> Most LLM users seem willing to change from Chat-GPT to Claude, for example
There is some nuance to this. If you're building an application that embeds an LLM, then your "user" might be a prompt engineer, not necessarily the user using the application. It just so happens you can use the embedded magic using the prompt yourself.
Not a single mention of Chat-GPT or Claude, but if you google you'll see they use Claude under the hood. So I would argue the branding is actually "AI" not ChatGPT.
It's a bit like Crypto and Bitcoin. Not all Crypto is bitcoin, but all bitcoin uses crypto to power it. People recognize both the branding of Crypto and Bitcoin.
Second, the investment isn't a loan that they need to repay. They are getting equity.
Third, Anthropic is exclusively using AWS to train its models. Which, yes, means if AWS gives them $4B and it costs them $500M/year to pay for AWS services then after 8 years, the cash is a wash. However this ignores the second point.
Fourth, there is brand association for someone who wanted to run their own single tenant instance of Claude whereby you would say "well they train Claude on AWS, so that must be the best place to run it for our <insert Enterprise org>" similar to OpenAI on Azure.
Fifth, raising money is a signaling exercise to larger markets who want to know "will this company exist in 5 years?"
Sixth, AWS doesn't have its own LLM (relative to Meta, MS, etc.). The market will associate Claude with Amazon now.
A quibble: AWS _does_ have an AI story (which i was originally dismissive of): Bedrock as a common interface and platform to access your model of choice, plus niceties for fine tuning/embeddings/customization etc. Unlike say Azure theyre not betting on _a_ implementation. Theyre betting that competition/results between models will trend towards parity with limited fundamental differentiation. Its a bet on enterprises wanting the _functionality_ more generally and being able to ramp up that usage via AWS spend.
WRT titan my view is that its 1) production r&d to stay “in the game” 2) a path towards commoditization and lower structural costs, which companies will need if these capabilities are going to stick/have roi in low cost transactions.
The big tech companies are spending enormous amounts for part ownership in startups whose only assets are knowledge that exists in the public domain, systems that the companies could have engineered themselves, and model weights trained with the buyer's own capital. The people who will get hurt are public investors who are having their investment used to make a few startup people really rich.
Knowledge is quite the useful asset, and not easily obtained. People obtain knowledge by studying for years and years, and even then, one might obtain information rather than knowledge, or have some incorrect knowledge. The AI companies have engineered a system that (by your argument) distills knowledge from artifacts (books, blogs, etc.) that contain statements, filler, opinions, facts, misleading arguments, incorrect arguments, as well as knowledge and perhaps even wisdom. Apparently this takes hundreds of millions of dollars (at least) to do for one model. But, assuming they actually have distilled out knowledge, that would be valuable.
Although, since the barrier to entry is pretty low, they should not expect sustained high profits. (The barrier is costly, but so is the barrier to entry to new airlines--a few planes cost as much as an AI model--yet new airlines start up regularly and nobody really makes much profit. Hence, I conclude that requiring a large amount of money is not necessarily a barrier to entry.)
(Also, I argue that they have not actually distilled out knowledge, they have merely created a system that is about as good at word association as the average human. This is not knowledge, although it may have its own uses.)
If they could build it themselves, why haven't they? Say what you want about Amazon, but I find it hard to believe that Anthropic bamboozled them into believing they can't build their own AI when they could do it cheaper.
Curious - so how do you manage client-specific schemas then? Do you just have mappings in postgres (column1, column2, column3, etc.) or maybe store a client specific schema in bson per client?
It's all JSON in two Postgres tables: `collection` which represents a Notion Database and has a `collection.schema` JSONB column, and `block` which has a `block.properties` JSONB column that stores the property values - the stuff in the Notion Database cells - for each row. We apply "schema on read" when querying or rendering a Notion Database, and we have a service on the backend that builds indexes/caches for hot collections on the fly. The service handles all the queries for collections larger than X rows. For smaller collections, we just give the client the whole thing modulo permissions and it does the query locally.
There are endless amounts of stories and situations in which selling something before it really exists has helped businesses. It's totally plausible that a team working on video streaming at the scale of Netflix could figure out live streaming.
Pre-optimization is definitely a thing and it can massively hurt (i.e. startups go under) businesses. Let's stop pretending any businesses would say 'no' to extra revenue even before the engineering team had full assurance there was no latency drop.
In my language we have a saying that roughly translates to "Don't sell the hide until you've shot the bear".
And sure, there have probably been lots of examples where a business made promises they weren't confident about and succeeded. But there are surely also lots of examples where they didn't succeed.
So what's the moral of the story? I don't know, maybe if you take a gamble you should be prepared to lose that gamble. Sounds to me like Netflix fucked up. They sold something they couldn't provide. What are the consequences of that? I don't know, not do I particularly care. Not my problem.
> actively thinking about this and dedicating part of their work to misinforming the poor intentionally.
