I worked in this field since long before LLMs. Nobody outside of the field really cared about GPT2, and even insiders knew the "too dangerous" part was a PR gag at best and the first dig of the moat at worst. After all, they released smaller versions of it along with detailed instructions on training it in the paper, so anyone with a lot of compute and a bunch of internet scrapers could try to recreate it. But basically noone did, even though it would have only cost ~50k back then (and less than 3k today). A few normal users started to take notice with GPT 3, but even then it was super limited. Even instructGPT didn't cause real shockwaves, despite being very close to the final product. Only ChatGPT/3.5 finally lit the fuse and people suddenly cared about having this too.
Since we’re doing anecdotes I definitely agree GPT2 lit the fuse. It woke up a sizable chunk of people paying attention. GPT3 is when I and many others got into a full blown existential crisis - it was the bang after the fuse. Then we got a long tail of laggards and people without vision. Even today you can find a significant chunk of folks in denial still.
I’m conflicted after reading this comment, because I think I would be that personality in my workplace, largely because I believe that’s the only sane position to take as a worker with ~0 power over the decisions made that can entirely destabilise your life.
On the other hand, my priority isn’t maximising my personal career benefit, but the collective benefit of my team, so I suppose I either see it more as a 2v1 sorta game, or perhaps my “player” is an amalgam of myself and my teammates. Playing this way, outsourcing everything you do to an LLM is the worst move, because you lose the touchpoints that tell you where the friction is in your team.
I think everyone should be looking to balance their work effort against the payout of the job. They should also be changing jobs when the effort to reward ratio starts to become unfavorable compared to other jobs on the market.
The problem with the personality above is that the person isn't playing like a team (like you said) but as an individual maximizing their own visibility while loading their coworkers up with the review effort. They found an asymmetry to abuse (they generate text easily, coworkers get a lot of extra work to review it). They don't care what it costs their coworkers. They just like that it makes them look good.
Whenever I try to articulate this issue to people during more casual AI discussions, I always refer to “study guides” in college.
I don’t know how many of y’all did these, but I’m sure I wasn’t the only person. At my undergrad it was very common for a group of students to all to get together, compare notes from lectures and readings, and basically come up with a group study guide of sorts. People were given specific sections to share, you didn’t just send all of your notes - usually 2 people per section’s take on that portion. You could always tell who just copy and pasted their shorthand (usually indecipherable) and who actually took the time to edit it/clean it up. This was at a time when almost everyone did it on laptops.
The people who took the time to make their portion(s) digestible for others were asked back, the others weren’t.
> They should also be changing jobs when the effort to reward ratio starts to become unfavorable compared to other jobs on the market.
The problem here is that all tech companies look alike. Take for example the interview process (copied by almost any company out there that thinks they are google). Another example: the under/meets/above expectations BS. And now the most recent example of “token usage as sign of productivity”.
So, it’s getting tremendously difficult to simply switch jobs that offer something different
My experience couldn't be more different. The tech companies I've worked for in the past 10 years have been so completely different from each other, from interview process to company culture, that I can't agree that all tech companies are the same.
You can also look to change to different roles (product management, even sales) or jump to a different career completely.
There are options if you look. You're not going to find a dream job that pays $600K for 4 hours of no-pressure work per day and perfect coworkers, but there are a range of job options with tradeoffs along the compensation-effort pareto front.
> For human code review requests, I always review my AI-generated code first.
I remember a time in the ancient past (2025 maybe) that your PR was your responsibility, whether or not you typed it with your meat fingers or cranked it out of the Giant Plagarism Machine. It’s absurd to think that the above quote is now something approaching controversial.
I don’t think this analogy holds. The whole way through the processing pipeline in the brain, different sensory data is ingested separately and processed separately; and we still don’t understand how that data is then integrated into a cohesive experience.
LLMs have the same fundamental input regardless of modality, tokens. There is a preprocessing step before the “brain”, which is more akin to some super-synesthesia where all senses are translated into sound before becoming experience.
Can't you say the same about the connectivity between the brain and your senses? Your eyes do 'preprocessing', but in the end the connection to your brain is just through electrical impulses in the end. All senses get translated to some sort of electrical signal, just like in an LLM with tokens.
Believe it or not, that other $60b isn’t one chunk of money, but other similar chunks of ~$20m that’s probably being spent in similarly idiotic ways. If this were an article about those, yes I probably would.
I get the joke, but CSUs are fairly inexpensive as far as universities go. Tuition at SJSU is about $9500 per school year for California residents. That's a year of education for almost 1700 students. It may not seem like a much money to some, but it certainly covers a lot.
The idea that AI is somehow at fault for the absolute fiscal disaster the UC and the CSU systems find themselves in is laughable at best and damaging at worst. These systems (and I say this as a graduate of UCLA that was on a full academic scholarship) have been taken over by parasitic administrators and bureaucracies-on-top-of-bureaucracies that have milked not only the students, but also the taxpayers, completely dry. Tuition has consistently gone up since the 70s, while housing, facility, classroom quality have all gone down.
