AI Recomping Controversy, PC Games on Switch and more (ft. Scarlet) - Tempcast #99
Theme by VinsCool and AlanJohn
The podcast is also available on Spotify!
Sorry buttercup, but we can't put pandora back in the box.
We need ai models to shift over to HBM modules so they leave PC RAM alone, and then RAM will be cheap again.
It doesn't help that all of the RAM manufacturers are in collusion to basically do a price-fixing scheme.Memory companies have all the incentive in the world to build out production capacity, the bigger problem here is that the memory market is cyclical and, up until 5 minutes ago in historical terms, sales were high whenever standards changed (in-between DDR/GDDR versions) and moderate when they stayed the same. Now all memory that’s manufactured is either already spoken for or sold instantly which is *unusual* for that market. This causes a cooling effect in regards to otherwise obvious and surefire investment as there’s no foreseeable timeline for how long the market will need to stabilise (if it will stabilise at all - this may be the new normal for all we know and the usual cyclical pattern will flatten out long-term). On one hand the more you produce the more you sell, on the other you don’t want to build new fabs that may end up being dormant in the future. They’re cautious to avoid over expansion which causes shortages. Market testing phases like this are a pain in the ass for all parties involved. I assure you, if memory companies had the capacity to make *more money* without the prospect of long term loss, they’d jump right to it.

That’s what they’re being sued for, not necessarily what they’re doing. They’re private companies, they have no obligation to supply the market with anything, and it stands to reason that with limited fabrication capabilities they would focus on what sells the most, and for the most. They’re booked solid until what, end of 2027? They physically cannot produce more memory, that’s the extent of “manipulation” here.It doesn't help that all of the RAM manufacturers are in collusion to basically do a price-fixing scheme.
There are 3 ways to decompile code. Byte matching, instruction matching, and non-matching. All give you source code in the end that you can manipulate or use to port games.for that sort of things, I am pretty sure the AI creates deterministic tools that will not hallucinate.
But I could be wrong? maybe someone who has done this before can confirm?
It's hard to believe that they are not doing it, when it's not their first time. They have a pattern of behavior of doing this sort of thing.That’s what they’re being sued for, not necessarily what they’re doing. They’re private companies, they have no obligation to supply the market with anything, and it stands to reason that with limited fabrication capabilities they would focus on what sells the most, and for the most. They’re booked solid until what, end of 2027? They physically cannot produce more memory, that’s the extent of “manipulation” here.
A nonsenical code can be forked using AI to be better and renamed correctly,cleanened,improved.So the point is moot.If your goal is just to create PC port, sure. The AI can do that. It will probably eventually create a functionally-equivalent but nonsensical codebase full of fakematches (things that compile to the same binary, but are not anywhere close to what was in the original code) and without naming stuff properly.
Said codebase is effectively useless to humans afterwards (and possibly also AI, lacking context from the previous runs), because trying to understand what it does requires you to repeat the reverse engineering effort.
I mean just look at this AI Decomp of Banjo Kazooie. Does any of this code actually tell you anything about how the game works? https://gitlab.com/banjo.decomp/banjo-kazooie/-/blob/master/src/core2/code_C4B0.c
Compare with a decomp done properly:
https://github.com/zeldaret/oot/blob/main/src/code/z_collision_check.c
https://github.com/SMGCommunity/Petari/blob/master/src/Game/AreaObj/CollisionArea.cpp
I don’t.I'd rather have lower RAM and SSD prices myself.

