The A.I. Thread

[Fuzz]

Originally Posted by calgarygeologist

Beautiful and a fact that can't be argued by anyone.

[opendoor]

Originally Posted by CliffFletcher

Housing is bad atm. But as opendoor has shown, it wasn't until the covid surge that housing became less affordable than it was in the early 80s.

There are of course some caveats with that. That was strictly comparing the mortgage payment on the average house vs. median wages through time. It's important to keep in mind that:

1) The early '80s was not the norm. Outside of that point and the 1989-1991 period, housing has always been more affordable than it has been the last 5 years, sometimes significantly so.

2) The unaffordability then was driven by high interest rates, not high prices, so it wasn't as hopeless as it may seem now. When you could buy a house for 2-3 years' salary, a good saver working the average job could buckle down and make it happen regardless of what interest rates were doing. Now that houses are 7-8x the median income, it's a lot tougher with less room for error. For instance, in some cities, it would take as long to save up to buy a house for cash in the early '80s as it would to save up for just the down payment now.

The point of the comparison isn't to show that current housing is affordable, but rather that we've been where we are now before and it has corrected over time through stagnant prices and inflation.

[Geraldsh]

What has grown astronomically also is the size that a livable home should be. Does every child really need their own room complete with desk and tv besides the family room, theatre room, play room, dining room, …

[ripTDR]

Originally Posted by Geraldsh

What has grown astronomically also is the size that a livable home should be. Does every child really need their own room complete with desk and tv besides the family room, theatre room, play room, dining room, …

Disagree. Grew up in the 80's in a typical 80's household suburb. Father worked full time mom worked part time. House had a formal living room, kitchen, dining room, family room, three bedrooms and a basement laundry room with a two car garage. Our bedrooms were bigger than condo den's of the nineties which are bigger than most townhouse condo bedrooms now.

[Geraldsh]

Originally Posted by ripTDR

Disagree. Grew up in the 80's in a typical 80's household suburb. Father worked full time mom worked part time. House had a formal living room, kitchen, dining room, family room, three bedrooms and a basement laundry room with a two car garage. Our bedrooms were bigger than condo den's of the nineties which are bigger than most townhouse condo bedrooms now.

You were rich! Like 10 percenter or better! Good for you

[Russic]

Originally Posted by Fuzz

That's why it's such a misnomer to call what we have now AI. It's not intelligent. That's not to say it isn't useful for a lot of things, but it doesn't posses intelligence. It posses vast amounts of information that it can do a lot with. But it has zero comprehension, and this is proven out time and time again. If you think it has fixed that, please give it spatial reference challenges like maps, directions, orientation. And it will be hyper-confident it is correct, while being consistently wrong.


I just tried it again the other day to OCR an old data table, and it failed spectacularly in multiple ways. It saw pastern in the first few numbers, and instead of trying to read the rest, it would just run with the pattern it "found". Some times it would do a great job on really hard to read numbers, but then apply the pattern to the decimal. And my instruction to flag uncertainties? Completely ignored until the last line, where it tossed a couple, then just...didn't do the rest of the file. I could probably have kept fighting with it, but it was quicker to type the numbers myself.


There is no intelligence on display here, and if you think there is because you use case shows it, you need to explain why it it fails so badly on basic intelligence type operations, like spatial awareness. It can't be both intelligent and inept. So the appearance of intelligence is just that, facade humans have convinced themselves is true. And it makes perfect sense when you understand a bit of how they are programmed and function, and that there is no path from the way these models are constructed to actual real AI. Great marketing, though!

The question I keep coming back to: does it matter? Does it need to emulate intelligence as humans define it in order to change everything?

Sometimes it feels like we’re bees watching the pyramids being built, and we’re unimpressed because the things constructing them can’t see UV light.

If humans experimentally solved about 185,000 protein structures by 2021 and AI can generate 200 million plus predicted structures today, does it really matter if it occasionally trips over riddles or miscounts the letters in “strawberry”? We don’t really need these systems to solve those types of problems, because they’re not real problems.

[Cain]

I feel like people are missing the boat in evaluating AI a lot of the time. It isn't perfect, and yes, it currently fails (sometimes spectacularly) at things that a human would be less likely to.

However, the progress that has been seen since GPT3 made huge waves a couple years ago is pretty staggering. I think many people tried it once or twice a long time ago and haven't paid much attention since. It isn't "yet" at the point where it is going to replace jobs en masse, but the signs of that coming are definitely there. I think it is easy to overlook how fast it is improving if you aren't watching closely, or in one of the fields that is really being targeted.

I have a buddy who is a very talented developer/software engineer and the increase in productivity for him has been immense. It has gone to "maybe it can help reduce some tedious work/boilerplate" to "it reliably one shots most things I throw at it". And it is becoming more and more the case, all the time where it takes on larger and larger workloads.

With the focus of the tech companies on having it be a reliable software engineer, I think that is the first spot that we start seeing big impacts. I believe we already are, just limited to the junior level mostly. If it gets to the point where it an reliably assist in its own development...that is where the takeoff is going to happen, and accelerate extremely fast from there.

