• 5 Posts
  • 570 Comments
Joined 3 years ago
cake
Cake day: October 20th, 2023

help-circle


  • Which is why people who actually look at trends tend to compare it more to the Dot-com bubble.

    The short version? A few early internet adopting sites (like Amazon…) set up online retail presences. People were ecstatic because you could now do most of the monthly shopping online and even re-buy pants that you know will fit and so forth.

    Seeing money, EVERYBODY made an online retailer or service website and EVERYONE wanted to invest in that.

    Then the market was oversaturated and companies with no right to exist went bankrupt and it was a bloodbath.

    Except… not really. Because while the massively overinflated stock market did indeed “downturn” and a LOT of those scam companies went away, the actual fundamental premise of online first companies was a very sound one. I mean… just look at “Cyber Monday” and so forth.

    And “AI” will almost definitely go the same route. Because, yeah, LLMs are HORRIBLE for accounting and finance. But they are actually really good for replacing the early career folk who translate earnings into reports. And ML in general is excellent at detecting patterns which can mean potentially billions of dollars in investing. But, like all things, it is about verification and caution. You actually need a human to read that earnings report before you send it to the investors. And you only give your “AI” a small portion of your portfolio. Same as with any team.


  • What you are describing is something different… that is “close enough” to Moore’s Law for all but the most pedantic.

    The (I forget the proper economics term so) base price of RAM/Storage does indeed go down as new processes and economies of scale are developed. But the cost of a “laptop hard drive” remains pretty steady in the sense that a couple hundred MB was enough back in the day but you REALLY want at least 500 gigs now. The price per byte does indeed drop rapidly but the price per “drive” is far more stable (not fully stable due to inflation and how many people are buying them, but within spitting distance).

    Its why a good rule of thumb was to always just spend roughly the same on storage during an upgrade and that would result in faster technologies and larger capacity drives and so forth.

    That isn’t what is happening with RAM in 2025. A much better comparison is GPUs because… it is the same problem. It is ridiculously high demand from businesses (often startups pouring dump trucks of VC money into their only hope… well, VC money or drug money in the case of miners but they matter a lot less these days) driving this. A quick search didn’t yield an easy graph and I can’t be bothered to go dig through Gamers Nexus’s twelve videos on it, but the price of an “entry level” GPU has drastically changed in the past decade.

    But just for two-ish data points?

    • The GTX 980 and 970 had an MSRP (probably) of 550 and 330 USD, respectively, back in 2014
    • While there is some other bullshit involved, the RTX 5080 and 5070 have MSRPs of 1000 USD and 550 USD in 2025
    • Adjusting for inflation, the 980 and 970 would still only be about 753 and 451 USD in 2025 dollars
    • And let’s not forget that basically no cards were sold at MSRP back in early 2025…

    The last point being what is, by all accounts, going to be the new normal. Barring outside impacts like… RAM going through the roof. Vendors will sell the cards for the ACTUAL MSRP rather than the inflated demand prices. And they will still be considerably more expensive as a result.

    All of which is to say… my current card is definitely good enough but having a hard time deciding if I do one “final” upgrade for the decade. But I am an AMD boi so those are at least “reasonable” in terms of price per performance.


    1. Prices rarely, if ever, go down in a meaningful degree. Stuff like this is partially necessity and partially a REALLY good excuse to see what the price ceiling actually is… and then turn that into the floor moving forward. Just look at gas prices
    2. The “AI Bubble” is likely to be on the same level as the Dotcom Bubble and the like. It is going to be brutal and a LOT of people are going to lose their jobs… and then much of the same tech will still dominate just with more realistic expectations. And that will still need large amounts of memory
    3. If the “AI Bubble” really is as bad as people seem to want it to be: A LOT of the vendors who make the parts you are buying RAM to use are going to be gutted. And then RAM production will drop drastically. Which will decrease supply and…





  • Yes. Which is why I specifically mentioned the need for grants and incentive programs to allow those companies and research groups/universities to pivot.

    Because, yes, we very much need better social programs. But there are going to be a LOT of people out of jobs. And a LOT of early career/new grads who just spent the past 4-8 years of their life literally training themselves to do what the government et al demanded of them. And they’ll be shit out of luck.

    All of which will very much overwhelm whatever half-assed social programs we rapidly implement.

    Like… to be blunt, what you are suggesting is very comparable to the “trump healthcare plan” of giving every 2 grand and telling them to figure it out themselves. On the surface, maybe that sounds nice. But that is not much of an insurance premium and would get cleaned out the first time someone gets sick. It is not a solution.

    The reality is that we desperately need actual social safety nets and we have needed them for decades at this point. But “We’ll figure something out in a few years, for real this time” doesn’t help people in even the medium term.





  • I… actually disagree with this even though I don’t want to.

    Yes, fuck OpenAI and Meta.

    But the problem is that the government has been 500% enabling this “bubble”. Basically any governent grant for the past five or six years has more or less required “AI” to be front and center. The more senior folk have generally been good about “I am doing Machine Learning enhanced construction of that office building, sure, whatever” but this, in turn, has more or less gutted academia as more and more programs have essentially been built around providing what funding is literally asking for. Flipping NOBODY thinks “I am a prompt engineer” is worth anything but scorn and mockery. But when those grants and CFPs basically require a prompt engineer on the proposal to have any chance of being accepted…

    It’s easy to say “you should have known better and not lied to us” but… they were telling the lies that were demanded of them. And just cutting them free and letting capitalism sort it out is going to mean pretty much an entire generation lost and countless companies down the drain as well. Which… also won’t be all that useful for getting people healthcare.

    Similarly, Jensen is a prick (and always has been but…). But if the rug gets pulled and the “bubble pops”? nVidia is not a US company but it is essential to the US economy and capabilities. We can’t afford to let that go the way of Intel.

    Do I think giant checks should be written? No. But we need a LOT of transition grants and programs and possibly outright stimulus checks so that the entire tech and research and science industries can pivot to what is being asked of them now.


  • A LOT of this narrative reeks of “See, chuck schumer is actually a heroic genius”

    But the basic idea is that trump can still pocket veto this. I forget the specifics but essentially it is the idea that he can ignore the bill for 10 days which would expire it and punt it back to the House. And if The House is out of session (which they will be because it is thanksgiving and our politicians would NEVER dare inconvenience themselves for The American People) it basically kills the bill entirely and they have to re-pass it.

    By getting it through the Senate ridiculously quickly it makes it MUCH harder to argue that the white house ran out of time. And it does sound like House republicans may have been counting on that since johnson was talking about how the Senate would take forever to pass this.


    The other theories I have seen are all based around “investigating the Democrats”. Either insist that NO files can be released because they are part of an active investigation or insist the reason it is predominantly republicans and bill clinton is because all the Democrat files had to be held.

    Personally? I think both immediately trigger massive leaks because of just how widespread they are and how bipartisan the access has been in the past. And any token sacrificial republicans will IMMEDIATELY narc on everyone else before they can hang themselves in a jail cell.

    And if the big beautiful bill ever faces even a chance of consequences for his actions, you can bet that he (and mostly Hilary) would start blabbing about trump sucking him off.