> It's a sad indictment of our society that there is always a shortage of money for medical care...
It has nothing to do with society; there is infinite demand for medical care. The upper limit is whatever it takes to live until the universe's heat death in good health. That takes a lot of resources.
However much society spends on medical care, there is always more that could be spent. The modern era has the best, most affordable medical care in history and people are showing no signs of being satisfied at all.
While war spending generally just causes pain for no gain it doesn't change the fact that there will never be enough available to satisfy people's demand for medical care. Every single time people get what they want they just come up with a new aspirational minimum standard.
I suppose I'm neutral on the topic of strategic helium reserves; but what aspect of this is supposed to be pound foolish? What exactly is the buffer meant to be for?
A strategic petroleum reserve makes a lot of sense, petroleum is part of the food supply chain and it'd be stupid to be in a position where a short disruption could cause people to starve. Not to mention the military implications if an army can't zoom around because the petrol stations run dry for whatever reason.
I don't see anything on the list of uses for helium that looks particularly time- and helium- sensitive in the way that a strategic stockpile would help with.
AFAIK, it was created to fill airships in case of a war. So the original intent is completely outdated.
Also, last I knew about it, the reserves were only reduced, and the US still has some. I have absolutely no idea on the cost/benefit of it, but I don't think any other country keeps a large reserve of Helium.
The article for example mentions MRI macines, aerospace engineering, fiber optics and semiconductors, so I guess it depends on if you want those things to still be available in a crisis
That does sound kinda minor? A worst-case scenario of a month or two without MRI machines or "aerospace engineering", whatever that means doesn't sound particularly scary. And that is making some pretty unrealistic assumptions like there is literally no helium, hospitals don't have private reserves that can last a few months and there are no replacement gasses or alternative options of any sort. And people can make do with limited fibre-optic or semiconductor manufacturing. We have crisises in various computer components every few years (I can think of HDD, RAM & GPU supply shocks over the last few years). Doesn't seem to be a major problem. A couple of months of disruption isn't a strategically interesting event.
If you're worried you can keep your own helium reserve? Then if there is an emergency and it turns out that you don't need an MRI you can sell the helium to whoever does and feel really good about your foresight.
I'm not seeing any need for a strategic reserve here. There aren't any strategic issues. It is a bit far-fetched that a helium shock will even lead to the end of MRIs.
Everyone also could keep their own supply of gas and their own batteries for electricity but it turns out that is not expensive and foolish compared to centralizing such backups.
A lot of people die every month. We're talking about a probability near-0 event where I imagine it'd be difficult to pick that deaths out from general background mortality - admittedly just based on the fact I don't recall anyone I know who needed a life-saving MRI but I know a few who died. That isn't much of a justification for a strategic helium reserve. Some level of risk just has to be tolerated, we can't afford to have a contingency for every possible hypothetical.
> But just because a company has appointed a CRO doesn’t necessarily mean that it has made risk management a high priority.
Priority or not, it suggests the company doesn't understand risk. In a company that doesn't look at risk-adjusted rates of return as a natural part of how they do things a CRO is mild bad sign.
An analogy might be helpful. Testing code is, with some squinting, a form of institutionalised risk management. Any particular test doesn't necessarily do anything useful, but they apply a certain level of pressure that means the code in general fails less and force people to think more about how they're writing their functions. If a company tells you that it has a special pool of coders who add tests, separate from the ones that write the actual code, that is a bad sign that they know how to do testing. A huge chunk of the value is forcing the person who makes the front line decisions to think about what they are doing. Not to say a dedicated testing team doesn't sometimes make sense in some unusual companies, but it is an exception to the rule. Risk management isn't the type of responsibility that should be separated out into a separate role for most companies because that is much less valuable than the people doing the work being part of a management chain that understands risk.