It's been literally the biggest grift of the past 50 years[1]. Education should be free.
In real terms, tuition fees in public universities peaked in the early 2010s. They have not kept up with inflation since then. That explains a large part of the fiscal disaster.
Can you source this? My cursory research shows the opposite[1]. Imo, the fiscal disaster is in part due to enrollment declining (which, ironically enough, mainly affects low-income households).
Declining enrollment affects universities very unevenly. University of California is still under pressure to increase enrollment, which is mostly constrained by physical capacity. The housing situation is particularly bad on some campuses.
Amusingly, education is free and I’ll die on this hill. There is nothing you learn at a university that you cannot learn, for free, at a library and online.
> Amusingly, education is free and I’ll die on this hill. There is nothing you learn at a university that you cannot learn, for free, at a library and online.
There exist parts or even degree courses in university education that cannot really be learned this way. Think of laboratory courses or courses where you need access to expensive equipment.
Also, there exist topics and degree courses that are much harder to learn by yourself than others.
Finally, keep in mind that computer science is "special" in the sense that:
- What the university teaches you or should teach you (a degree course at a university rather prepares you for an academic career in the field) makes you quite overqualified (in the academic sense) for many programming jobs. Such topics are possible, but in my opinion far from easy to learn by yourself.
- Many employers want very different skills from applicants, which often involve "fashionable" skills with a very short half-life. A university system is likely not the best kind of education system to teach this kind of skills: it rather (ideally) excels at teaching topics that are complicated, but have a much longer "half-life" before becoming outdated.
> Free AI like ChatGPT can assist with offering many different explanations personalized for someone to make it easier to learn.
What I can tell you is the following: a lot of academic topics are quite subtle - to get to more than a basic level, you have to learn things that are very subtle, and where you only can judge the correctness of the information years later (basically when you have finished your degree or even PhD).
Because of this, I would rather read the most renowned (and ideally hardest) textbooks in the respective area (if you really need to cheap out, download them at some shadow library) instead of trusting some AI.
I can tell you that for quite a lot of questions in my area of expertise, the answers that AIs gave were far from being sufficiently reliable for learners who want to get a deep knowledge about the topic, and the errors were often quite subtle.
In mathematics, for example, it is not uncommon to hang for hours over a page or even a paragraph, trying to understand why the statement holds - and this in a situation where the proof is for sure correct. Now imagine the situation of hanging over a page of text that you will need hours for understanding when you cannot even rely on the prior that the information in the text is correct ...
> Now imagine if AI can explain that page better so someone can understand it in a minute.
The problem is the "understanding" part: from my experience oneself is the huge bottleneck here: one realizes very fast that the lacking component is one's own brain capacity.
This is also why many mathematicians and physicists are so obsessed about IQ: mathematics and physics are disciplines where IQ points really can give you quite an advantage.
So, the really helpful thing to ask to the AI for is not better explanations, but methods for getting a huge increase in brain capacity.
Because of all this, your point is rather a mixture between a nice science-fiction story and a marketing pitch for an AI company.
The information is how to use a lab, so you can do research, you know, the thing that happens largely on university campuses. (Now why taxpayer funded labs end up patenting things for private corporations, that’s what’s peculiar to me!)
Another example is history. It's theoretically possible to become an academic historian through private study and there are certainly no legal barriers to it, yet amateurs almost never make the transition except through higher education.
> These systems (and I say this as a graduate of UCLA) have been taken over by parasitic administrators and bureaucracies-on-top-of-bureaucracies that have milked not only the students, but also the taxpayers, completely dry.
That quote shows an utter disregard for basic human decency.
It is the responsibility of the person running the coding agent to make sure the resulting PRs are high quality. Putting that on your team mates, or worse, random open source project maintainers on the internet, is the definition of an extractive contribution.
> It is the responsibility of the person running the coding agent to make sure the resulting PRs are high quality.
And
> he's proposing a method for how to do so using agents
Are not in agreement. The claim being made is that you shouldn't be sending PRs you haven't personally vetted to be high quality. Definitionally a bot cannot be used to personally vet something.
This is not a contradiction; it's an augmentation. As an operations guy, I can tell you that well-constructed automation to reduce the amount of manual checking a human has to do almost always increases the quality of the overall process's output.
As a software developer, you must never subject your team mates to a PR that you yourself believe to be low quality. The point of code review by others is to catch things that you missed.
There are multiple lines of defense for quality. Yes, automation can and should be one of them, but your own self-review always has to come before review by your team mates.
And for a dev, that's essential professional ethics, and good personal pride as a craftsman.
However, from an operations perspective, a dev is a piece of the qa pipeline with a nonzero error rate, and an optimal throughput rate, above which that error rate rises dramatically.
As a dev, you'll never merge a bad PR; in ops, we want to help you with that goal, and also have plans for what happens when it fails.
