HNHacker News
TopNewBestAskShowJobs

morleytj

519 karma · joined February 29, 2024

https://tjmorley.com/
submissionscomments
morleytj··on GPT-5: Key characteristics, pricing and system card
I wouldn't say model development and performance is "leveling off", and in fact didn't write that. I'd say that tons more funding is going into the development of many models, so one would expect performance increases unless the paradigm was completely flawed at it's core, a belief I wouldn't personally profess to. My point was moreso the following: A couple years ago it was easy to find people saying that all we needed was to add in video data, or genetic data, or some other data modality, in the exact same format that the models trained on existing language data were, and we'd see a fast takeoff scenario with no other algorithmic changes. Given that the top labs seem to be increasingly investigating alternate approaches to setting up the models beyond just adding more data sources, and have been for the last couple years(Which, I should clarify, is a good idea in my opinion), then the probability of those statements of just adding more data or more compute taking us straight to AGI being correct seems at the very least slightly lower, right?

Rather than my personal opinion, I was commenting on commonly viewed opinions of people I would believe to have been caught up in hype in the past. But I do feel that although that's a benchmark, it's not necessarily the end-all of benchmarks. I'll reserve my final opinions until I test personally, of course. I will say that increasing the context window probably translates pretty well to longer context task performance, but I'm not entirely convinced it directly translates to individual end-step improvement on every class of task.

morleytj··on GPT-5: Key characteristics, pricing and system card
It's cool and I'm glad it sounds like it's getting more reliable, but given the types of things people have been saying GPT-5 would be for the last two years you'd expect GPT-5 to be a world-shattering release rather than incremental and stable improvement.

It does sort of give me the vibe that the pure scaling maximalism really is dying off though. If the approach is on writing better routers, tooling, comboing specialized submodels on tasks, then it feels like there's a search for new ways to improve performance(and lower cost), suggesting the other established approaches weren't working. I could totally be wrong, but I feel like if just throwing more compute at the problem was working OpenAI probably wouldn't be spending much time on optimizing the user routing on currently existing strategies to get marginal improvements on average user interactions.

I've been pretty negative on the thesis of only needing more data/compute to achieve AGI with current techniques though, so perhaps I'm overly biased against it. If there's one thing that bothers me in general about the situation though, it's that it feels like we really have no clue what the actual status of these models is because of how closed off all the industry labs have become + the feeling of not being able to expect anything other than marketing language from the presentations. I suppose that's inevitable with the massive investments though. Maybe they've got some massive earthshattering model release coming out next, who knows.

morleytj··on At 17, Hannah Cairo solved a major math mystery
In the US a doctorate usually doesn't require a master's, most people go straight from undergrad to phd.
morleytj··on So you wanna build an aging company
I was under the impression that the majority of recent analysis pointed to the interpretation that most of the claimed blue zones were primarily marked as such due to poor record keeping rather than true super ager status.

Is Sardinia an exception to this?

morleytj··on Backyard Coffee and Jazz in Kyoto
I've thought the same thing. Would love if other people were into that.

I'd ask you to consider the following thought experiment. If the regulatory barriers are lower to establish a nightlife business establishment without alcohol as an option, and a large number of people want to do activities without alcohol, why do these businesses not exist in significant numbers?

It's certainly not a novel concept to open a cafe or a late night board game location. If they were sustainable businesses economically, which barriers currently exist that prevent them from being present in cities to the same extent as bars?

morleytj··on Backyard Coffee and Jazz in Kyoto
Businesses that sell alcohol make a lot of money, businesses that don't tend to go out of business.

Wish it weren't so (I don't drink alcohol, personally), but that's the economics of it.

morleytj··on Backyard Coffee and Jazz in Kyoto
A similar situation in the US can be seen in Boston. Historically terrible nightlife and for easily explained reasons. Liquor licenses are distributed by the state at a capped amount that can potentially be increased each year, meaning the majority of new businesses wanting to have alcohol sales will need to purchase an existing liquor license from another business, often at an exorbitant price (over 500k USD on average I believe)

This makes it extremely difficult for any new businesses to start, and massively advantages large chain businesses that have the ability to make the initial investment in securing a license, versus small or quirky businesses which just have no chance getting started.

morleytj··on Should we design for iffy internet?
Really interesting perspective, thanks for sharing.

I think in so many fields the datasets are by far the highest impact thing someone can work on, even if it seems a bit mundane and boring. Basically every field I've worked in struggles for need of reliable, well maintained and open access data, and when they do get it, it usually sets off a massive amount of related work (Seen this happen in genetics, ML of course once we got ImageNet and also started getting social media text instead of just old newspaper corpuses).