They certainly were fined for forming a cartel in the early 2000’s and found guilty, they were also sued for the same thing in 2016, but the charges were dropped due to no evidence of collusion, so it’s up in the air whether they learned anything from the experience or not. My guess is that a fine is lower than potential profit in the current environment, so nobody’s in a rush to expand on a “maybe”. I also have to consider the element of government coercion here because it’s not just consumer desktops that need memory, not to mention all the market regulators that, for some reason, believe that companies should always cater to demand even if it leads to lower profits. Either way it is unknown how long the AI buildout will continue, or if it will ever stop, so the market is in an interesting position right now. It’s certainly comfortable for the chip manufacturers since their stock is sold even before it rolls off the conveyor belt, so the shortage is very much real and the demand seems steady. All market indicators show that the situation won’t improve until at least late 2027 since all stock they’re capable of making in that timeframe is already allocated, leading to *buyers* trading allocations amongst each other. All the memory-related stock tickers are very volatile at the moment for this exact reason - I should know, I have a small investment in Micron. In theory it should be shooting up into the stratosphere but it keeps oscillating up and down which shows uncertainty. On the flip side it’s trading at all-time highs, so it’s all swings and roundabouts.It's hard to believe that they are not doing it, when it's not their first time. They have a pattern of behavior of doing this sort of thing.
It's true that they are not obligated to produce anything, but if they really are having under the table deals like this.......well I don't think I need to say why market manipulation and oligopoly is a bad thing.
Yea, it's hard to tell when will this nonsense end. The bubble could for all we know pop tomorrow...... The AI craze is simply unsustainable.They certainly were fined for forming a cartel in the early 2000’s and found guilty, they were also sued for the same thing in 2016, but the charges were dropped due to no evidence of collusion, so it’s up in the air whether they learned anything from the experience or not. My guess is that a fine is lower than potential profit in the current environment, so nobody’s in a rush to expand on a “maybe”. I also have to consider the element of government coercion here because it’s not just consumer desktops that need memory, not to mention all the market regulators that, for some reason, believe that companies should always cater to demand even if it leads to lower profits. Either way it is unknown how long the AI buildout will continue, or if it will ever stop, so the market is in an interesting position right now. It’s certainly comfortable for the chip manufacturers since their stock is sold even before it rolls off the conveyor belt, so the shortage is very much real and the demand seems steady. All market indicators show that the situation won’t improve until at least late 2027 since all stock they’re capable of making in that timeframe is already allocated, leading to *buyers* trading allocations amongst each other. All the memory-related stock tickers are very volatile at the moment for this exact reason - I should know, I have a small investment in Micron. In theory it should be shooting up into the stratosphere but it keeps oscillating up and down which shows uncertainty. On the flip side it’s trading at all-time highs, so it’s all swings and roundabouts.
A nonsenical code can be forked using AI to be better and renamed correctly,cleanened,improved.So the point is moot.
Post automatically merged:
I don’t.
Post automatically merged:
Like someone said,any AI "slop" code can be cleaned and commented in minutes,improved and optimised in hours.I can bet my Arm that the flow use it too.imagine you have some repetive tasks that will takes 2 weeks of your times,but Using AI can be done in minutes,who would refuse that? As for wannabee coders that think because they learned few things on C that they can lecture AI users.I say to these wannabees,there is a 1 million context LLMs that they can never compete with.
Easy to answer that,because nobody cares about that project.Not everyone is sitting waiting to correct other people mistakes.If it was a mario,you will get 2 recomp a day.Well, if it's that easy, then why hasn't anyone done it in the months after the BK decomp was completed? Instead those fools seem to be doing it manually.
You should go set them straight.

They said the same thing about the combustion engine at one time. Imagine that, having to drill into the crust of the earth to pump out crude oil and having to refine it into petroleum just to spin a wheel when a horse can spin it just as well and eats grass which just grows out of the ground with zero assistance. I have the opposite view - much like the introduction of the Internet to the masses, AI infrastructure will continue to grow - it must grow in order to meet increasing demand, both from industrial customers and regular consumers. Even once it is built out, it will need regular maintenance, just like web hosting does. Demand will stabilise at some point, but it will always be elevated going forward and the only relief available is a permanent increase in manufacturing capacity. As it was said earlier in the thread, this genie is not going back into the bottle - yesterday AI was making funny memes, then it began making full-blown videos, now it’s decompiling binaries, and accurately too. It’s the new normal.Yea, it's hard to tell when will this nonsense end. The bubble could for all we know pop tomorrow...... The AI craze is simply unsustainable.
It being a part of our lives does not mean it's not a bubble. You heard of dotcom bubble? I am quite sure 'dotcom' is still around, but it was still a bubble, it still popped and it still brought a ton of trouble with it.Easy to answer that,because nobody cares about that project.Not everyone is sitting waiting to correct other people mistakes.If it was a mario,you will get 2 recomp a day.
As for the bubble thing,AI is not a bubble waiting to pop like crypto or MLM scams,its something that will be in our lives like internet and social media.Its a tool ,an invaluable tool.Some people haven't found a use for them.With time computing will not need that much datacenters.when enough are builts,prices (the thing that people hate) will go down on ram and storage.
Issue is......the demand is not truly there. Not to the extent AI companies are trying to sell it as. And that is with it being sold at a huge loss. They bring a ton of issues by themselves and is insanely costly. After more than half a decade, the whole thing is still a net loss. A massive loss. It's not producing any money and it won't produce any in quite a long time unless things change drastically.They said the same thing about the combustion engine at one time. Imagine that, having to drill into the crust of the earth to pump out crude oil and having to refine it into petroleum just to spin a wheel when a horse can spin it just as well and eats grass which is a just growing out of the ground with zero assistance. I have the opposite view - much like the introduction of the Internet to the masses, AI infrastructure will continue to grow - it must grow in order to meet increasing demand, both from industrial customers and regular consumers. Even once it is built out, it will need regular maintenance, just like web hosting does. Demand will stabilise at some point, but it will always be elevated going forward and the only relief available is a permanent increase in manufacturing capacity. As it was said earlier in the thread, this genie is not going back into the bottle - yesterday AI was making funny memes, then it began making full-blown videos, now it’s decompiling binaries, and accurately too. It’s the new normal.