You completely miss the role of CROs or risk function in an organisation. Using your analogy, the Chief Testing Office would not write the tests. They would establish how test coverage is defined and measured, and the target coverage. They would monitor the progress of each team in meeting these targets. It is a governance role that sits as a second line behind the first line that has the immediate responsibility to manage the risk.
Risk adjusted rates are not traditionally in the mandate of a CRO. They sit with Finance or Treasury. And they should be abstracted from front line, who would experience them only through optimisation of their funding.
This sounds well lined up with what I was saying? The CRO doesn't manage risks. Having him in with the executives is a signal that the company is putting resources into communicating with the regulators rather than that they are committed to managing risks in any way. That isn't what these regulatory-heavy roles are for. Their job is to make sure the regulators don't investigate. That is in no way a signal that the company has any ability at risk management, and is a slight signal that they might think "risk" just means that the government will sue them or shut them down.
If a company were actually serious about managing the risks it'd be some relatively quiet role reporting to someone responsible for operations like a CTO, COO or head of product. Maybe part of the CEOs personal staff but not an exec.
>If a company were actually serious about managing the risks it'd be some relatively quiet role reporting to someone responsible for operations like a CTO, COO or head of product. Maybe part of the CEOs personal staff but not an exec.
Actually, that is the real red flag. That quiet little role is completely overridden by the first inconvenienced exec. Having a C-level at least means the role is considered co-equal, and if outweighed by the rest of the C-team they at least have the resources and discretion to do the best they can with what they have.
The approach you mention is what I call "ablative armor for management" or an accountability sink. Responsibility is delegated, but no authority is actually invested. If they can't say no with sufficient gravitas to upset operations, then they're nothing but a figurehead.
> If a company tells you that it has a special pool of coders who add tests, separate from the ones that write the actual code, that is a bad sign that they know how to do testing.
I disagree this is necessarily a bad sign. The people writing the code have blind spots and they may also not necessarily be experts at testing. Probably the highest quality software I ever worked on was in a setup where we had a combination of the developers writing tests plus dedicated people who wrote only tests. That said, I think this setup is secondary to the quality and experience of the teams and the individuals.
> A huge chunk of the value is forcing the person who makes the front line decisions to think about what they are doing.
I would look at this differently and say a huge chunk of the value is coming from making sure you have the right person in the front line. The wrong person being "forced" to make decisions they're not good at is not going to help you a lot. The right person doesn't need forcing to make the right decisions. People and culture drive outcomes and not process.
I would argue that testing code is “Risk Mitigation” not “Risk Management”.
It is nuanced, but at least in large Systems Engineering orgs, Risk Management is typically a different thing entirely.
It entails documenting known risks, evaluating the likelihood and potential impacts, defining mitigating actions, tracking the closure of those actions and the resultant reduction in the likelihood of the risk manifesting.
This is both centralized and distributed. The specific SMEs provide most of this input/definition, but it is also useful to have a centralized understanding of all the system risks by someone with a system level purview.
Agreed. This maps directly to the white-box vs black-box testing distinction: either you own your priors and trace the full data lineage from training through validation, or you're relying on an opaque validation set of unknown provenance. And that's before factoring in the organizational politics.
I find this line of thinking similiar to copmanies with "innovation" officers. That is, having an employee who is in "charge of" innovation implies all other employees dont?
I meant that in the sense that a typical SaaS company has no reason to be formally thinking about risk adjusted returns and therefore has no need of a CRO. If anyone cares product can do a guesstimate or something. Most companies shouldn't have a CRO.
If you’re a B2B SaaS with no CRO, good luck with vendor assessments. B2C you can skip it before reaching a critical mass where regulatory pressures will mandate it.
> While handling the legal defense of Pastor Cai Zhuohua, who was charged with “illegal business practices” in 2004...
I can't help but call to mind the concept of "illegal enemy combatants" who could be sent to Guantanamo Bay back in the War on Terror days [0]. It always struck me as a uniquely American concept that one could travel half a world away, attack someone's homeland, declare them to be fighting back in an illegal manner and get a pat on the back for upholding justice.