That would definitely be advice I'd give to many people searching for a project in a field -- high quality data is the bedrock infrastructure for basically all projects in academic and corporate research, so if you provide the data, you will have a major impact, pretty much guaranteed.

morleytj··on Should we design for iffy internet?
This is a huge issue for me with a lot of sites. For whatever reason I've spent a lot of time in my life in areas with high latency or jist spotty internet service in general, and a lot of these modern sites with massive payload sizes and chained together dependencies (click this button to load this animation to display the next thing that you have to click to get the information you want) seriously struggle or outright break in those situations.

The ol reliable plain HTML stuff usually works great though, even when you have to wait a bit for it to load.

morleytj··on FSE meets the FBI
The article technically does have a TLDR in the second paragraph, though not directly labelled.
morleytj··on What was Radiant AI, anyway?
One of the classic emergent behaviors (which I loved) that came from Dwarf Fortress was the cats dying of alcohol poisoning.

There was a patch note years and years ago about a bugfix that had to happen because all the cats were mysteriously dying in people's fortresses -- it was tracked down to the fact that cats would walk through the taverns, in which visitors would be drinking and occasionally spilling alcohol (There was a feature that had been added at one point for spilled liquids to form pools which could get on entities passing through, like getting mud on you by walking through a large puddle).

The cats also had a piece of functionality where they could self clean by licking after they became dirty with something, and would ingest some amount of it due to using their tongue to clean themselves.

The cats' fur would become damp with the alcohol as they stepped through the spills, and the cats self cleaning meant they would regularly get extremely drunk and die from the alcohol poisoning trying to clean themselves. Not intended at all, but two completely different systems colliding in an emergent behavior of interest.

I also always loved the behavior of undead zones. In them, any dead creature could be revived by dark magic in the area. It leads to the question though, what counts as a dead creature? Well, it would be anything with a tag indicating that it came from something that died. This does in fact include small bones though, or hair from a butchered animal. Fortresses in these areas would have quite the casualty rate trying to butcher a pack mule as its hair would come to life and kill the butcher.

morleytj··on How to live an intellectually rich life
The Sound and the Fury is an incredible piece of art with a beautifully structured narrative, in my reading of it. Why do you say he's a terrible writer in your opinion? Who would you rank higher?
morleytj··on OpenAI releases image generation in the API
It's probably horrifying!
morleytj··on Harvard's response to federal government letter demanding changes
Who hates you in particular? What do you mean by that? Also, that's fine to have a conversation about funding, but it's fairly different from arrests, deportations, and shipping individuals to foreign prisons. Rather motte and bailey to earlier suggest that the government was going to go "Henry VII" on universities, and then say that you just want to change funding. These are very different positions.
morleytj··on Harvard's response to federal government letter demanding changes
So you're fine with them arresting dissenters as long as you disagree with the dissenters? That's fairly antithetical to the ideas expressed in the US constitution.
morleytj··on Cottagecore Programmers
This is a really beautiful sentiment, it's interesting to look more at our careers as stewards of a lasting organization in some ways, where every iteration of some product is dependent on the existence of work done before. Even with ephemeral software we all leave our mark in that way. Thank you for writing that.
morleytj··on Cottagecore Programmers
Now that you mention it, I have always enjoyed that general feeling of resourcefulness. I've talked about this with other people, but given my fairly eclectic background I usually feel like I can handle a pretty wide variety of problems. It's a very centering feeling, so it would make sense as to why other people worry about being overinvested in just one skill.
morleytj··on Cottagecore Programmers
Will do! :)

Ascutney is great -- I definitely recommend starting from the Brownsville trailhead, there's an oddly beautiful view partway up where the abandoned mining operation has some remaining tracks leading out along a thin little cliff and you get an incredible view of all the fields and forest because of the rocks that blocking the trees from growing beneath the cliffside.

morleytj··on Cottagecore Programmers
That's my bad! Thank you for the correction.

It is an incredible comment, haha, as you can tell I still think about it semi-regularly. Incredible that the issue is still open as well.

Best of luck with the dining table! I took a good look through your site while I was looking back up that thread that I remembered and your pieces look great. Actually, if you're in the Upper Valley, you probably aren't far at all from where I grew up. It's over near Windsor, around Mt. Ascutney.

Edit: If you or your partner like knitting, I'm sure I could prevail upon my parents to offer some of our yarn to you as a gift for the inspiration!

morleytj··on Cottagecore Programmers
I'd never heard of that, that's an extremely cool connection! I'll check out acoup.blog now, it sounds intriguing. Thank you for telling me about it!
morleytj··on Cottagecore Programmers
Your point is a bit strangely stated. If just doing physical activity and sensory engagement is all you think a person needs to balance, why not just lift weights or go for a run? Why fantasize about milking a cow? There's clearly more to this, in my opinion.

