I do have to wonder what the mix is between "our data show this is how most people want to be talked to" and "these tokens lead to better responses on objective measures of correctness." That is, in the training data insightful questions are tangled with insightful answers, so if the bot basically always treats the user like a genius it gets on the track that leads to better answers.
Yeah, I similarly doubt that LLMs are going to directly lead to AGI just via scaling and might almost be a dead end in that direction.
But they're still quite useful tools and accelerators or force-multipliers.
And you're still going to need humans in the loop.
And I'm very worried that the capex buildout will implode once we hit diminishing returns and good-enough models can be run on substantially smaller footprints.
It all isn't going away, though, and it will still continue to improve.
Also, all the cloud models don't have to be the best frontier models, and you don't need to focus on hitting the benchmark of shrinking Opus 4.7 down to a single MBP to make significant improvements. If you get it so that an Opus 4.7 benchmark-compatible model can run in $250k of datacenter capex (and associated reduced opex for power+cooling) that'd be a massive cost improvement that makes the cloud models cheaper. And for most consumers that'll probably be good enough. You don't need to run on a $5k laptop to make a big difference.
However, inference costs for entirely good enough models are likely to keep declining in the future. We're probably hitting diminishing returns on model size and training. The new generations aren't quantum leaps anymore, and newer generations of open source models like DeepSeek are likely to start getting good enough.
There's going to be a limit to how much they can raise prices, because someone can always build out a datacenter and fill it up with open source DeepSeek inference and undercut your prices by 10x while still making a very good ROI--and that's a business model right there. Right now I'm sure there's a lot of people who will protest that they couldn't do their jobs with lesser models, but as time goes on that will get less and less. Already right now the consumers who are using AI for writing presentations, cooking recipe generation and ELI5 answers for common things, aren't going to be missing much from a lesser model. That'll actually only start to get cheaper over time.
Also for business needs, as AI inference costs escalate there comes a point where businesses rediscover human intelligence again, and start hiring/training people to do more work to use lesser models--if that is more productive in the end than shelling out large amounts of cash for inference on the latest models. [Although given how much companies waste on AWS, there's a lot of tolerance for overspending in corporations...]
> because someone can always build out a datacenter and fill it up with open source DeepSeek inference and undercut your prices by 10x while still making a very good ROI-
Not sure how it all works out. Currently trillion dollar companies can't make a native app for platforms. Everything is just JS/Electron because economics does not work for them.
And here companies can make GW data center running very expensive GPUs for 1/10th of current prices. Sound little fanciful to me.
The price you pay for anthropic must include the price of training new and better models which is incredibly costly. If you use the models someone else already spend money to develop you don’t need to pay this price.
They've been like that for a while actually, I think at least since the big hype around ChatGPT 4.5 (or was it 5?) and that underwhelming, lukewarm, oversanitised presentation by Altman and his team.
Yups... Mythos is the smallest possible leap. Not a standard model generation advance, not even a version point advance. Just the smallest possible quanta of a change. We are absolutely hitting a plateau any day now. Any day. Any time. Any second now. Yup. Right now! Surely!
I mean let's be realistic - all that we know about the "mythical" Mythos is the carefully curated and release stuff by the Anthropic's PR team. Is it really a huge leap they are making it to be? I doubt it. In fact I bet if it was indeed that powerful and dangerous, as they imply, they'd find a way to release it immediately, devastate OpenAI and DeepSeek and secure a leading position in the market. Why is it not happening? I suspect because Dario is again at it, peddling his bullshit.
Yeah. AI progress is insanely fast if you compare it to anything else. Where else is a one year old technology already hopelessly outdated? 10 years ago is basically stone age.
I am continually tripped out by the fact when I was 16, I didn't have a 'smartphone' beyond a Windows Mobile 6 phone that had no internet on it.
Now, I have this high-resolution shiny object that can near instantaneously get any information I want along with _streaming HD video to it_ *anywhere*.
15 years even feels like a stone age. I can't fathom what it has to feel like people in their 60s and 70s.
I'm not quite 60, but it's always interesting to me that I feel quite the opposite of this. When I was 16, I didn't have a computer, didn't have a phone, had never used the Internet, but when I think of how life has changed, it's frankly not much. I woke up this morning, scooped my cats' litter boxes, took out some trash, made myself breakfast, ate that, read some news while eating, then lifted weights in my garage, had some work meetings, wrote up some instructions per a customer request from Friday, and am about to go drive to the lake to go do a 9 mile longboard loop.
