It might seem obvious but in truth is not - to be honest, a strictly typed language where models can play adversarial competitions against a strict compiler/linter does have advantages, as mentioned in the article itself but that's not related to training size, the advantage is exactly that it can generate synthetic useful data due to compiler/linter guarantees - regarding training size as long as some logic showcasing the constructs exists that is all that is needed. If you have 2 correct examples for each feature or language construct then the probability of the a LLM learning it and applying it is very much guaranteed, if you have thousands of diverging uses of patterns and constructs in the training data (with a lot of bad examples or wrong uses) a LLM might, I dare say will, actually perform worse, since the probabilities of what particular variation being the "correct" one to use are all over the place.
What can happen is sometimes patterns that are unique to certain languages aren't surfaced and so can't be "learned" but that is a different case, since you can add examples - or patterns that go beyond syntax (threaded code, process isolation, interaction between different modes of resolution, etc) where you need the agent to be able to understand and plan higher-level logic.
I wrote CSSex as a css pre-processor (similar but not the same as SASS/SCSS) and my "boss" at the time fed the existing cssex files to GPT (+2 years ago) and it was able to write CSSex just as fine as if it was writing something that was present in its training data set. It never introduced syntax bugs, the bugs it introduced were all relative to complex cascading styles rules in an existing large (for a definition of large in webapps) codebase, rules interactions (CSS), browser quirks, and complex logic to do what we "humans" wanted (or him in this case) and so on.
I loved my first 100h (maybe 200h?) of Factorio and thought it was like crack, reached a point where I started designing city blocks as monads because I wanted to solve the game - or well, tried, because the blueprinting system becomes unwieldy once you try to do some advanced blueprinting and things that should have been easy to became just deep time sinks to iterate on.
One can't really point the finger at an hugely successful game - on the other hand I think it's another example of how bad architectural decisions in the underlying systems creep up and cripple future development.
The following is from the point of view of the "end-game" stage, of when you're doing things because you're trying the limits of the system - not your first runs, those were fun and I guess it's reasonable to expect a game to have a finite life.
With blue-prints you can't really customise them beyond parameters - which in turn are limited in the number of them that you can use, don't have any way of ordering, don't have any way of copying/applying them, don't have any way of relating them to other aspects of the blueprint itself (besides very basic stuff, that sometimes was broken too).
You can't apply "sub" blueprints inside a blue-print, making the whole idea of "blueprinting" quite limited. Like if you want to make a city block with train stations that can either take liquids or solids, you end up having to create 2 (almost identical blueprints), or 3 (a general layout city block and 2 "internal" for the stations).
To work around the parameters limitations you need to create monstrous circuitry abominations, that can be fun the first time you're implementing them, but are just horrendous in interface afterwards - doing a simple circuit change can take an hour and endless clicking, keeping memory of it all, double checking it and so on... The same to updating them if you have plastered N city blocks on your way to factorio greatness and realised later it needs a "slight" change in the circuitry control/flow.
The train system was also severely hampered by (although better in v2) their underlying choices - even the "wildcard" improvements of v2 still show those design issues popping up all over their interfaces and makes you reach for awful designs due to the way wildcards work with the stations naming, connections and many other details. Part of it is due to - from what I read - the way they implemented IDing the train/wagons - and other is just the inexistent connection between the seemingly connected aspects of trains, stops, and stations, resources, and units.
The network/automation (bots) was also pretty limited in functionality, yeah, it felt like a great advancement once you get to it, and then you have a learning curve until you master it but it just ends on a cliff - you can't really program the network in ways that seems obvious you should.
Some of this might be to keep the game "flow" mostly manual, but to me, that I don't like doing click & run endlessly and like games that progress further, it all got old at some point and reaching those hours of game-play others consistently talk about looked like subjecting yourself to endless punishment rather than enjoyment. I tried and in part succeeded to create city blocks that you just had to paste and pick your inputs, choose your outputs, place some trains and choose the type of input and output, and it all worked - no stuck trains, not the highest throughput by space area, but just endlessly expandable without bottlenecks - at that point I considered the game solved and moved on but would have spent (probably quite many) more hours on it optimising and testing designs if it didn't end on that functionality/UI/flow cliff once you reach a certain stage (and blueprint bugs too, like them disappearing after being saved and so on). And I had to use the map editor mods to be able to design the blueprints and iterate on them otherwise impossible...