The negative effects of the “dot com bubble” were grossly overblown - 4 years in the red followed by rapid recovery and now an all-time high.It being a part of our lives does not mean it's not a bubble. You heard of dotcom bubble? I am quite sure 'dotcom' is still around, but it was still a bubble, it still popped and it still brought a ton of trouble with it.
You are insanely charitable and far too optimistic of what AI is and is capable of. It's absolutely a bubble and it will absolutely pop. AI is here to stay, but the way it is, is not sustainable.

It would be silly to introduce the concept of a loss leader here, but needless to say, it doesn’t have to “make money” at this stage of development. What AI companies want is to grow and integrate, at any cost, which is what they’re doing. Profit taking comes much later when the tool becomes indispensable.Issue is......the demand is not truly there. Not to the extent AI companies are trying to sell it as. And that is with it being sold at a huge loss. They bring a ton of issues by themselves and is insanely costly. After more than half a decade, the whole thing is still a net loss. A massive loss. It's not producing any money and it won't produce any in quite a long time unless things change drastically.
I also doubt it can decompile binaries in any accurate or useful way. Looking at that decomp someone sent.......It's a mess that is frankly useless. And this is without even considering all the potential legal issues with generated stuff. Copyright is not a joke and it ain't getting abolished. That alone is a huge poison pill.
It's a hype cycle that will at one point break and then will stabilize.
We will see what will be the case in the end, but I lean on it being a fad that will pop, make a huge financial crisis, then after a few years will actually stabilize and its use will be far more reasonable.
The point was merely that it was a bubble and that it popped, then it got stabilized and wasn't being shoved down our throats.The negative effects of the “dot com bubble” were grossly overblown - 4 years in the red followed by rapid recovery and now an all-time high.
View attachment 593646
If anything, “dot com” is more in trouble now than it ever was in the late 90’s/early 00’s because the value of having a domain grossly diminished on account of the advent of social media. In the past every business and many individuals had no option but to market themselves with their own website - now they can just have a Twitter/X page and run with that. If they really need a storefront, it can be powered by Shopify or some other universal tool. Why have your own website to post videos on when you can just upload them to YouTube? Everyone has access to YouTube, it would be crazy to do it any other way. The list goes on and on, everything is increasingly aggregated on a handful of sites. In that sense the Internet is shrinking rather than growing, and yet tech stocks are soaring. COVID alone was nearly as bad, if not worse, for the global economy, although admittedly the recovery was much faster since there was no shortage of confidence, we just “took a break” in terms of economic activity.

I think people waiting for the bubble to “burst” are going to wait for a very, very long time. I’m sure the unprecedented growth will eventually slow down, but AI is not going anywhere. I consider this to be the ground floor, if anything. Going forward it will change our lives in ways that we can’t even imagine yet.The point was merely that it was a bubble and that it popped, then it got stabilized and wasn't being shoved down our throats.
Negatives of the AI bubble popping however.......will be devastating, as you have shown here. The insane speed of growth is unnatural and unsustainable. 50% of the market if not more is AI........... I just cannot be convinced that it can remain like this.
Again......I agree that it's not going away, but this kind of valuation in the trillions is simply not reality. Also......I cannot really see your vision of the future where "AI" will be nearly as foundational as you claim.I think people waiting for the bubble to “burst” are going to wait for a very, very long time. I’m sure the unprecedented growth will eventually slow down, but AI is not going anywhere. I consider this to be the ground floor, if anything. Going forward it will change our lives in ways that we can’t even imagine yet.
I love the novel idea that before 2026 people somehow cared about code quality.If your goal is just to create PC port, sure. The AI can do that. It will probably eventually create a functionally-equivalent but nonsensical codebase full of fakematches (things that compile to the same binary, but are not anywhere close to what was in the original code) and without naming stuff properly.
Said codebase is effectively useless to humans afterwards (and possibly also AI, lacking context from the previous runs), because trying to understand what it does requires you to repeat the reverse engineering effort.
I mean just look at this AI Decomp of Banjo Kazooie. Does any of this code actually tell you anything about how the game works? https://gitlab.com/banjo.decomp/banjo-kazooie/-/blob/master/src/core2/code_C4B0.c
Compare with a decomp done properly:
https://github.com/zeldaret/oot/blob/main/src/code/z_collision_check.c
https://github.com/SMGCommunity/Petari/blob/master/src/Game/AreaObj/CollisionArea.cpp