That's very close to a normal day in 1996. The biggest difference is I read the news on my phone instead of a physical newspaper. The news was not any more interesting or informative because of that. I guess I can also still do the loop reasonably well, but I'm a lot slower than I was in 1996 when I was a cross-country state champion.
My parents are closing in on 70 and I guess I can't speak for them, but I'm at least aware of the daily routines of their lives, too. Walk the dog, do housework, DIY building projects, visit kids and grankids. Seems much the same, too, with the biggest difference being they're now teaching my sister's sons to play baseball rather than me, but shit, one of her sons even looks like exactly the same way I looked when I was 7! The more things change, the more they stay the same.
If your parents are closing in on 70, I would have expected you to be closer to not quite 50 than not quite 60.
I am just over 50 myself and I agree with your points. Technology has changed but life is largely very similar to wear it was in the 90s. At least day to day. Attitudes are way worse now.
Depends on where you live. My dad is almost 80, grew up in a very rural area, and when he was 16 they'd just gotten indoor plumbing. Up until he was 14, his school was a one-room school house with no heating other than a wood stove. If you were the first kid to arrive for the day, it was your job to get the fire going in winter months.
Housework meant no laundry machine, no dishwasher, and possibly no vacuum cleaner. That means hand washing everything, and beating rugs with sticks and brushes to get the dust off of them.
The early lives of my grandparents (in their 90s) are so fascinatingly different to that of mine. But even by the time my parents were growing up in the 60s, life was not so different in the west. The real differentiators in living standards - energy, household appliances and cooking, modes of transport - were more or less figured out then. By the time my parents were young adults in the early 80s, so many of the aspects of "modern life" had been figured out.
I look at the life my kids live, and it's not so different to my childhood. The toys are similar, their housing is similar. Probably the biggest difference is the availability of content on demand rather than much more fixed TV schedules.
The big difference in the last 30 years hasn't so much been in the kind of middle class life you can live, but the number of people who live that kind of life. In the 90s 40% of people globally were living in extreme poverty. Now its under 10%. The kinds of lives the middle class live in China and Vietnam are closer to those of Europeans today, when even 30 years ago most people in those countries were living much closer to the way your dad grew up.
I wonder if AI will result in a step change of living standards? Perhaps along with robotics we'll finally get to do nothing at all at home? I'm not convinced it'll be quick though. Maybe another 30 years.
General agree... I still do the things (mid-50's) I used to do when I was a teenager with no computer, no phone.
But - now they are easier - I can read books on an e-ink screen and pretty much instantly find what I want to read next. I get my news on a phone. I used to watch TV/movies broadcast or on tape rentals. Now, I have just about everything I could ever want available - without ADs... those were such a time-waster.
What has changed is that I have access to MORE information than my local (or school) libraries could ever provide - in a variety of more accessible formats. Whatever tools I need to get "work done", I can find a myriad of free and open-source options.
But - the overall days and household family routines are the same - now, instead of reading a paper book while waiting to pickup my kids (or other family members) "back-in-the-day", I can read my device, or connect with my DIY communities online on my phone - or learn something new. I don't have to schedule life around major broadcast events, I can easily do many tasks while I am "out-and-about".
I always wonder the views of older people. My parents are very technology forward and have been my entire life so it is difficult to gauge how different life is compared to when they were growing up.
It's easy to hear "Oh well I only had 640kb of memory and typed programs out of a magazine I got in the mail!" and see as distinct from having 'unlimited' resources and the internet.
Your insight is good ("The biggest difference is I read the news on my phone instead of a physical newspaper") that life sort of stays the same but the modality changes. People still go to the store like they did in the mid-1800s but now it is by car.
I wonder what our "industrial revolution" will be where the previous generation lived (ie: out in the country on a farm) totally different lives to the current (ie: in the city in a factory). Maybe when space travel and multi-planetary living is normalized?
> It's easy to hear "Oh well I only had 640kb of memory and typed programs out of a magazine I got in the mail!"