Obviously it'd be better for China - and the rest of us - if their internal dialog was more coherent. However it seems less useful to ruminate on one particular example. The country is so large that without estimates of magnitude and broad statistics all that can really be said is that everything happens somewhere in China. Any one example is ultimately an internal matter for them. Authoritarianism isn't much fun but the problem isn't the one-off miscarriages of justice but rather the seeping pervasiveness of forcing people to make stupid decisions instead of letting them better themselves in the manner that they see most fit.
> In today’s China,... judicial independence [is] labeled as erroneous ideological trends of the West. In fact, justice is justice, and doesn’t distinguish between East and West.
And I will say that if this part is what people wanted to focus on, the dry distinction between dependent and independent judiciary is actually one of the major advantages that the West maintains over China. But in the article it is more of a one-line thing and part of a longer, vaguer list
For anyone interested, the dynamics are something like:
1) Commercial entities generally try to sell their products for a profit (positive price).
2) Negative prices make this quite hard.
3) To get back into positive profit, they're going to need to charge a higher price later on.
There are 3 ways the overall market can go if prices are negative - either prices go positive at some other time to make up for the losses the power providers are making, someone figures out a use for the free power and the price stops going negative or power providers go out of business. Typically what this seems to result in for renewable-heavy grids is occasional (even regular) negative wholesale prices and some impressively high retail prices. Free electricity turns out to be very expensive; it shows the grid isn't coping.
They've been asking this question since before 2013. The writing has been on the wall since the US started demonstrating that it thinks debt monetisation is an acceptable strategy.
Have outlets like Deutsche Welle been speaking like this since 2013? I don't really think so. Not to defend the past administrations, but that the question is starting to hit mainstream media does in fact tell a huge lot about this current one.
This has been a topic of continuous debate since at least ~2000 in Germany. The German Wikipedia has a whole section covering it¹. Obviously, the debate gets more intense every time the relationship between Germany and USA gets strained.
Sovereignty is a matter of constant maintenance, but only recently has the discourse been affected by lack of trust in the admin itself, not only for transparency or logistics.
> think there is substantial difference between the two sub-parties
You said something close enough.
Yes, they're both terrible. Biden's admin allowed a genocide to take place in Gaza while pretending to give a damn about it and repeating propagandist lies from Israel. But the Trump admin is openly looting everything they can get their hands on.
The real answer is something HN doesn't like so I won't advocate it openly, but it involves society paying to take care of people to provide homes, provide medical access, things like this. Neither party is interested in that.
But only one is brazen about harming peopler and making content from it or unashamedly posting AI-generated memes.
> The real answer is something HN doesn't like so I won't advocate it openly, but it involves society paying to take care of people to provide homes, provide medical access, things like this. Neither party is interested in that.
There are enough people on HN who think working social democracy is a great option; not everybody here is a libertarian cryptobro, an eastern european with decades-long PTSD or a hardcore conservative.
I’m learning the opposite lesson. The US is surprisingly fragile and a single president with no morals or ethics can do far more damage than anyone could have imagined.
It's a lot more than "a single president". It's the two thirds of American voters who either voted for him or couldn't be bothered to vote against him, even after having seen how he behaved in his first term.
That's why he's not the biggest problem. He's more of a symptom than a cause. He won't be around for ever, but even after he's gone, most of those same people will still be around.
no i believe we were taught correctly that the powers of the presidency are limited. the issue is that he is the leader of a group that is embedded into every branch of government at multiple levels. it is not hte case of a random crazy president abusing presidential power and everyone is just at the mercy of a lunatic. It is the case that every wild thing he does is upheld and supported by a large network of people who otherwise would have the power to absorb and dismiss his attempted actions.
Nah, it's more like the United States is vulnerable to 50 years of reactionaries, upset about things like civil rights for African Americans and women in particular, actively working to dismantle democratic processes.