In the end, I feel like the game provided me enjoyment worth its cost (and that's what matters) but at the same time, paradoxically, was a bit of a let down because the more "engineering", "architecture" like aspects of the game were not really fleshed out or deep at all. Like all the ingredients are there, but... It's mostly aimed at spreadsheet min-maxing and I prefer other sorts of puzzles.
If china that is the number one producer of solar panels and batteries is investing in nuclear perhaps that shows that indeed it's not such a bad bet? Specially taking into account the military applications of such capabilities.
It would also seem that investing into it may perhaps render significant gains into the future as R&D becomes more involved with actual running systems and funded (skin in the game?)
It seems the only ones who lost by abandoning their nuclear programs were European countries (specially Germany) in detriment of Russia and other Oil trading countries... It's even probable that much of the "bad press" nuclear has had for the past decades was in fact instigated by oil-producing countries doesn't it?
Even if China is the number one in installing nuclear power, the installed capacity dwarfs compared to solar. China is "betting" on anything that produces electricity that doesn't depend on imports, and if that is nuclear they go for it. But it is clearly not their main bet:
Wind overtook nuclear production in 2016, Solar FV in 2022. Today both more than double it.
Tables in https://en.wikipedia.org/wiki/Electricity_sector_in_China#Ch... are also very telling. Total power production: 15% wind, 17% solar, 5% nuclear. Nuclear is almost stagnant, but they will probably finish what they are already building, or have planned. That is the way they work. But they are slowing the rate of nuclear installations, it took 5 years to go from producing 400 TWh to 500 TWh, while wind in the same time went from 500 TWh to 1.1k TWh and sun from 260 TWh to 1.2k TWh. These growing rates will drive nuclear from being ~5% of production to be irrelevant in 10 years. But you will read from pro-nuclears that they "plan" to double nuclear output in those 10 years, even when they barely put a +20% in last 5 years. I wouldn't be surprised it they pivot to solar, wind and batteries suddenly (a very China thing to do), abandoning all nuclear that is not already almost finished.
IMO China is not betting on nuclear. Just before solar was economically sound (c. 2015), they believed nuclear could be a solution for their dependency problems, and bet on it. As of today, they are no longer betting on nuclear, or at least the bet is crawling to a halt.
No nuclear debate without Germany. This is my opinion: Germany actually is winning. Not a lot, but their current situation is far from the catastrophe some predicted (massive blackouts! outrageous electricity price!). Nuclear was quickly replaced with solar and wind after an initial coal and gas spike. But now coal usage is going down, and gas is stable. Dependency from Russia is less than 10 years ago. OTOH, France is having serious problems to 1) increase their nuclear production and even maintaining what they have (average reactor age: 35 years, younger reactor: 18 years) and 2) to deal with the install of cheaper energy sources because their base load is so high and can't be easily phased out.
And please, stop with the "bad press" tune, as if everyone that doesn't blindly believe nuclear is the future is dumb and can be manipulated with ease, meanwhile pro-nuclear are very smart and impossible to manipulate. Do you think there is no nuclear or fossil lobby also releasing "bad press" about everything else and selling nuclear as the only solution? It only causes high friction unproductive debates: it is like Democrats firmly believing that Republicans are dumb, and vice versa. Either you accept that you are also being manipulated by the nuclear lobby to support them, or stop accusing others of being manipulated. Italy is running without nuclear power for more than 30 years, but only recently, when it makes absolutely no economical sense to start new nuclear plants if you have your production-demand figured out (not booming like China), some politics, not at all "manipulated" or "bought" by the nuclear lobby, plan up to 10 new reactors against people will (94% of voters against installing new nuclear in 2011). Yes, people are dumb, politics are honest and smart. That is the only possible cause.
I would imagine that China not having a world wide market for selling/dumping their reactors would have some relation too.
I have no horse in this race but anyway, I hope Europe goes back to nuclear, and then get nukes enough to wipe out anyone in the world some 35 times if needed.
Europe is already heavily nuclearized. Even the nuclear lobby recommends a "10% nuclear power in the mix", which many countries en Europe already meet... or way more than that. On average 24% (varies between sources) of European mix is nuclear. Yes, there are a couple of countries (Germany and Italy) that banned nuclear, but there are also the most nuclearized country in the world (France, 60% nuclear), and countries like Slovakia (60% nuclear), Hungary (40% nuclear).
Europe produces almost no uranium, and France has a historical problem with uranium mining in Africa, is not like Europe is giving up a free-of-trouble energy source. Honestly, sometimes it feels that some pro-nuclear won't be happy until 100% of the base load is nuclear. They are asking for more nuclear than the nuclear lobby does.