I'm reminded of the quote
> “No one involved in an extralegal activity thinks of themselves as nefarious. I'm a businessman, okay?" - Quark, DS9 S6E25
I don't believe anyone nefariously sits there and says "lets make sure people aren't educated" but I genuinely believe there are people who say "I did this thing and people keep voting me in to keep doing that thing or keep paying me to do that thing, so I'm going to continue doing it that way"
You are 100% correct, people mostly follow incentives. The problem is that, for many politicians, "this thing" in "I did this thing and people keep voting me in, so I'm going to continue doing this thing" refers to the starving and/or defunding of education.
> It's what a lot people here have already said. Education is what is failing, EdTech didn't magically solve this.
To expand this more globally - anything that requires human interaction fails at scale. Healthcare, trade skills, education, housing, etc. is all "failing" to some degree no matter how much technology we throw at it. The costs continually go up and the value isn't paired to it.
To put a name on this, it's the Baumol effect, https://en.wikipedia.org/wiki/Baumol_effect. Essentially, as productivity increases in most industries (from automation), it drives up labor costs in the industries that can't be automated (healthcare, education, performing arts, etc), which drives up the price of those services (healthcare), or drives down the quality to find a market clearing price (increasing student to teacher ratios).
> anything that requires human interaction fails at scale
What do you mean by "fails"? It doesn't have the zero-marginal-cost dynamic favored by the investor class of the tech industry?
Don't confuse "you can't scale human interactions" with "human interaction fails at scale". The former is talking about the venture-capital-accelerated winner-takes-all business strategy, the latter is a misunderstanding of civil societies and living a rewarding life because you can only envision such things through the lens of owning a zero-marginal-cost process.
These industries, more or less, need to be highly regulated because we've faced the alternative and it's worse. The quality of education went up SIGNFICANTLY when the state took more control over it.
Did it really improve, though? There's evidence that Colonial/Revolutionary-Era American populace was highly educated, and that education had been gradually going downhill as the United States gradually adopted the Prussian model of education -- an education system designed to produce loyal soldiers and interchangeable worker widgets.
There is a strong tendency to recognize "market failure" as a problem -- but it is rather rare to see people discuss "regulatory failure". Yet there is this endless cycle of increasing regulation, where a politician or bureaucrat identifies a problem with the market -- real or imagined -- and then offers a solution. That solution the situation even worse, but politicians and bureaucrats will blame this on "market failure", and then insist on another solution ... and so it goes.
Yes, it's true that some regulation is necessary. But how much effort is put into place to make sure that the regulations put in place do what they are supposed to do? What effort is put into place to identify the nasty side effects of otherwise good regulations? The assumption that all regulations exist because the alternative is worse is deeply flawed.
And that's even before we get into the fact that regulations have their limits: people can handle only so many rules, particularly if they are onerous, and will start ignoring regulations when it's the only way to get things done. An example of this is a report from the FAA, where they identified around 2,000 pilots who didn't report disqualifying health conditions. Why? Because the conditions were quickly resolved, and their livelihoods depended on flying, so they didn't want to go through the hassle of waiting three months for their license to fly to be cleared again. What's worse, of the 2,000 or so pilots who did this, only 4 or so had serious-enough conditions that they shouldn't have been flying!
> There's evidence that Colonial/Revolutionary-Era American populace was highly educated
I've never seen this evidence, but to speculate: probably since the colonist were, well, colonists, and literally building every industry from the ground up, they had to be.
After this though, in the 19th century, we knew most people could not read or write.
> an education system designed to produce loyal soldiers and interchangeable worker widgets
True, but worker widgets who can read.
> There is a strong tendency to recognize "market failure" as a problem -- but it is rather rare to see people discuss "regulatory failure"
I experience the opposite, but granted I live in Texas. The default mode is regulations are bad, and the less the better.
> The assumption that all regulations exist because the alternative is worse is deeply flawed
It's not an assumption, it's an observation. When it comes to education and others.
The thing is the free market relies on there being a free market. Not every market can be a free market. That's a problem. I've already explained how education can never be a free market above, but another example is healthcare. I can't compete in healthcare because of the HUGE time investment in schooling. I also have zero choice as a consumer - I just go to the nearest hospital. There's also no visibility in price because you have to diagnose me first, and that can be wrong or things can pop up later.
Healthcare is not, and will never be, a free market. Even now the US has a largely socialized healthcare system, it just so happens to be the worst socialized healthcare system. Primarily because we've handed it off to the private sector, who have incentives to actually make it worse. Meaning, insurance thrives when you receive less care. The inefficiency is built into the profit motive, and you can't get around it.
> I've never seen this evidence, but to speculate: probably since the colonist were, well, colonists, and literally building every industry from the ground up, they had to be.
I would strongly recommend looking into John Taylor Gatto's "An Underground History of American Education", at least as a starting point. Indeed, that is where I learned that a literacy drop between WWI and WWII raised eyebrows, but when it came to the Vietnam era, the Army had to investigate "How can so many people fake illiteracy?" and found that the answer was "They aren't faking it."
> After this though, in the 19th century, we knew most people could not read or write.
Now it's your turn: I'd like to see your evidence for this.
> True, but worker widgets who can read.
That's assuming that everyone (including educators and politicians) think reading is important.