That's why farmers vote for politicians and policy that put them out of business. In my hometown, the county went from 150 dairy farms in 1990 to 3. They voted to keep their guns or whatever bullshit they were told to worry about, and were wiped off the map by the deregulation they "wanted". My little town is a rural ghetto, the richest people in town now are the school principal and state trooper.
POTUS is an acute condition. The sickness lay deeper.
Debt monetization is not happening. There's been significant expansion of the money supply, which got the US through COVID at a one-off cost of about 10% inflation in one year, which I think was a reasonable cost of covering the crisis (especially compared to the GDP response in less generous countries!)
What's happening is a much simpler, more drastic concern: does the US respect its international commitments?
I think it's also worth noting that a lot of people have 'forgotten' other factors happening besides stimulus and monetary (central bank) policy—especially when it comes to energy and food prices:
The US stopped respecting its international commitments in Trump's first term when it single-handedly withdrew from the JCPOA (singed by way more countries than just the US) without a single valid reason to do so.
Since then, it flip-flopped on the Paris Agreement, single-handedly put tariffs on goods imported from literally every single country in the world, withdrew from WHO and so on and so on.
Not only does the US not respect the commitments it already agreed to, it hasn't done so for the past 10 or years.
> Not only does the US not respect the commitments it already agreed to, it hasn't done so for the past 10 or years.
These commitments were commitments by the administration in the White House at that time. They were not accords or treaties ratified by Congress. Agree with them or don't, they should have been understood as what they were: limited in legitimacy and free to be canceled by the next administration with no discussion.
This isn't a matter of arcane political skullduggery, it's spelled out in the US Constitution's definition of treaty, Article II, Section 2, Clause 2. The people were not formally heard.
The administration is administering the USA. It's the USA that commits. To other countries it does not matter what your internal rules are to what your country does.
Well I can't speak for all the people who own gold, but I expect the order of actions they'd generally prefer is move the gold to a vault somewhere they think is financially stable first and then engage in a relaxed debate about the merits of storing gold here or there second. It doesn't cost that much to move a bit of metal around.
If countries enter the lunatic phase of money printing you never quite know what they're going to do next. But it probably isn't going to be good for asset owners and it may well already be too late to get things out of well known vaults. Better to be a bit early.
It appears that the gain mentioned is a realization of their asset value. I would also speculate that what happened is that they wanted LBMA bars because those are a standard variety and therefore easily tradable. An arbitrary LBMA bar is generally fungible. I would also speculate that they held many bars in the US from ancient times. After 2008, they repatriated 200-ish tonnes and 'upgraded' them (which I would speculate again is 'ensured they were LBMA-standard').
These articles all have the flavour of the game of telephone common in this style of article where the currency that the gain is in changes wording, the motivation seems to shift, and phrasing lacks real detail instead relying on 'upgrading' and 'refining'.
I wish there were a good LLM agent that were capable of tracing all this back to the real original source that spawned all these things, but the information environment is currently full of smoke and getting real news is quite hard.
I can't realistically conclude whether this was politically motivated or not. The original motivation is sufficiently strong on its own, but it is completely normal for governments to move something to be earlier, or to do a marginal thing if there is other gain.
They started the process in 2005 [1]. The goal has been to upgrade all their goal to modern purity standards (99.999% purity). The repatriation to France may have been done for national security reasons, but not political as in ideological.
I'm not seeing how this matters, they were already doing that - the market is a big auction to work out the overlap between lowest salary employees will work for and the highest salary employers will offer. In that process employees also use data to figure out the highest salary that will be offered. The thing forcing employers to pay the salary they do is that if they offer less someone else will gazump them for the employee's time. It has nothing to do with the circumstances of the employees lifestyle. The lifestyle adjusts to the salary.
