Can AI categorically not be a "real coder" or is the limitation in more trivial things such as scope and breadth? My experiences suggest to me that it is technically quite close.
I didn't want to get into the details, because I've already talked about BitGrid here endlessly, and was trying to stay on the topic of AI usefulness, but since you asked.
I'm trying to build a software stack that can eventually take something like a PyTorch model, and unwind everything, resulting in a directed acyclic graph of individual bit-level operations (OR, AND, XOR). That graph will then be compiled into a bitstream suitable for an FPGA-like substrate that eliminates the memory/compute divide, the BitGrid[1].
FPGA routing is a non-trivial problem, I'm hoping to get it down to seconds. I'm currently trying to build the software stack to make it usable.
The goal is to answer questions about BitGrid:
How efficiently can I pack a program into the hardware?
Is the model I've chosen for a cell optimal?
How many femtojoules per operation would a cell actually take?
If the answers are favorable, then in the deep (and improbable) future, it's possible that there could be a set of racks with an array of thousands resulting in a system that could stream ChatGPT at aggregate rate of a gigatoken per second, for far less than the Trillion dollars Meta plans to spend.
This isn't just some CRUD application with a web front end. There are a number of layers of abstraction at play, and the LLMs seem to handle it well if you limit the depth under consideration.
[1] BitGrid eliminates the traditional memory/compute divide that causes most of the energy consumption of CPUs, GPUs, and other accelerators. Even FPGA systems tend to focus on emulation of these models, and routing fabric for minimum latency, instead of maximum performance. Because all the active lines only reach nearest neighbors, power consumption for a given operation can be far lower than the traditional approach.
PS: I pay $10/month for GitHub CoPilot, which apparently now includes ChatGPT5
>I’m often asked, “what is the future of programming?” Should people consider entering software development now? Will LLMs eliminate the need for junior engineers? Should senior engineers get out of the profession before it’s too late? My answer to all these questions is “I haven’t the foggiest”
I just want to point out that this answer implicitly means that, at the very least, the profession is at least questionably uncertain which isn't a good sign for people with a long future orientation such as students.
Students shouldn't ever have become so laser-focused on a single career path anyway, and even worse than that is how colleges have become glorified trade schools in our minds. Students should focus on studying for their classes, getting their electives and extracurriculars in, getting into clubs... Then depending on which circles they end up in, they shape their career that way. The thought that getting a Computer Science major would guarantee students a spot in the tech industry was always ridiculous, because the industry was just never structured that way, it's always been hacker groups, study groups, open source, etc. bringing out the best minds
It has never been anywhere close to certain, we just had 20 years of wild, unsustainable growth that encouraged people to cover their eyes and pretend the ride would go on forever. 20 years of telling everyone under the age of 30 that of course they should learn to code and that CS was the new medical or legal degree. 20 years of smugly acting like we are the inevitable future when we are, in fact, subject to the same ups and downs as every other career.
The answers are even kind of contradictory. If the answer to "will we ever need software engineers again in the future?" is "we don't know" then the answer to "should I spend time and money to enter software engineering?" should be "no".
The requirement for a voting franchise was that one joined the Federal Service, of which the military was a very small part --- the protagonists best friend was working as a science researcher on Pluto and the young lady who joined at the same time might have become part of the Skywatch if she hadn't qualified as a pilot.
Nor was "you have to join the military to have basic rights" the premise of the book. One of the themes of the book (I wouldn't call it the premise, just one theme among them) was that to wield authority over others one must first demonstrate that they are capable of acting for the good of the whole even if it is not in his own personal best interest. Military service was one way, not the only way, to demonstrate that ability to act selflessly.
I think Heinlein actually has a very interesting point. To wield the power of the government (which is what voting is), it is important to be able to act selflessly. If someone can't do that, even for a couple of years of their life, why should they be able to wield that power over others? The universal franchise is not a religious dogma, it's good to ask these questions and think about whether our society could be better if we organized it differently. Unfortunately, a lot of people completely missed the point and just rounded it off to "Heinlein thinks the military should run society", which isn't at all true.
