% if bonds issued today yield 3.5% and the Fed interest rate is 4.75-5.00%.

So there just may not be a market for them, unless you discount them enough to make the purchase price yield profit in the current inflation environment. At which point, you're selling 20B worth of bonds for e.g. 15B.

This was thrilling.

Sometimes, one person's bug is another person's feature :)

LLMs have a very clear issue of making false predictions, versus knowledge graphs that represent the information mapped within them.

We should be careful to not treat the current hot thing as a hammer.

> Anytime you have an ML system where humans are designing how the information is organized (feature engineering, linking, graph building), it will work poorly.

This is a very broad assertion that I think is not accurate.

> The algorithm needs full control over the information at the lowest level (characters, words). Only then can something that truly works be built.

But data preparation and management is fundamentally important which means there's a human shaping the information at the lowest levels. There's no single ring to rule them all. This is also just fundamentally erasing the utility of any supervised model or hybrid system which is a bit silly.

> What would you do if you were a board member of multiple companies and you found out their bank was financially unsound and couldn't do normal bank stuff?

Founders Fund has investments in some of the largest internet age companies that exist.

Airbnb, DeepMind, Lyft, Facebook, Flexport, Palantir Technologies, SpaceX, Spotify, Stripe, Wish, Neuralink, Nubank, Twilio.

A call for them to move their money out of SVB meant putting pressure out on a massive percentage of SVB's customers by value.

It's not clear at all from news reports why Founders Fund put the call out. Fast Comapny's article notes that SVB had north of 32B in net assets before the run that Thiel's fund started.

IF FF had the facts wrong, this collapse would demonstrate the effect of a few poeple's outside influence in SV, and should be a warning to the broader industry of the power that these investment vehicles are wielding.

A tool to get you to do your browsing and shopping on Amazon, so that they can improve your buyer profile and sell it across their network.

The chess example may though count as "shallow" thinking, in that your quick games are pattern matching out of experience without much time for deep or self-reflective novel analysis.

There is an issue with how people are personifying ChatGPT and assigning it agency.

Some want to talk of these LLMs as approximating an intelligent actor. If that's the case, then we also need to assign metaphors for things like deceit and coercion. We also need to consider assignments of novelty to what's generated and think of their rights as quasi-sentient, etc.

Some want to talk about them as probabilistic text token generators, which brings the benefit of not being intelligent or independent actors at all really but also then comes with the issue of intellectual property theft in training them on information not licenced for reproduction or commercial use.

The industry prefers to thread the needle between these as the former case brings some pretty wild conversations and the latter may mean lawsuits.

That's the trouble though isn't it—leaving it to mature gracefully in the cellar gets you to a point where only an expert somilier in that particular varietal knows really what to do with it, and you pay through the nose for their knowledge.

I struggle with chasing after the MacOS appearance.

It feels in passing like a lot of work to build a skin/UI that will always be a few generations of the design language behind and not quite there, just given the volume of staff at Apple and them being the ones defining where their design language goes.

Like a choice between a Porsche and, according to Google a Zotye -- they'll look close but one will be the obvious "genuine article."

Would it not be better to take all that talent and consider a fresh approach to the user experience of a personal computer?