It seems like it should have been checked by Lockheed not NASA since supposedly NASA provided a specification, that specified the units, and paid for the software no?
> "The output from the SM_FORCES application code as required by a MSOP Project Software Interface Specification (SIS) was to be in metric units of Newtonseconds (N-s)"
(MSOP = Mars Surveyor Operations Program).
One of the recommendations was
> "Conduct software audit for specification compliance on all data transferred between JPL and Lockheed Martin Astronautics"
So yes, NASA should have checked the provided software more thoroughly, but also Lockheed should have actually followed the spec they were given. I doubt the SIS is available online to check any harder
No, if NASA chose not to triple check everything before sending something into space then they are as much to blame as the supplier. Blindly assuming something works like it should without testing it makes an ass out of u and me.
Very fitting for the AI age as well.
"Yeah I just put the specs of the project you're paying me to do (in NASA's case, probably 100% public funds) but Claude the gimp missed the units because all previous training data use Stones and Yards as units, you should have verified it yourself! I just prompt!"
Should have been checked by both. Lockheed should have checked because of contractual obligations; NASA should have checked because of minimal engineering practices (bugs happen in all your dependencies).
They happen and some are put there on purpose too. In this case it's not a "bug" in your dependencies, and yes NASA since it's using public funding should have been more careful, but ultimately it's Lockheed that was paid to do something no?
Lockheed's job is to follow the customer's specifications.
NASA's job is to check to make sure what they paid for is what they received.
I do this every single day as a quality inspector here. I don't know why a bunch of highly-degreed engineers can't do a simple job that a person with oonly a GED does without fail.
Actually no, it's Lockheed's that should have had a test-suite and conformance-suite for it, but probably only has a C-Suite. If the program is non-trivial proving its correctness can be very expensive to prove, both money and time wise, and something that is done contrary to the spec is obviously on the one executing the spec and being paid for it, which probably wasn't cheap and probably these expenses add to the "NASA only burns money..." narrative.
Sleep paralysis is when you're having an intervention in the monitoring realm but your matrix being becomes aware while so. Sources can't agree if it's done on purpose or just the analgesic dosage that was too short.
Next time you have one pay attention to hear if there's sound of steps as if moving away from where you are.
I personally don't understand what's bad about AI being able to do much of the route, boring, crappy parts of programming.
The same way SAP, salesforce, etc were never good products from the point of view of a company, other than enabling a corporate workflow and legal compliance, that was needlessly complex, they still got used when for a fraction of the "subscription" + consultants a better more fit to purpose solution could have been built - much leaner and meaner and to the point, actually enabling empowerment of the company's staff and market differentiation. AI without proper guidance will be the same.
Just a few days ago I checked out https://colonist.io due to it having some open positions (saw here on hn "who's hiring"), it's all AI driven, but it looks and feels awful. Perhaps that's the original catan playing experience but I for sure wouldn't want future software to be like this. Even when using AI, even looking at that, I know it's possible to do much better with AI, it's just that it won't by itself so I don't think there will be a lack of opportunities for those that are into building stuff.
It's actually amazing that an AI can guide me through a mountain of incongruent, difficult to use and understand without years of study, nerd interfaces and programming decisions others have made and allow me to try out a complete new OS (for instance microOS) and write a WebGL tile drawing engine. At the same time, it still stumbles with simple stuff, while at the same time being able to generate a JS + wasm tile game skeleton. But when you look at the skeleton it's actually not that amazing and you can see the training/stolen code leaking everywhere, things you didn't ask are there too, choices are made for things that weren't present in the specs/prompts when they weren't needed either... The upside is that it cost me $0.02. But to be decent I need to spec it properly. I need to know how and why it sucks, and how and why to make it not suck, even if when writing out a spec.
I don't have a horse in this race, I know this architecture can't lead to AGI, it could theoretically solve programming (in absolute), but it doesn't seem so either due to the nature of LLMs. What will happen is that it will be "SAP/salesforce" software for everyone. To be honest, it mostly sucks already and that's not fault of the AI models. What I guess you won't be able to do is charge enough to pay your house on a business model of switching CSS colors, or doing wordpress updates, and pretending you're some sort of wizard and that pounding sand activity is worth what you were being paid.
I think that those who want to build things will have way more leverage and reach than ever. Maybe not in a corporate environment - as the same **** that would tangle basic stuff into 2 weeks of work, instead of a single afternoon may actually be more savvy politically to keep their positions - but everywhere else it feels like it will be so.