Heinlein isn't even saying anything completely new in starship troopers. It is in essence an evolution of the Active vs Passive Citizen distinction. Merely living in a society doesn't necessarily give all of the responsibilities of governing, aka voting. A citizen aught to have be an active citizen within the society to gain that privilege. The US Constitution was originally for only Land Owners (ignoring the other race and sex based stipulations). Heinlein is treating that active - passive distinction as being based on service instead of property.
> Nor was "you have to join the military to have basic rights" the premise of the book.
It was not even the premise of the film - only one right was conditioned on service, the right to vote (and possibly hold political office). I.e. actions that wield authority over others. I argue that is not "basic".
In the book it was said that if a blind deaf person in a wheelchair volunteered for service, the state would find something for them to do. Maybe tediously counting hairs on a caterpillar, or testing chairs in Antarctica.
Now for me the asspull from Starship Troopers that I still think about every now and then was the notion of mathematical proofs of morality at a high school level (or any academic level, for that matter). This was a society that somehow discovered provable objective morality, and I really wish that idea could have been fleshed out more by Heinlein.
It was one of his juveniles (what would now be a "young adult" novel). I don't think it was ever going to get fleshed out.
You see him put much more thought into political systems and economics in other books like The Moon is a Harsh Mistress, Stranger in a Strange Land, or For Us, the Living.
This [1] lays out pretty conclusively that even though the book plays lip service to "federal Service", the entire rest of the book makes clear that it is only military that get the right to vote, not "government workers".
I don't think that is correct at all. I read the book and it was very clear to me that it wasn't just lip service, you could do your federal service in many different ways.
I felt the opposite when I read it, which led me to looking around and finding that essay. It felt like he said you could do any federal service, but no federal service actual existed outside of the military. I personally think this is more of the case of the author intending one thing and writing another. Clearly Heinlein intended for it to be about Federal service, but over and over in the actual details, there just is not a Federal Service. It is only the military. That is my main issue with the book, it proposes one society, but it builds a different one. They tells us you can join Federal Service, but those people don’t exist anywhere in the text. Not current Federal workers or former ones, now citizens.
It's not great, but worth recognizing that it's a notion as old as democracy itself. See ancient Athens, Greece. Among the many other requirements that made voting exclusive, only men who had completed military service were granted the right to vote.
I think it would be interested if only those who had completed military service could vote to go to war since I think you'd def see less of them. That said, I don't see this actually happening.
Fracturing a population is a time honored way to maintain oppression.
For example, let’s say you want to prevent wealth redistribution, but there is lots of popular support for it. Then you give it to some, for example old people, and it removes wealth redistribution as a priority for them.
Now, you can more easily withhold the benefit from the remaining population. Or give them all various tiers, such as white collar and government workers with more employer subsidies, and lower paid workers with few or no subsidies.
At the end, the opposition group will waste a ton of energy trying to make their cause a priority, and will probably never achieve the full goal. The more politically important will get theirs, and the less politically important will suffer without. That stepping stone becomes a barrier.
In the example, you can change age to skin tone, gender, religious tribe, military status, and even arbitrary geographic boundary. See the recent tax benefits Alaskans received in exchange for their Senator’s vote.
That's not what's happening here. Being in the military won't magically give you a right to repair your iPhone. This law puts a requirement on DOD procurement (beyond what's already present, as I mentioned in another comment) to acquire tech data (or whatever's required) to make purchased systems maintainable by the military itself. So you buy a Generac for your house and you're in the military you get the same manuals the rest of us get. If the DOD buys a Generac for a facility, they can get more data.
React isn't the problem, the problem is literally every dev I've ever worked with who writes react good except for one or two don't understand or care how to write performant react code at all.
I don't think it's unreasonable or onerous to shift the burden of hiring onto companies well-suited and minimally impacted by this process compared to individuals.
The fines should definitely be proportional though, with larger companies facing very severe infractions.