And no real insult taken if you see me as being pretty verbal, I am, though I can rotate a shape too as evidenced through my job.

morleytj··on Why do programmers want to quit programming and start a homestead?
RIP, I didn't realize there was a flamewar detector, I probably shouldn't have replied to so many comments, haha.
morleytj··on Cottagecore Programmers
I've definitely struggled with that feeling in my work as well.

When I was growing up, the metric for having done something were things like: All the hay is baled and we moved it all into a loft, or we finished shingling the roof.

Working in a technical position, it feels much vaguer. Maybe I finished a project, but then again, maybe there's a bug, and I'm back working on what I thought was done. Maybe I made a tool accessible online, but now there's a new tech stack and my tool is irrelevant unless I update it. It feels very Sisyphean at times I suppose, in a way that making a physical product and then seeing someone use it or eat it really doesn't.

It's interesting that it feels like that though, because I certainly didn't have to throw hay only one time, and I didn't have to scrape paint from a wall only one time. I suppose I'm interested in why it feels more real and less repetitive, and maybe part of that is what you point out, that people outside of the field just don't understand what the work that tech workers do consists of in actuality.

morleytj··on Cottagecore Programmers
Running a farm is a ton of work, exactly. The difference of having the ample injections of cash and not having them is pretty huge, especially when it comes to how common the issues that pop up when trying to run a modern farm are, and how expensive they can be.

I think when you're someone who grew up with exposure to the lifestyle of farming, it gets easier to see that the escapism is possible because of how rare it is for people to interact with people whose main employment is farming on a regular basis.

It is honestly pretty interesting from a historical perspective to think about what this means as a shift in the populace's opinion towards certain kinds of work, because we're really entering unforeseen territory in US history where no one will even really understand first-hand what a version of the US where the vast majority of humans living there are engaged in agricultural labor on a regular basis lookd like, if that makes sense.

morleytj··on Cottagecore Programmers
That's fair, I think there are some people (like that initial hn post I reference) who such as yourself really are just people who really enjoy that sort of lifestyle. But I also think that societally we put a lot of value and for lack of a better word, "coolness factor" on manual labor. You could imagine in certain time periods and cultures that something like working in a field wouldn't be viewed positively at all, and maybe something like writing poetry is seen as a very masculine and rugged endeavor.

I think that for certain individuals that sort of mythological status ascribed to being a frontiersman or a person who can do everything themselves definitely influences how they perceive a potential job or method of living.

Thanks for reading though!!

morleytj··on Cottagecore Programmers
It depends on the job for sure, but you're certainly not wrong that many technical jobs have a negative societal impact. I had a hard time finding jobs I actually wanted to apply to after college because I strongly disliked the idea of working for a company where the sum total of my contribution was making people more likely to click an advertisement or something like that. Felt pretty awful in terms of "meaning" in that way.

But there are a lot of tech jobs that aren't like that as well, and I think people have similar feelings about wanting to escape into a more natural world in those ones too. The sum total of which are positive and which are negative is certainly up for debate though. I actually have another post I started writing about the negative second-order effects of certain outcomes of software development in the last decade or so.

morleytj··on Cottagecore Programmers
I think the dual interesting aspects for me are as follows:

Why is farming or woodworking seen as less alienating than being at a computer?

And why does it feel so impossible for us to form the communities that would give us the kind of meaning that we really seem to desire when we're working in these sorts of positions? It seems almost universal that working at a computer means we feel isolated, even when we talk in meetings all day.

I think the nature part makes sense, after all, being in a field is certainly more natural than an office, but I think lots of farming is actually pretty lonely from when I've done it, with the exception of animals. But when thinking about the profession, it just feels more social. There's something there about the way we view types of work and their importance, almost metaphysically speaking.

morleytj··on Cottagecore Programmers
Thank you, I appreciate it! I've been wanting to write things more often that aren't just for work, so I was thinking I'd try to write blog posts like this to practice.
morleytj··on Cottagecore Programmers
Definitely a big part of it -- I think that sitting in front of a computer all day is a situation that makes you feel detached from your work, and it's harder to feel like it has meaning. I wrote this largely because I was curious to drill down into why that was the case and whether it was unique to tech workers, or if everyone was feeling this to some degree, and perhaps tech workers were just the most able to transition to other fields like you mentioned.
← PreviousPage 3 of 3