They are probably reacting to the laughable idea that by making PRs 20% better (or whatever), devs will continue to review the code with sufficient rigor to catch even the bugs they're supposedly now preventing. Assuming such rigor was ever present in their work!
Put another way, who are they supposed to hire to tell these low quality PRs apart from the high quality ones? Who even knows how to do something like that?!
I really have to wonder what can you use 10G for? I have 500M down from my ISP, and it is faster than I can imagine ever needing, unless I get into data-hoarding 8k movies.
My homelab has a 10G fabric (switched) for NFS, iSCSI, NVMe-OF, etc. and a 25G fabric (a mix of back-to-back and switched) for clustering (Ceph, DRDB, ZFS replication, migrating VMs).
I spun up some iSCSI-backed SQL Server a few months ago and 10G couldn't keep up with the workload, so I dropped in a pair of 100G ConnectX-4 cards with iSER (iSCSI Extensions for RDMA) support for that particular use-case.
Just because your uplink is less than 10G doesn't mean the rest of your network can't be a bit more capable. :)
True, I don't really feel limited by my existing 500Mbps down, but knowing I'll be having 2500Mbps up/down soon means I want to have the infra to handle it.
Basing things on 2.5GbE would certainly have been cheaper but some things don't support it (they either do 1GbE or 10G SFP+) so settling on 10G where possible made more sense to me. My future ISP also has a 5Gbps up/down option, but even I can't justify that right now.
My wife and kid just want their phones/laptops to work, and to be able to stream stuff to watch, they don't care about the underlying speed.
Having a faster network may make some of my work related things run a bit quicker. A few times a day I'll need to pull something big down (either an ISO or a bunch of docker images) and that can take up to 2 minutes with 500Mbps down. Having those take a fifth of that time will make it seem less of a roadblock to doing work. 2 minutes meant I went and got a cup of coffee and often got more distracted, 30 seconds should keep me at my desk and focused on what I was doing. That's not a big enough reason to justify it on its own obviously.
I also want to do offsite backups with/for various family members, so something better than 75Mbps up is going to be a huge boost. Getting 1Gbps+ out will be huge (assuming whatever is at the other end can support that).
I don't do any kind of data hoarding, I think I've got something under 4TB of data that I actually care about, and most of that are family photos/videos.
Deep down it's mostly because I'm a networking geek so it's fun to play with some new kit and make blinkenlights.
Going for a cup of coffee means physical walk. Detaching from focussed mode means your mind gets in diffused mode. This is where/when creativity ensues.
One thing to remember is 2.5 gbit/sec uplink is shared between all clients. So if one client is on 1 gbit, and one client could saturate their 1 gbit while switch and router can handle better. An advantage of that is QoS isn't needed to be applied manually.
So, for example, it maybe worth it to have higher than 1 gbit uplink on switch to router, and maybe a server to switch, but devices such as your TV or WLAN clients do not need such.
75 mbit up is pretty good compared to DSL (I bet it is cable), and yes 1 gbit up is nice for off-site backups. But the upsell of going above 1 gbit symmetric IMO isn't there.
Cable providers know this. Which is why they sit below 1 gbit symmetric, at a level average subscribers are comfortable with.
> Going for a cup of coffee means physical walk. Detaching from focussed mode means your mind gets in diffused mode. This is where/when creativity ensues.
Sure, but I want to choose when I do it, not have it forced upon me.
> 75 mbit up is pretty good compared to DSL (I bet it is cable)
It is FTTP not DSL or cable. BT Fibre 500 in the UK. Almost all of the deals through the legacy/monopoly provider (BT/Openreach) are asymmetric like this.
The 2500/2500 at the new property is a different provider that has their own network and so isn't tied into reselling Openreach's GPON infra.
It's less "what new thing can you do" and more "what things involve noticeably waiting, how long is the waiting, and what else is impacted". E.g. updating a game on Steam practically takes slightly under half the time for me (1.2 Gbps actual rate) and has absolutely 0 impact to any other traffic in the house. If it was 10x the price to get 10x the bandwidth I wouldn't bother but it was actually about the same as my old cable modem plan.
It just makes everything feel faster. I went from 500m to 2.5g thinking I would immediately go back (I really just wanted the upgrade to XGS-PON to run my own network) and then I couldn't go back. Its very much like using a higher refresh rate or a faster CPU...
I went from one dev machine to two at my desk so I connected them via 25GBe. With about 2.8GBps TCP throughput and RDMA available I don't have to think too much about task placement or cross-traffic. (specific hardware: Mellanox ConnectX 4 LX cards + a DAC cable).
For most people, 500M is probably fine. But once you have a few family members, each streaming 4K movies to their devices, and a parent that needs a video call to work seamlessly, you start to see the benefits.
10G is probably overkill, but it's also future proofing. The way things are going, loading the NYtimes will require 10G just for the advertising alone...
reply