And even in other aspects. I paid a bit more than $3k to different artists as freelancers for getting in total less than 100 images for my game ($30-$45/img), of which, prior to finding an artist that was actually willing to listen and do what I asked, was about 70% not worth even the small amount I was paying. With that artist the rate of satisfaction went up, but still not always as I had envisioned. Some of the first artists I worked with were just plain stupid to work with even though I was paying them ok (taking into account their countries). With AI, even though it's mostly built out of stolen work, with a yearly sub to any of those image gen platforms, I did in 2 months more than 2k images of which maybe 1k was usable and I was way happier with them... for about $24. I left thousands of images in credits unused for about $120/year.
There's just so many things that don't require an artist but instead adherence to a spec under a known guard-railed path. This does suck for the artist himself, it does suck for me too since my hourly rates got butchered too, but as a creator, this is all amazing - as long as we get local, free models. For the artists and programmers that were just milking it - well, you kinda prompted this, no pun intended.
that's why after 1 year of product development of these AI 20x maxxed speed, we reached AGI 'wizards', there's really no difference in output, outstanding bugs no longer get solved and sites still suck, even doing things that were just regular development 20 years ago. Are you sure they aren't only producing 2.5% of your output that you manage just by farting into your phone? Are you sure it's 25% really? Seems way to high, days when I have diarrhoea my AI agents move even faster
It's not you, it's X... but what would you expect of a nepo-baby economy of little swines. This is like the nepo wet-dream on steroids. Incompetence and delulu
If you are too dumb or too lazy to figure out what this guy has, then you are the problem and will be looked at as a relic.
Watching the AI slop my sales reps put in their emails is disgusting but the reply telling them how great of a job they are doing and how insightful their email was says differently.
Many people are laughing to the bank while you are still running `--help` to figure out how to run a complex command.
Maybe it's you that needs to learn how to run `--help` or ask an AI how to not cry about burnout on open source instead? I don't get it, you should just be cruising on auto-pilot now.
The problem is retards that can only function on a cocktail of drugs, and as they were never good at anything other than anal retentive stuff built and continue to build these retarded systems. Those peddling RoR apps even when they couldn't serve more than 3 or 4 concurrent requests, JS backends to handle complex workflows that even after 2 years of dev. still have bugs and accrued a sprawl of crap to hide the issues of their own making, etc, and yet charge thousands of dollars, those that write shit software that's not even worth to clean your ass with, even though they have 20 years of experience, but then go give conferences and write books about their amazing architectural skills, those that write utils behind the "oh, it's open source, if you don't like it just fork it" and due to marketing get their crap everywhere, while making holes everywhere for their paycheques. Or the nepo babies that need their mexico border run to get their fix so they can have these "humanity changing" ideas? I bet they're the same that before would weasel a 2 week sprint to change the borders of a button. Or burn through 10k in meetings for irrelevant crap. Or get VC funding for a CSS styling company or a two prompt company. Or go on about the value of ideas, but then can't even get that going without outsourcing or an AI to help them have those same "ideas".
Ultimately, you just need to turn into a little pig and party in the pigsty, it's not that difficult either, they say pigs are very close anatomically to humans.
At least AI can help untangle the crap the anal retentive retards have built, and thank god, the pig-mor, this society can't even fuck to replacement levels (perhaps they'll manage now with AI).
I'm sure people do that, I myself do, but just a couple weeks ago Github issue navigation was broken, you couldn't get past page 1 by the UI, a couple weeks before that github pages domain associations were broken too, just right now I saw on upwork a job ad where they were saying "we use AI to speed our workflow and you'll need to too" and then the list of urgent fixes ranged from: forms on administrators pages not working, values are being misplaced, forms somewhere else not saving on submission, ui bugs with panels not opening, sections on pages X with bugs on... So on and so forth.
I mean, these problems were there before (careless development) and it's probably possible to stabilise at some point with proper guardrails, it could also be that the first version was done with AI at meaningless cost compared to hiring someone, but I am a bit weary of this extending to ever more areas of software production without an accompanying increase in "true" model capabilities.
The more it kinda "works" the less people will be incentivised to review it properly and understand it - there's already people saying the future is agents talking to agents - when I saw minority report it sounded a bit too far-fetched but it might not be that far-fetched after all.
This is interesting and should make you think but it's the same issue with infiltrators/agitators in other groups. If all it takes for a group to start resorting to violence and eventually establishing as part of their identity the killing of "others", when instigated by some authority figure/small group of instigators - specially with lies and historical distortions - this itself points out that the nature of that group is problematic doesn't it?