> the overlap between lowest salary employees will work for and the highest salary employers will offer
There is still an element of unknown because both parties do not know each others numbers, which allows employees to still negotiate. You are now talking about information asymmetry where the party with the information will now have all the bargaining power.
When I went from working a $150K job to getting offers from Meta at $300K, the initial number they offered was $250K, and we worked upwards. I absolutely would’ve taken the job even if they offered $200K and not negotiated. But they did, based on information asymmetry. Now imagine a world where meta knows exactly how much I make and all the other information about me. I’d probably get a minor bump over my previous salary.
Edit: I ended up taking a different offer. I don’t work for and have never worked for Meta.
You might want to rethink your example if the counterparty offers you 50% more than you wanted then you reject the deal; it makes adding the framing a bit pointless because it is clear you weren't ever going to accept the job for $200k.
And you're underestimating how much of an impact the broader market is having on Meta's thinking in this scenario. If your silver tongue or secret number was a factor here then everyone would end up being overpaid because they wouldn't reveal that they were happy to work for a reasonable amount. It doesn't matter how much or little Meta knows, they're only going to offer $300k if they have a reasonable belief that you can find a job for $300k somewhere else; informed by a pretty detailed analysis of the employment market. And in fact that appears to be exactly what happened in your story. Nothing about that dynamic has anything to do with your salary history or spending habits and them getting better information on those things doesn't change your negotiating position. Since a key factor is the future, even if they know you'd say yes to $200k, they'd still be best served offering you more money. I've had that happen to me 2 or 3 times because I'm a sloppy negotiator and don't try very hard to optimise salary.
> You might want to rethink your example if the counterparty offers you 50% more than you wanted then you reject the deal
I rejected the deal because I got even more elsewhere. My framing still stands. In a case when only one employer has the information, sure they’re better served by offering me more money. But in an environment where all of them have the information, this no longer is a problem. At a system level, this is a problem for employees.
But if Meta wanted to hire you and had perfect information, it sounds like they'd discover they needed to offer you salaries in the $350-400k range? That sounds like it might be good for you.
The story you seem to have told is they just wasted time low-balling you because they didn't have enough information to offer a competitive salary. You weren't ever going to settle for $250k, they didn't have enough leverage and they lacked the information to identify that. I'm not sure how you're seeing this story as one where more information to Meta leads to them offering you a lower salary. It seems like you'd have rejected them regardless unless they went higher.
All the employers knowing that you'd have "taken the job even if they offered $200K" seems to be completely useless to them. They're locked in an auction and the market price for your time is nowhere near $200k.
> they'd discover they needed to offer you salaries in the $350-400k range?
No, such a discovery wouldn't be possible, because nobody would pay that amount to someone who was willing to accept $200K.
> They're locked in an auction and the market price for your time is nowhere near $200k.
There is no magical market price that exists outside the market dynamics. When bidders know that one's current salary is $150K, their willingness to offer higher salaries will diminish accordingly.
>Now imagine a world where meta knows exactly how much I make and all the other information about me. I’d probably get a minor bump over my previous salary.
Not necessarily. People don't change companies for just any value greater than current TC. There is a big cost to switching companies -- it's going to shake up your lifestyle, you might lose some relationships, reset your company-internal network and reputation, reset technical and organization context etc. Possibly even moving your home (even if a new job is in the same city, people often move to be closer to it anyway).
As a matter of policy I wouldn't switch companies for less than a 30% monetary premium over my current TC (I'm a SWE), and other soft criteria like type of work and company culture. In my early career I've gotten 50-100% premiums each time I made a hop.
My policy is the opposite – I switch companies every two years. Especially if I get too comfortable. Usually because I’m bored or because the company grows too much (>100 people is too much for me).
Are you worried that you never get to see the results of architectural decisions you make? Two years is not much time to make an impact and see it through if you’re senior+
This is usually a nonsense argument unless you're in a very slow-moving company. Two years is not only enough to see the results of your architectural decisions, but by that point the architecture would already be due for change.
