revel

LLMs are bad at creative work and I don’t see them improving any time soon. Try asking an agent to write a story about raccoons. It will almost certainly involve either stealing food or raiding trash with a 50% chance of having a character named Pip. If a location is mentioned, it’ll be Elm Street.

I guess this is the average story and, similarly, the average game is boring and predictable

Awful news for everybody, particularly those in speculative tech jobs.

We may well look back at the attack on Iran as the American tipping point that caused the same kind of financial stagnation as Brexit caused for the UK.

Running benchmarks at scale and protecting against reward hacking is non-trivial.

What exactly do you mean by "career politician?" Being a member of the opposition in a dictatorship means giving up material wealth and putting your life & liberty on the line.

This is one cause but another is that agents are mostly trained using the same sets of problems. There are only so many open source projects that can be used for training (ie. benchmarks). There's huge oversampling for a subset of projects like pandas and nothing at all for proprietary datasets. This is a huge problem!

If you want your agent to be really good at working with dates in a functional way or know how to deal with the metric system (as examples), then you need to train on those problems, probably using RFT. The other challenge is that even if you have this problem set in testable fashion running at scale is hard. Some benchmarks have 20k+ test cases and can take well over an hour to run. If you ran each test case sequentially it would take over 2 years to complete.

Right now the only company I'm aware of that lets you do that at scale is runloop (disclaimer, I work there).

I was a big enough believer in crypto to literally start a company in this space only to leave it completely disenchanted and deeply pessimistic about the direction of the industry. I felt that there were many real legal and regulatory challenges that governments just didn't want to deal with. No government wants to enable money laundering, black markets, corruption and terrorism; or so I thought!

Now we're in a situation that's so much worse than I ever imagined -- Trump coins are vehicles for naked bribery and corruption with a sprinkle of encryption on top. I was worried about black markets, Trump has literally been using his office to grant access to top holders of his scam coin.

This is a big lesson for everyone about why some degree of regulation is necessary.

They used RFT and there's only so many benchmarks out there, so I would be very surprised if they didn't train on the tests.

> Now, imagine the uproar that might ensue from some corners today if a textbook made this sort of direct comparison between two groups, with one group happening to be white and the other black. Some social justice warriors would no doubt raise an outcry. The book would be branded as Eurocentric, racist, and white supremacist, since it doesn’t give equal space to describing the intellectual achievements of the Zulu or to recognizing the validity of indigenous ways of knowing. The picture of the African witch doctor might be described as a “culturally insensitive caricature”. The book would also be criticized for not describing ways in which Zulu society is more healthy than contemporary American society.

This textbook is from 1929 and was probably being written at the exact same time as the Snopes monkey trial (1925). You're reading science hype because science was under attack; and not from "social justice warriors" but from conservative Americans. Notice that the single most transformative discovery in all of biology, evolution, isn't mentioned.

That attack on evolution is still taking place. In Texas there was an attempt to strike all references to evolution and climate change from the curriculum, along with other attempts to introduce biblical references. When did that take place? That was this year; oh, and in 2023: https://www.wsj.com/us-news/education/texas-education-board-...

Getting mad at "social justice warriors" for an imagined reaction while ignoring the actual attacks on science is... well, it's what I've come to expect.

That last point about differences between dev, test and prod should be right at the top. It's rare to find teams that have set themselves up for success (for reasons I do not fully understand)

I didn't get that impression.

Although you're right that there's likely no relationship between wives developing heart conditions and subsequent divorce, there's not enough information from the article to know whether there's anything statistically meaningful about heart conditions specifically. It seems more likely that it's just statistical noise. I read that section and got the impression that the relationship is interesting but it doesn't necessarily mean anything, rather than implying that the original study still has some validity.

In my opinion kubernetes is fundamentally hamstrung by the overly simplistic operator model. I really like the general idea, but it's not really possible to reduce the entire model down to "current state, desired state, next action." It means that an entire workflow ends up in the next action logic, but with so many operators looking at the same system state it's not really possible to know how the various components will interact. The problems with helm are a subcase of this larger issue.

By analogy, this is the same issue as frontend programming faces with the DOM. Introducing a VDOM / reducer paradigm (like react) would go a long way towards solving these problems.

QAnon (often shortened to Q) is the name given to the persona that has been adopted by a few different individuals and is almost certainly currently controlled by Ron (principally) and Jim (secondarily) Watkins. Calling the posts "Q intel drops" is coded language the tells me you are intimately aware of the QAnon conspiracy group. Let me be clear: Q is some random dudes larping on the internet to give their own pathetic lives some level of importance they do not deserve. Q is a fictionalized persona.

I don't mean this as an insult, but after looking through your post history, you should consider psychological counseling. I recognize the coded language you are using. I'm sorry to tell you this, but you are in a cult. These people are taking advantage of you.