Since I was there (young, but there), I want to point out that this crosses three eras which all felt quite different:
1978: typed programs in from a magazine or loaded from a cassette (16kB, TRS-80)
1983: loaded programs from a floppy (64kB, Apple ][ and C64 etc)
1988: loaded programs from a hard disk (640kB, IBM PC and Mac).
Exact years vary but these eras were only about 5 years each. Nobody had a floppy in 1978 but almost computer user did by 1983; nobody had a hard drive in 1983 but almost everyone did by 1988.
To some degree this already happened with the move from the industrial city to suburbanization and then re-urbanization. In particular one of the most notable recent developments is that urban waterways are now pretty desirable places to be with parks and recreation; in most industrializing cities the waterfront was actively avoided because the industrial use made it polluted, smelly etc.
And at some point even frontier model costs will hopefully come down (if there is still a meaningful difference between closed and open source models at that point) as all of the compute that's being built out right now comes online.
One question the city probably can't answer is what disabled persons in the taxi are supposed to do. If you strictly enforce bike lanes the result is probably the rider needing to walk a few blocks. If the rider is disabled, that could actually be a huge burden. Since I've got an 80+ year old disabled parent with a walker this is an issue for me that does compete pretty aggressively with my support for bikes.
Designated drop off points with disabled person priority is the answer. How do you dropp off a disabled person in a lane with clogged street side paking? Shouldn't you be against street side parking by the same logic?
> Designated drop off points with disabled person priority is the answer.
You have to actually put those drop off points on every street corner, though, and is the lowest priority for city planners, and gets rolled out at a glacial pace over decades--while disabled people have needs today. Having been forced to confront how well designed our infrastructure is for disabled access and put it to the test, I can tell you that a lot of existing ADA infrastructure can best be described as "performative" and a lot of it is entirely missing.
First of all, the walk would rarely be more than half a block. Bike lanes go down a small number of streets, so one can usually unload on an intersecting street. Not ideal, but ...
... bike lanes are not the only thing that creates this issue. Any road that lacks parking, with or without bike lanes, will have the same problem. Even when there is parking, all of the parking spots may be occupied. In both cases, people may have to walk a few blocks. While they may be grouchy about the lack of (sufficient) parking, you don't see many people blaming motorists for placing a burden on the elderly.
Finally, it is always possible to make accommodations. Having a carve-out for loading and unloading taxis will do far more for safety of everybody than letting people stop anywhere in bike lanes. It is also possible to have exceptions for people with disabilities, as long as non-disabled people don't abuse it.
In the UK (which the article seems to be about, although it keeps talking about other cities too), it depends on the specific road markings. If it's a solid white line between the bike line and the road, absolutely no vehicles are allowed into the cycle line except bikes. If it's a dashed line, they can enter it as long as it's not being used by cyclists.
In general though, special affordances for parking and dropping off disabled passengers is only given to those with a blue badge. AFAIK in general taxis carrying disabled passengers without a blue badge have no more rights than any other vehicle. Only companies that specifically care for disabled passengers can get their own company blue-badge rather than relying on their passengers having one.
Right, but in general, what is a car driver supposed to do when he needs to stop and unload (anything) in a place where there isn't a special provision for that? Here in Slovakia they can stop in the car lane, blocking it, as long as that's not explicitly forbidden by a sign there.
Because we also have streets with a car lane and a bike lane with a solid line where it's forbidden for any other vehicle to enter or stop there, so of course I had to explain to a few drivers what are the rules they should alrady know and follow.
It depends on the road markings and signage. In general double yellow lines mean no waiting or parking, but there is generally a special exemption for loading/unloading or letting passengers in or out. However, there is frequently additional signage that will restrict hours when loading is permissible, or state no loading.
Double yellow lines (waiting / parking) are different to bike lines however. If a bike line is separated from the road by a solid white lines, motorised vehicles are forbidden from entering at all. If you want to load/unload or drop off passengers, you need to do that somewhere else. If the highways agency considered it safe for vehicles to block the cycle route in that place, they'd use dashed lines instead.
What is missing is a denominator and some standardization.
There could be 100,000s of people in the US who have jobs where their disappearance could be considered "concerning".
And then we need a base rate for people of similar socioeconomic status. They're probably disappearing at a far smaller rate than the general population, since they're not poor, not sex-workers, not troubled teens, etc. However, there is still a base rate, and you still need to show that it exceeds that base rate--and I kind of doubt that it actually does.