1-1.5 years in a fast-paced software team is plenty enough to learn 95% of everything you're ever going to learn there and live through one major system lifecycle.
This allows all sorts of normally illegal discrimination via ai pass through. Never hire pregnant women, sick people or employees over 30 again. Target for race and religion whatever you want. Basically everything that’s scary about chinas social credit score except private run with zero accountability.
If that’s true and this has a null effect, why would a business pay for it? There must be some utility for them. Like others already pointed out: information asymmetry undermines worker’s ability to negotiate, resulting in lower wages for everyone.
Same. This is how any market works. I think one of the problems (US only maybe?) is that talking about salary is generally taboo. The first thing you should do with all your friends is be open about how much you make, and then go online and look for more data. That way you start with some good data points. Next is talk about negotiating. I was taught growing up that everything is negotiable, but later in life I learned that not everyone knows that.
That doesn't sound like much of an issue. Bob was already going to encounter problems that are too large and complex for him to solve, agents or otherwise. Life throws us hard problems. I don't recall if we even assumed Bob was unusually capable, he might be one of life's flounderers. I'd give good odds that if he got through a program with the help of agents he'll get through life achieving at least a normal level of success.
But there is also a more subtle thing, which is we're trending towards superintelligence with these AIs. At the point, Bob may discover that anything agents can't do, Alice can't do because she is limited by trying to think using soggy meat as opposed to a high-performance engineered thinking system. Not going to win that battle in the long term.
> The market will always value the exact things LLMs can not do, because if an LLM can do something, there is no reason to hire a person for that.
The market values bulldozers. Whether a human does actual work or not isn't particularly exciting to a market.
> we're trending towards superintelligence with these AIs
The article addresses this, because, well... no we aren't. Maybe we are. But it's far from clear that we're not moving toward a plateau in what these agents can do.
> Whether a human does actual work or not isn't particularly exciting to a market.
You seem to be convinced these AI agents will continue to improve without bound, so I think this is where the disconnect lies. Some of us (including the article author) are more skeptical. The market values work actually getting done. If the AIs have limits, and the humans driving them no longer have the capability to surpass those limits on their own, then people who have learned the hard way, without relying so much on an AI, will have an advantage in the market.
I already find myself getting lazy as a software developer, having an LLM verify my work, rather than going through the process of really thinking it through myself. I can feel that part of my skills atrophying. Now consider someone who has never developed those skills in the first place, because the LLM has done it for them. What happens when the LLM does a bad job of it? They'll have no idea. I still do, at least.
Maybe someday the AIs will be so capable that it won't matter. They'll be smarter and more through and be able to do more, and do it correctly, than even the most experienced person in the field. But I don't think that's even close to a certainty.
There's no good definition of superintelligence. A calculator is already way more capable than any human at doing simple mathematical operations, and even small AIs for local use can instantly recall all sorts of impressive knowledge about virtually any field of study, which would be unfeasible for any human; but neither of those is what people mean when they wonder whether future AIs will have superintelligence.
General superintelligence is more well defined, I assume that is what he meant. When I hear superintelligence I assume they just mean general superintelligence as in its better than humans at every single mental task that exists.
> But it's far from clear that we're not moving toward a plateau in what these agents can do.
It is a debatable topic, and I agree with you that it's unclear whether we will hit the wall or not at some point. But one point I want to mention is that at the time when the AI agents were only conceived and the most popular type of """AI""" was LLM-based chatbot, it also seemed that we're approaching some kind of plateau in their performance. Then "agents" appeared, and this plateau, the wall we're likely to hit at some point, the boundary was pushed further. I don't know (who knows at all?) how far away we can push the boundaries, but who knows what comes next? Who knows, for example, when a completely new architecture different from Transformers will come out and be adopted everywhere, which will allow for something new? Future is uncertain. We may hit the wall this year, or we may not hit it in the next 10-20 years. It is, indeed, unclear.