If you have a subscription, the HBO documentary I mentioned, Q: Into the Storm is excellent. I think that the struggle mainstream news orgs have in talking about QAnon is that it's very difficult to encapsulate all the interlocking conspiracies and figures. There's also a wide degree of variance for what various bits and pieces mean, but there are a few central tenets. The main one is that there's a mysterious high ranking figure in the government, Q (who gets his name due to the Q-level security clearance he or she has). Q makes very abstract posts on 4chan -- and then other sites -- that seem to suggest that there's a massive pedophile ring inside the government and that Donald Trump is working to expose it from within. Many figures are implicated and, by strange coincidence, they happen to be figures that the far right detests. The Clintons and Bill Gates take starring roles but honestly it's impossible to keep track of it all.

Into the Storm is centrally focused on the identity of Q, but in the process of pursuing the answer the creator of the documentary, Cullen Hoback, ends up going on an absolutely wild hunt. I find documentaries of this sort to be a bit hit-or-miss, but this is one of the best of its subgenre. The subculture and background surrounding QAnon is interesting in its own right but, of course, the real highlight is the characters. Fredrick Brenan, Jim & Ron Watkins, and Hoback himself are fascinating individuals.

The contents of the letter are interesting in their own right, but there are 2 aspects that strike me as particularly interesting. First off, he doesn't seem to regret killing his wife, more the consequences of his crime. He doesn't mention her by name or say that she didn't deserve it.

Secondly, the person he's corresponding with, Fredrick Brennan is fascinating in his own right. He's one of the 3 central characters inside the HBO documentary, Q: Into the Storm, and has a totally bizarre relationship with Jim and Ron Watkins, the two figures currently steering the ridiculous QAnon set of conspiracy theories. It's a very strange confluence of interests.

Although no lead is insurmountable, the fixed capital investment and mature software ecosystem specific to this sector makes it harder to imagine what a competitor would look like.

Given how large the prize is, the next chapter of chip development is likely to be nvidia vs state sponsored projects. China, in particular, will funnel further resources into acquiring this technology by any means necessary, including (more) industrial sabotage and outright theft. It's going to be interesting to see how this will play out. Up until a few years ago China was viewed as being a formidable competitor for projects of this nature, but as the country has moved to become increasingly authoritarian, so too have its decision making and execution declined in quality.

this fits well with my vision: every website will have unskippable land acknowledgement prompts that list the entire history of that nation

There are some really profound misunderstandings of how bankruptcy and credit work in this thread.

Creditors are ranked by seniority and get paid out in order by seniority. Equity is below every creditor and has been completely wiped out. Companies are not allowed to take cash or sell the furniture to pay out employees. That money is legally owed to creditors and attempting to stiff them is theft. I'm sorry to those affected at Convoy, but that is the reality of working at a startup that goes through a hasty liquidation. Convoy is not being "cheap" about this, as some are suggesting. Any "retention bonuses" are to keep executives around for long enough to unwind the company in an orderly fashion. Nobody is getting rich off failure; if everyone were to walk away there'd be nothing left and therefore nothing to recover and distribute. It's bad for everyone, but the alternative is worse.

As for the larger situation: Convoy were a digital freight brokerage. They acted as intermediaries between shippers and carriers and make money on the spread between the two. They got into trouble because the entire freight sector has been suffering a double whammy of cost increases due to inflation (ie. diesel costs) and a slowdown in demand. This has caused a number of carriers and brokerages to go bust. Freight brokerage has some other properties that make Convoy's situation particularly serious. In particular, carriers typically securitize their accounts receivable. In freight, this is known as "factoring." Convoy got stuck holding the bag after their partner carriers went under, having already paid them for their service, but still waiting for payment from the shipper.

Worst of all, Convoy had no exits because their only potential acquirers are in the same industry and are also getting completely crushed.

This thread is already a damning indictment of google cloud. Price cuts alone won't attract customers -- google has sustained significant reputational damage with their crappy customer service and support. That's hard to fix

The CFO leaving is always a big deal. It can mean nothing but it can also mean a really spectacular blowup is about to take place.

I don't think that the problem is that JavaScript may or may not have this feature, it's not even that the language is large and all-encompassing; it's that mixing and matching features is highly appealing and very rarely works out well in practice.

JavaScript is not really a single language built around a single specific style or set of needs any more. In this day and age it's an amalgamation of different techniques and styles. You've got classic inheritance and composition for your OOP crowd; you've got your map and reduce for your functional crowd; you've even got your observables for your reactive crowd.

This is all well and good, but I've found that it's hard to write some practical code that blends styles. At some point it's tempting to start adding types, generics and dependency management into a functional project, but it's my experience that this blending ends up getting in the way of itself. Similar story for wanting to do things like having a service that listens to queues in an async way with sync rest APIs. It seems like having a common set of middleware would make it easy to support both layers; however this is easier said than done. These things sort of work but it feels deeply unsatisfying, requiring constant switching between observable and async/await styles with error handling being a constant concern.

this is standard industry parlance

Is there an accompanying architecture diagram that goes with this?