We have a large population, and over the course of a few years 10 weird things happening seems entirely normal to me.
> an authoritarian regime, with no respect for human rights
I mean Alex Pretti was murdered because he was exercising his 2nd amendment rights and was getting beaten after having been disarmed and never displaying his firearm.
Headlines are full of people who are dying without medical care because they're being incarcerated and shipped overseas due to legal technicalities.
Nice whataboutism, nothing has changed since the USSR, I see! The US also has its problems; this doesn't give the PCC a free pass. And those issues will likely go after the elections, something people in China haven't been able to do since the PCC took over the country.
Anyway, shall I discuss the cultural genocide being committed on the Uighurs, or the Tibetans? Or, in a quite bizarre fashion, the descendants of the Mongols? [1]
The goulags created during covid? And let's not talk about all the horrors of the Mao-era PCC. Etc, etc...
You are aware that there are not only two countries in the world, right? And that China could have been something in between the current maoist regime and America? Like how Taiwan is, or Hong Kong was?
Yeah, AI-enabled surveillance capitalism is likely to be every bit as bad as what people imagine China is doing with their social credit scores.
And the scary thing is that you can probably easily sell it to Democratic voters if you track racism scores for people, so you can filter people out of your dating pool or job/rental applications. Most people don't care about privacy as a fundamental right, and they'll roll over and compromise if you give them a way to track what they hate. You just need to make sure it is "bipartisan" and it'll be wildly popular.
> Is it not sufficient to employ good engineering and design practices? Git...
Git has horrible design and ergonomics.
It is an excellent example of engineers designing interfaces for engineers without a good feedback loop.
Ironically, you just proved your point that engineers need to better understand how users are actually using their product, because their mental visualizations of how their product gets used is usually poor.
People say this and never has written about the supposed failure of design. Git has a very good conceptual model, and then provides operations (aptly named when you know about the model) to manipulate it.
Most people who complains about git only think of it as code storage (folder {v1,v2,...}) instead of version control.
If you don't want to look at what people write you can't say that they haven't written about it.
> the supposed failure of design
I don’t think people complain about the internals of git itself as much as the complexity of all the operations.
If you want to read about complaints, you really don't have to look further than the myriad of git GUIs, TUIs and otherwise alternative/simplified interfaces.
> I don’t think people complain about the internals of git itself as much as the complexity of all the operations.
The complexity is only there when you want to avoid learning what you’re doing. Just like find(1) is complex if you don’t know stuff about the file system or sed(1) is complex if you don’t know regex and line based addresing of a text file.
A lot of people who are using git don’t want to know what a commit is and their relation to branches. And then they are saying rebasing is too complex.
> If you want to read about complaints, you really don't have to look further than the myriad of git GUIs, TUIs and otherwise alternative/simplified interfaces
Git is a cli. The goal is always for you to find your workflow then create aliases for common operations. It does assume that you want complete control and avoid doing magic (which is what jj is doing).
Wanting magic is great (I use magit which makes git magical ;) ) but it’s like wanting to fly a plane without learning the instruments.
I've written my own toy git implementation before, I know how the underlying design works.
You're assuming that I have trouble understanding and using git, when I actually have 16+ years of experience with it, and I can use it just fine.
It is horrible to try to teach to new people. Those operations sort of make sense one you already know how it all works, and can see how the coders wrote it. But that is backwards, since it actively makes it more difficult for neophytes to understand it. "It gets easy once you're an expert" isn't actually good design.
How many years of experience with git do you have? How much of git do you use? I bet you use 5 commands and 10 flags at most. Take a look at git's docs
I always felt like humans that were good at writing that way were often doing exactly what the LLM is doing. Making it sound good so that the human reader would draw all those same inferences.
You've just had it exposed that it is easy to write very good-sounding slop. I really don't think the LLMs invented that.
Sure some people could write well but didn't have a clue but they failed to maintain interest since once you realized the author was no good you bounced once you saw their styled blog.
Now they don't care as they only want the one view and likely won't even bother with more posts at the same site.
I'm one of those aspy people who immediately don't trust other humans who try to fluff up my ego. Don't like it from a chatbot either.
But the fact that all the chatbots do it means that most people really crave that ego reinforcement.
reply