Yes, and this is an incredibly powerful idea. Running fluid flow simulations inside an optimization loop (monte carlo + gradient descent) revolutionized aircraft design, nuclear simulations and geophysics.
When the tool being called updates the LLM's training data, or runs experiments that the LLM can learn from, then that potentially becomes a self-improvement loop.
I went the compensation route: a set of regex patterns that check generated code for known anti-patterns before it runs. The agent can ignore a prompt telling it "don't use mocks here." It cannot ignore a failed exit code from a pattern check. Started with 40 patterns and grew to 138 over a few months as new failure modes appeared.
> It doesn't matter if Bob can be normal. There was no point to him being paid to be on the program.
Yeah, I'm surprised at the number of people who read the article and came away with the conclusion that the program was designed to churn deliverables, and then they conclude that it doesn't matter if Bob can only function with an AI holding his hand, because he can still deliver.
That isn't the output of the program; the output is an Alice. That's the point of the program. They don't want the results generated by Alice, they want the final Alice.
And then you realize that most of science is unnecessary. As TFA points out, it doesn't matter if the age of the universe is 13.77 or 13.79 billion years. So you ban AI in science, you produce more scientists who can solve problems that don't matter. So what?
> we're trending towards superintelligence with these AIs
I wouldn't count on that because even if it happens, we don't know when it ill happen, and it's one of those things where how close it looks to be is no indication of how close it actually is. We could just as easily spend the next 100 years being 10 years away from agi. Just look at fusion power, self driving cars, etc.
Fusion isn't a good example. Self driving cars are a battle between regulation and 9's of reliability, if we were willing to accept self driving cars that crashed as much as humans it'd be here already.
Whatever models suck at, we can pour money into making them do better. It's very cut and dry. The squirrely bit is how that contributes to "general intelligence" and whether the models are progressing towards overall autonomy due to our changes. That mostly matters for the AGI mouthbreathers though, people doing actual work just care that the models have improved.
This is just trash, like almost any AI benchmark. E.g. it says since around 2015 speech recognition is above human yet any any speech input today has more errors than any human would have.
If I would not type but speak this comment maybe 2 to 5 words would be wrong. For a human it is maybe 10% of that.
we have to look at what LLMs are and are not doing for this to be applicable; they are not "thinking", there is no real cognition going on inside an LLM. They are making statistical connections between data points in their training sets. Obviously, that has born some pretty interesting (and sometimes even useful) results but they are not doing anything that any reasonably informed person would call "intelligent" and certainly not "super intelligent".
"Things that have never been done before in software" has been my entire career. A lot of it requires specific knowledge of physics, modelling, computer science, and the tradeoffs involved in parsimony and efficiency vs accuracy and fidelity.
Do you have a solution for me? How does the market value things that don't yet exist in this brave new world?
> There's a common rebuttal to this, and I hear it constantly. "Just wait," people say. "In a few months, in a year, the models will be better. They won't hallucinate. They won't fake plots. The problems you're describing are temporary." I've been hearing "just wait" since 2023.
We're not trending towards superintelligence with these AIs. We're trending towards (and, in fact, have already reached) superintelligence with computers in general, but LLM agents are among the least capable known algorithms for the majority of tasks we get them to do. The problem, as it usually is, is that most people don't have access to the fruits of obscure research projects.
Untrained children write better code than the most sophisticated LLMs, without even noticing they're doing anything special.
I'm willing to bet 10% of my net worth on this. But my claim was not about any given untrained child (for instance, a child who does not want to program would do poorly): a fair bet would allow me to choose the child, you to choose the LLM, use a task and programming language of the child's choice, and have a neutral third-party familiar with the programming language judge "better code". (I would, of course, want to ensure that the judge used an appropriate rubric: RLHF can produce a sophisticated turd-polisher. Perhaps the evaluation process could involve modifications made to the program?)