There are rules like this in other countries around the world, and the impact is that it's much harder to change full-time jobs, because companies work around them by replacing full-time roles with contract positions, something that's much harder to regulate.
But the big thing here is: obviously there's a cheering section for any rules that make things harder for hiring managers, because most people here are on the other side of that transaction. Ok, sure, whatever. But none of this has anything to do with the "ghost job" phenomenon, where job postings are literally fig leaves satisfying a compliance checkbox so that roles can be sourced to H1Bs.
I completely agree that there isn't any legislation that will "fix this problem" (other than perhaps abolition of H1B).
If it helps soothe the feelings of those who think hiring managers are demon spawn out to get them, I can add, as one of those who spent a whole season this year doing basically nothing but conducting first-round interviews all day, to hire four open SWE seats, it's misery out here for us, too.
It was a mix of laughably unqualified people; people in India lying and pretending to be in the US ("Which neighborhood in NY do you live in?" - "I live by the Statue of Liberty"); people sending ringers to do their coding exercises for them including different ringers or two ringers at once (oops); and entirely made-up resumes (we have your resume from an application 3 years ago but your entire life history has changed since).
> people sending ringers to do their coding exercises for them including different ringers or two ringers at once (oops)
Found this part mildly fascinating. That it's more lucrative for a significant portion of the population to be so highly skilled at programming that they can regularly serve as "ringers" to falsify entrance exams, than it is to simply complete such an exam and get a job in America. Must make great money as exam falsifiers.
Quick check on Google says it runs upwards of ₹50,000 ($570) for valuable falsifications (data is mostly from Indian exam falsification though). If you can actually manage to get one per day it's $200,000 USD in India.
It's not being skilled at programming, it's being skilled at interview programming, which is very different.
Also ringers are black market, so you can be hired everywhere and not blocked by not being in the USA in the first place. You can pass the interview, but you can't find a job that hires you in the USA because of non-interview reasons or market conditions. Anybody could become a ringer because it's a purely skill based job you can self teach, while getting a job at an Indian big tech might require a degree from an ITT or even just a degree at all might be too much for you.
Honestly they would be a great hiring pool on some level if the reasons why they can't get hired had nothing to do with skill but more socioeconomic barriers like that.
I've never hired a ringer, been a ringer or even looked into it, so it's total conjecture on my part.
The topic set is very different and the conditions are very artificial. There are many engineers who are amazing, but are absolute shit about writing code, under pressure, while someone is looking over their shoulder, to a leetcode medium that they haven't practiced for. Their brains go blank.
Like seriously, go to leetcode, pick a random medium+ problem and solve it within 15-30m without bugs under pressure, on coderpad, without code execution or being allowed to look up syntax. Make sure you've never seen or been asked the question before. Now do 10 more. You'll see it's something you need to practice for specifically. Make sure you throw in a few binary search questions while you do it.
Anybody going for a leetcode interview without doing prep work first is not a good candidate to start with.
I coached a newly minted PhD engineer into spending 3 weeks studying leetcode prep materials before the interview, telling him that per-hour it would be the best investment of his career. Shockingly, he listened, studied for 3 weeks, and absolutely nailed it. And landed the big bucks job, too.
> Their brains go blank
Anyone with a degree has undergone exams that were critical to graduation. I know all about time pressure, high stakes exams. After all, I attended Caltech. I learned how to deal with them. Study the material beforehand, work all the problem sets beforehand, make sure you have it cold backwards and forwards. That's the cure for brain freeze.
Going in without prep for a $250,000 job interview is just lazy, no matter how smart or capable you are.
A crackerjack programmer that cannot learn leetcode is not a crackerjack programmer.
Just for fun, I found among my dad's papers the written tests he had to pass before he could even sit in the cockpit of an F-86 Sabrejet. They were quite comprehensive and full of minutia. For damn sure the AF wasn't going to let anyone fly one of those monsters without proving they could study and learn everything about them by heart. No excuses about brain freeze is going to fly (pun intended).