Specially if that group has a history with thousands of terrorist attacks across all the world, even when just counting the last 25 years. It's the same if you join a left-wing group that's "just demanding" some social justice, as an undercover paid actor, and then you steer it to start killing people who are deemed "against" their objectives, or the reason they need to act. Then some people go and said they were "infiltrated", but was it so? Or was it just waiting for some nut case to come forward?
If I tell you to go kill all the cats in the neighbourhood and you go, you're as or more dangerous than I am for suggesting that isn't it so?
While I'm sure Israel is no angel it doesn't seem to be the one putting criminals in hiding near schools and civilian infrastructure and I also don't understand what people want them to do? Have no country?
So others can then do with them as they please? Like telling them to "go back to their country". While it's obviously all part of a political ploy, jews should have a right to a homeland. No-one talks about Egypt and the other surrounding countries interests in this, in having a "Palestine" identity that didn't even exist a few decades ago. In fact, if you go see the first declarations, made by some Turkish guy, looking like the brother of Abramovich, sent to Palestine at the time, there's not even mention of anything Palestine.
And lastly, it's very rich to live in countries that have benefitted from this (oil) and follow ways of life that require this, while sipping on their $15 expresso's machiatos... So in fact, perhaps there's many called "humanists" in the west that have way more blood in their hands than they would like to understand, or pretend not to, while filling their wallets and living as good samaritans. Maybe they should globalize intifada next to their homes to see if they also like living like that.
This is not to say that Israel should have the right to bulldoze, kill and take lands of others for building high-rise apartment complexes but as far as I know and seems to be documented, it was the supposed Palestinian Authority's that many times refused land agreements and split-state solutions... But I guess here in this website where most were covid infected it doesn't matter.
You missed the point. Israel banks on making Gaza look like the enemy, seeks reasons to seek war, and bets on a hostile takeover of Gaza and expulsion of the people.
At this point, Hamas’s stories are meant to feed Israel’s hunger for annihilating Gaza. No stories of dictatorship or instigation or terrorism or hostage taking or can even measure back to the violent nature of Israel. It’s not Israel “is no angel”, but Israel is this earth’s worst demon who is trying to delegitimize the Palestinians, and you are their allied propagator.
> Israel banks on making Gaza look like the enemy, seeks reasons to seek war, and bets on a hostile takeover of Gaza and expulsion of the people.
You seem to be hanging all this on what was described as a "credible rumour". To some people the idea that Jews control the weather is a credible rumour.
Well, it seems you're the one missing the point. Again you show your nature when you make it about "Israel is this earth's worst demon", surprised you didn't say the jews. Luckily we have records of the peaceful religion and translations of what their head figures say and think, in many places around the world, throughout history, along with their acts, and luckily not only by christians or catholics (in the sense that that shows us without a bias), and still going on today. In fact we have more than 1.000 years of history to confirm. While the left and its love for erasing history (goes back to Stalin's photoshopping pictures before there was photoshop) to make itself look good goes very well with Islam's political games. I don't think Israel has to make it look like the enemy, muslisms make it abundantly clear by themselves.
You fallaciously conflated what was said with religion. Religion was not mentioned anywhere. This is politics and war. And so is selfishly based on personal vendettas and agendas.
does that really help? I have had llms fumble pretty simple stuff in html/css and probably that's what there's more information on the internet, while with simple packages/languages with clear cut semantics it seems to be much better, even if off some samples pasted into context.
[edit] the case was simply, write me an html landing page, with a navbar, a content area and a footer, make sure the content on each area doesn't extend beyond 1200px wide , make each container centered (paraphrasing)
What can happen is sometimes patterns that are unique to certain languages aren't surfaced and so can't be "learned" but that is a different case, since you can add examples - or patterns that go beyond syntax (threaded code, process isolation, interaction between different modes of resolution, etc) where you need the agent to be able to understand and plan higher-level logic.
I wrote CSSex as a css pre-processor (similar but not the same as SASS/SCSS) and my "boss" at the time fed the existing cssex files to GPT (+2 years ago) and it was able to write CSSex just as fine as if it was writing something that was present in its training data set. It never introduced syntax bugs, the bugs it introduced were all relative to complex cascading styles rules in an existing large (for a definition of large in webapps) codebase, rules interactions (CSS), browser quirks, and complex logic to do what we "humans" wanted (or him in this case) and so on.
reply