I'm not sure which bits and pieces are included without poking through the code

EDX has been in the works for the better part of a decade at this point. Totally unrelated to binance

Authentication is easy and solutions are more or less fungible, but there’s no obvious and easy way to do authorization. OAuth makes authentication pretty easy but all the complexity comes from authorization, which is adjacent to authentication but not directly related to it. Although it’s not OAuth’s job to help you with authorization, it’s an inevitable next step that leads to massively divergent approaches

I think you're getting to the heart of what a smart approach to regulation would look like (for me at least). Try as I might, I can't imagine a situation where someone is able to subvert an LLM to come up with a world-ending scenario using its knowledge of onion-based recipes or by painting fingernails. There is no point to regulating every single potential use of an AI: to do so is to cede the future to a less scrupulous nation. We already know the kind of data and functionality that will likely lead to a dystopian hellscape because we already do those things and they are already regulated.

Paradoxically, a heavy-handed approach to regulation could guarantee a bleak future for humanity. We risk increasing the incentives for malicious use at the same time as we make legitimate use prohibitively expensive / regulated out of existance. If the war and crime machine is cheap and easy to monetize, don't be surprised when that's what gets built.

The future could be unimaginably better for everyone, but we won't realize that future unless we get the regulation right.

The real problem with AI, as I see it, is that we are not ready to give the average person the capabilities they will soon have access to. The knowledge necessary to build nuclear weapons has been readily available for decades, but the constraints of the physical world put practical limits on who can do what. We are also able to limit who can do what by requiring formal education and licenses for all kinds of professions. This is no longer going to be strictly true; or at least not in the same way. If we impose no restrictions, literally everyone on the planet will eventually be able to do anything, provided they have the resources.

The fact is that AIs are more likely to be accomplices than masterminds, at least for the foreseeable future. What we are afraid of is that they will end up being just as terrible and flawed as we are, but it's more likely that they will do what they do today: what we ask them to do. The greater risk is therefore from malicious internal users rather than from the technology itself.

Perhaps more to the point, there is no effective way to limit the output of an LLM or place restraints on how they work. I think it's foolish to even attempt that -- they just don't work that way. The better way to regulate this kind of technology is to focus on the people. Some data classification and licensing program seems a lot easier to implement than the road we currently seem to be going down; which is either no regulation or insanely restrictive regulation.

Twitter has remained functional largely because all the former employees did their jobs effectively. I've seen what happens when companies lay a tech department to waste and it looks like this. This is a one-way cost reduction program that will forever reduce the quality of workers available to the company.

The way this pans out is that in the short term, nothing much happens except deadlines start to get pushed out. As time goes on everything becomes more and more dysfunctional. All progress grinds to a halt; simply treading water created by ongoing churn and operational needs ends up becoming impossible. Next the business tries to get more and more out of the remaining staff until they get fed up with it and leave. Shortly before the wheels really come off, legal and compliance issues will suddenly start ballooning. That's where we are now. Nobody wants to work on these kind of projects -- they do nothing for your career and nobody will celebrate your work -- but they are not optional. Elon seems to have decided simply not to comply with laws he dislikes in regulatory environments where he thinks he can get away with it. This can work for a while, but companies with no internal governance and that regularly flout laws can don't tend to last for too long without making painful changes. Let's see how it turns out. Given the reported state of the company it sounds like he has about a year to figure it out.

Although they weren't mentioned, social media and streaming video are the most striking cases. There's one flavor of social media website and it sucks. It's the same infinite scrolling page of 30s attention grabbing headline/video/images posted by the same people on the same 5 or 6 websites made by the same set of engineers. There are minor variations, but they're better described as "positioning" than truly differentiated experiences. What makes instagram "better" (or worse) than tiktok or twitter? It's the same soup filling the contours of the same, slightly differently shaped bowl.

Streaming video is even less differentiated. Every player looks more or less the same now and the content is largely fungible between platforms. It used to be that Netflix had superior selection and had the best player, but content has been balkanized and every platform looks and works roughly the same way. Almost every platform is now targeting the same type of content: low cost, easy to mass produce, and mass consumable. I will give youtube and twitch credit for being quite different -- at one time. Now, for whatever reasons, these services seem like they're on a steady march towards homogeneity too.

There are signs of the phenomenon the article describes everywhere and it seems to have happened across most artistic and business disciplines around the same time period. I would be very curious to know the ages of some of the posters in this thread since there are some interesting differences in opinion/experience. If you see it, you see it and feel it everywhere, but if you never saw what life was like before there's no way to make any kind of comparison. There were places and services and things objects that did not even attempt to cater to average expectations. That difference was real, in ways both tangible and abstract. This article won't resonate with posters in their 20's like it does for those in their 40's.

The complaint is very damning. They have receipts showing that the CEO directed the company to set up a sham compliance department to evade restrictions and sanctions. Also there's all the historic fraud and market manipulation. Hard to see them survive this.

Where is the corruption in any of this? That is a strong accusation without any semblance of an argument.

During a downturn the stronger companies often acquire distressed competitors and consolidate. Evidence of corruption would be CS being propped up and executives allowed to pay themselves big bonuses. I'm not seeing anything like that, are you?