One day, he was flying along in his F-86 over the Arizona desert when the engine conked out. He radioed the situation to the tower, who advised him to bail out. But he knew how to figure out hour far he could fly given his speed, altitude, weight of the fuel, wind speed, etc. and calculated that he could make the strip. And he did, with a few feet to spare.
There's no way I would even dare fly one without mastering all that stuff, either.
> calculated that he could make the strip. And he did, with a few feet to spare.
Stories like that give me goosebumps. Your dad has (had? if so, I'm sorry) just absolute balls of steel. Not to mention an incredible mind. Thank you for sharing.
Yes, because flying a war jet under pressure and landing it in the desert is the same as writing software over 5 months in an office. /s
Test anxiety is real. Anxiety disorders are real in general, not to mention ADHD, autism and more where these disabilities can interact in bad ways in interviews. It does not mean that people with these issues are bad engineers. If your instrument is inaccurate, it does not mean the thing it is measuring is wrong.
One of the best and smartest engineers I have ever hired had visible anxiety and objectively did not pass the interview but we pushed for it anyway and was hired as a jr engineer. He is a staff engineer now, leads various large company library projects and is the go to expert about various systems in the entire company and will probably become a sr staff engineer. He also probably hasn't changed jobs because he knows he's bad at interviews, which is incredibly sad.
No, that optimizing and interviewing for performance under pressure is not needed for most software engineers, while it makes sense for fighter pilots. It's is definitely more stressful.
I'll repeat that the cure for test anxiety is preparation, preparation, preparation. Avoiding it, like your engineer employee, won't be helpful. Keep going to interviews, and the performance anxiety will also dissipate.
The first few times I did public speaking, I'd freeze up and squeak. I speak regularly now, and it's not a problem anymore.
(Again, I attended Caltech. Most of us were nerds, and many Aspergers. I have worked with many people "on the spectrum" and am not at all ignorant of their differences.)
When I worked as a firefighter/paramedic, one of the axioms was that you don't train until you get it right, you train until you don't get it wrong. That, at 3am, sleep-deprived, it's so ingrained in you what to do, you can do it automatically, leaving your mental capacity for the variables of the situation, not the bare minimum of "what do I do next?"
> because companies work around them by replacing full-time roles with contract positions, something that's much harder to regulate.
Yes, then a regulator sniffs on that, company is unable to prove absence of employment-like relationship, then is fined and owes backpay on all the unpaid taxes with interest.
>But these fears are frequently based on oversimplified or misapplied interpretations of economic models, and appear to be driven more by political agendas rather than evidence
A razor for people who think like this:
Do you think Japan and Japanese people would have a higher or lower quality of life if they had accepted unlimited sub-Saharan Africa immigration?
This should separate serious people from the not so serious people pretty rapidly.
It's not directly related to your post, but Japan has a very significant Brazilian population that live in very specific parts of Japan. Toyota[1], and its prefecture Aichi, are known for having a Portuguese-speaking population and they are, of course, working as cheap labor in automobile industries. In fact, some signs on the road are written in Portuguese, too.
There's a significant number of Japanese in Brazil, and a large portion of the Brazilian population is of Japanese descent. They're called the Nikkei Japanese, and they're predominantly ethnically Japanese.
I wonder if there's the stat somewhere but I'm sure a sizeable portion is at least half Brazilian so it makes them a "half"—as Japanese like to say—which they don't tend to regard very highly. It is indeed a marginalized group in Japan and probably why there is such a big community in that area.
Since Brazil has a very significant Japanese population, that makes me wonder: Are these Brazilians in Japan ethnically Japanese or of other ethnicities?
How? OLED is still the gold standard while MiniLED is a rebrand of full array local dimming LCDs where the zone count is somewhat higher than past models. MicroLED is the true replacement for both. I do not see how MiniLED is on track to outpace anything.
I'm just referring to smaller and smaller LEDs, I don't really care about the marketing terms really. Whatever technology you want to call sufficiently small LEDs, how you market it isn't relevant to me. Pixel-LEDs are the goal.
reply