It is (rightly) difficult to get hold of one uninvolved child, for safeguarding reasons, so it would be better to run it as a school (or interschool) competition, where multiple children may participate. For fairness, you may also provide multiple LLM participants (however you define that). The winner of the contest, as determined by the judge, would then determine the winner of the bet – unless the winning child had been trained, in which case we would fall back to the next-highest-ranked participant. The number of LLM candidates would be equal to the number of eligible children.
However, I don't see a good way to allow each child to pick a programming language and task, without leaving the competition results incomparable. So perhaps each child should be paired with an LLM, and the judge should determine which submission from each pair is better? But then if I only need one victory (to support my claim), this is clearly unfair. So each pair should be tested enough to determine whether they're consistently better than the LLM… but then we are demanding a lot of the child participants, for no real benefit to them.
If we can agree on a workable protocol, I can try to pull some strings and see if we can make this happen. I could use the money.
The rate of hallucination has gone down drastically since 2023. As LLM coding tools continue to pare that rate down, eventually we’ll hit a point where it is comparable to the rate we naturally introduce bugs as humans programmers.
I wonder how much of the decrease in hallucination is because the models are getting better, and how much is because these massively over-funded companies are adding a bunch of one-off shims at breakneck speed. IE - are they truly improving the cognition, or just monkey-patching the hell out of it?
The recent article where the AI companies are paying experts in the field to help train the models makes me wonder if they're also manually fixing a bunch of post-processing errors as they come up.
LLMs are still making fundamentally the same kinds of errors that they made in 2021. If you check my HN comment history, you'll see I predicted these errors, just from skimming the relevant academic papers (which is to say they're obvious: I'm far from the only person saying this). There is no theoretical reason we should expect them to go away, unless the model architectures fundamentally change (and no, GPT -> LLaMA is not a fundamental change), because they're not removable discontinuities: they're indicative of fundamental capability gaps.
I don't care how many terms you add to your Taylor series: your polynomial approximation of a sine wave is never going to be suitable for additive speech synthesis. Likewise, I don't care how good your predictive-text transformer model gets at instrumental NLP subtasks: it will never be a good programmer (except as far as it's a plagiarist). Just look at the Claude Code source code: if anyone's an expert in agentic AI development, it's the Claude people, and yet the codebase is utterly unmaintainable dogshit that shouldn't work and, on further inspection, doesn't work.
That's not to say that no computer program can write computer programs, but this computer program is well into the realm of diminishing returns.
> That doesn't sound like much of an issue. Bob was already going to encounter problems that are too large and complex for him to solve, agents or otherwise.
I have literally never run into this in my career..challenges have always been something to help me grow.
I would take that bet on the side of the wet meat. In the future, every AI will be an ad executive. At least the meat programming won't be preloaded to sell ads every N tokens.
The multiple sources don't know either, the reason the picture is shoddy is it is necessarily composed of the primary information that is coming from ... people with the strongest incentive to lie. There aren't a lot of independent ways to assess the situation. And of course that is part of the fog of war - even the militairy struggles to put together a picture of what is going on. I'd imagine that defining where the front-line is presents a complex and uncertain exercise for the commanders involved.
The only thing I think can be said reliably is that this has been going on for weeks, the Strait seems to be more closed than open, Trump is clearly out of his depth and the US is sending more units to the area. All of those point to a serious problem for the US military.
It has nothing to do with society; there is infinite demand for medical care. The upper limit is whatever it takes to live until the universe's heat death in good health. That takes a lot of resources.
However much society spends on medical care, there is always more that could be spent. The modern era has the best, most affordable medical care in history and people are showing no signs of being satisfied at all.
While war spending generally just causes pain for no gain it doesn't change the fact that there will never be enough available to satisfy people's demand for medical care. Every single time people get what they want they just come up with a new aspirational minimum standard.
reply