>> There is a strong tendency to recognize "market failure" as a problem -- but it is rather rare to see people discuss "regulatory failure"
> I experience the opposite, but granted I live in Texas. The default mode is regulations are bad, and the less the better.
Have you considered the possibility that what is happening in Texas is an attempt to recognize harmful regulations, and weed them out so that people can have more freedom to function?
>> The assumption that all regulations exist because the alternative is worse is deeply flawed
> It's not an assumption, it's an observation. When it comes to education and others.
I cannot help but worry that your observations are cherry-picked. There's plenty of regulatory fail to observe for the few who look to observe it.
> The thing is the free market relies on there being a free market. Not every market can be a free market. That's a problem. I've already explained how education can never be a free market above, [ed: I can't find that explanation] but another example is healthcare. I can't compete in healthcare because of the HUGE time investment in schooling. I also have zero choice as a consumer - I just go to the nearest hospital. There's also no visibility in price because you have to diagnose me first, and that can be wrong or things can pop up later.
> Healthcare is not, and will never be, a free market. Even now the US has a largely socialized healthcare system, it just so happens to be the worst socialized healthcare system. Primarily because we've handed it off to the private sector, who have incentives to actually make it worse. Meaning, insurance thrives when you receive less care. The inefficiency is built into the profit motive, and you can't get around it.
I have both experienced and seen what health care in Great Britain is like, and based on the statistics I hav come across over time, to this day, to suggest Great Britain's health care is better than America's is a big stretch. Something similar can be said of Canadian health care (Pittsburg, PA, has more MRI machines than all of Canada!), and certainly, Cuba's awful health care system should be considered "socialized". To the degree that American health care is better than these systems, it's largely because of the private aspects of it -- and what's worse, to the degree that it's harmful, it is because of government regulations that kill competition. How is it that the law requires that I have the freedom to choose where I want my car repaired, but that I must find an "in-network" doctor? Why can I purchase auto insurance from any company in any State, but I must have insurance sold in my home State? How the heck are these specific kinds of restrictions supposed to make health care better?!?
Many of the concerns you gave about health care apply to the car repair market too -- both in terms of insurance and in terms of car repairs and maintenance. Why is it that mostly-privatized car repair is market-friendly, but health care is not?
I'm really not going to bother replying any further because I don't think it's worth it. You're fighting and uphill battle here - virtually every country has agreed that having the government control education is a good thing and has been successful.
Conservatives, in the past few years, have fallen into a new sphere of lunacy. Anti-vaccine, anti-education, anti-workplace safety, anti-everything.
I don't have the patience to stamp out every single opposition people come up with.
To me, the "problem" that you, and others, have with education is that the government is involved. Sorry, that is not an argument. Just as blindly supporting something makes one a sheep, blindly opposing something also makes someone a sheep.
Our education is as successful as it is due to our government, and that is not up for debate. I've already explained the problem with free markets and why some services, such as healthcare and education, do not apply to a market and literally cannot.
As a thought experiment, consider what makes a free market a free market. Now consider the feasibility of implementing that into healthcare and education (hint: impossible). Now, assume it is possible, and consider the consequences of it. How many millions of people would need to die before you start caring?
In my experience, it's been just as difficult to convince conservatives to let go of the notion that only government can solve problems, as it has been for liberals. For this kind of "lunacy", you generally have to go full-blown libertarian -- and very few people listen to libertarians.
As for free market health care and education: Are they really impossible, though? And if it's "purely" impossible, is it really impossible to introduce some fundamental free-market reforms, like school choice, or the ability to choose your own doctor?
As for "government-run education being successful" -- ever since I started school in the 1980s, government-run schools have been in crisis mode -- indeed, it's what's led to things like the Department of Education and No Child Left Behind. What's more, there is no evidence things are improving, and plenty of evidence that it's been degrading over time.
How many children's futures are you willing to cut off at the kneecaps until you finally admit that government education, particularly in the United States, has serious problems?
Part of that is questionable, because the state gets to set the means by which the quality of education is measured. However, a more objective measure is the extent of education; certainly, the introduction of national education has always produced a much greater literacy rate.
The thing is extent and execution are intrinsically linked. Meaning, unregulated markets mean the extent will be limited because people will just choose not to get educated. Which makes sense, if they themselves are uneducated they don't have perfect reasoning skills or future outlook.
A market relies on the participants having visibility and ability. They need to see the alternatives, understand them deeply, and have access to them. Turns out you can't do that in a bunch of markets, education being one of them. So, it can never be a true free market.
Is not the absolute cost that goes up per se. It’s the cost relative to cost of things that scale. Education, nursing care, even doctoring all have significant human to human interaction at a personal level for there to be success. When the cost of other things like commodities drop 100x, those human interactions services don’t and therefore become relatively more expensive even despite massive productivity gains those fields
I understand education and healthcare, but how do trade skills and housing require human interaction? For housing especially, a lot of foreign/remote investors can own houses and just collect rent checks from tenants in a pretty hand-off manner. Housing is supply-limited, sure, and heavily regulated, sure, but I don't think it requires human interaction, and certainly not to the scale of education or healthcare.
reply