I'm not a developer or software engineer, and I've used it to great effect in having it assist me in every day sort of office work. Not in an automated way, but in a way where it enables me to learn/create useful tools and reports that would have taken a ton more time if I had to do it unassisted. It is already a substantial tool, and I'd argue that it is still in its infancy outside of select targeted niches...and things have the possibility of getting a bit crazy in the next 2-3 years.

[Bill Bumface]

Originally Posted by Geraldsh

What has grown astronomically also is the size that a livable home should be. Does every child really need their own room complete with desk and tv besides the family room, theatre room, play room, dining room, …

Agreed, this trend must swallow a massive amount of the capacity to build new homes:

[Fuzz]

Originally Posted by Russic

The question I keep coming back to: does it matter? Does it need to emulate intelligence as humans define it in order to change everything?

Sometimes it feels like we’re bees watching the pyramids being built, and we’re unimpressed because the things constructing them can’t see UV light.

If humans experimentally solved about 185,000 protein structures by 2021 and AI can generate 200 million plus predicted structures today, does it really matter if it occasionally trips over riddles or miscounts the letters in “strawberry”? We don’t really need these systems to solve those types of problems, because they’re not real problems.

I think it does matter. AI implies it can substitute for human thinking. And if it is amazing and capable in some areas, it most have those same capabilities in all areas. And it is amazing at some things. Better than humans. But not universal. This leads to it being used, or demanded to be used in areas it should not, because it can not. This leads to all sorts of unnecessary problems.

If this results in a large economic crash when the promises can't match reality, that's a pretty big failure, right? and then it will set back work, as all tech does when it doesn't live up to promises(how's that 3D thing going?). It also makes us miss the picture of what real AI actually means for humanity. The average person won't care, they'll just think they were duped the first time.

There is nothing wrong with celebrating what we have created, it can be incredible. But the name starts from a false premise, and hides behind it. I dunno, I just find deceit to be a pretty poor way to usher in new technology. We see how humanity reacts to these things in the past, and we should be honest about what it is and isn't.

[Russic]

Originally Posted by Fuzz

I think it does matter. AI implies it can substitute for human thinking. And if it is amazing and capable in some areas, it most have those same capabilities in all areas. And it is amazing at some things. Better than humans. But not universal. This leads to it being used, or demanded to be used in areas it should not, because it can not. This leads to all sorts of unnecessary problems.

If this results in a large economic crash when the promises can't match reality, that's a pretty big failure, right? and then it will set back work, as all tech does when it doesn't live up to promises(how's that 3D thing going?). It also makes us miss the picture of what real AI actually means for humanity. The average person won't care, they'll just think they were duped the first time.

There is nothing wrong with celebrating what we have created, it can be incredible. But the name starts from a false premise, and hides behind it. I dunno, I just find deceit to be a pretty poor way to usher in new technology. We see how humanity reacts to these things in the past, and we should be honest about what it is and isn't.

That you’re comparing ai to 3D TVs is bold, and I respect it. I don’t agree, but ya, you’re correct, a lot of tech ends up being hype.

But so far, month over month for years, I don’t believe we’re seeing that here. I’m pretty sure when this thread started it couldn’t manage multiplication tables. Now it’s solving phd-level math but we don’t care… we just move on to the next thing. It couldn’t count r’s in “strawberry” but now it can do the wordle (the only metric that matters). Again, we don’t raise an eyebrow.

When this thing becomes indispensable to family doctors and saves lives, will anybody care? I think not, because it’s already doing that but we don’t talk about it for some weird reason.

That’s the strangeness for me. Everybody has some arbitrary line in the sand and says “when it can do _this_, I’ll be convinced.” Then it does that and they draw a new line. It’s been happening since before Turing and I don’t get it. We actually believe there’s been this steady march of progress since we invented fire, but today it’s going to come to a full stop?

[opendoor]

Originally Posted by Bill Bumface

Agreed, this trend must swallow a massive amount of the capacity to build new homes:

Not really, because that looks only at detached houses. Condos/apartments make up a larger share of units than in the past, so the average size per newly constructed housing unit is only up about 20% since the '70s.

They could build smaller detached houses, but that isn't really economical viable given how costs scale. A new 1,250 sq ft detached house might cost 80% of what a 2,500 sq ft one would when you consider land/site/development costs, so they're not really attractive propositions.

[CliffFletcher]

Originally Posted by Bill Bumface

Agreed, this trend must swallow a massive amount of the capacity to build new homes:

Which is compounded by the number of people per household getting lower and lower over recent decades (3.5 in 1971 vs 2.5 in 2021).

It takes a lot more housing units (including the smaller apartments and condos opendoor referenced) to house a million Canadians today than it did 50 years ago.

[CliffFletcher]

An alarming take on the speed and scope of job losses the rapid adoption of AI might cause.

https://www.theringer.com/podcasts/p...ill-do-to-jobs

[Russic]

Originally Posted by CliffFletcher

An alarming take on the speed and scope of job losses the rapid adoption of AI might cause.

https://www.theringer.com/podcasts/p...ill-do-to-jobs

I do think there's a fascinating collision coming between a very fast technology and very slow moving businesses. For a lot of people, Christmas was the first time they sat down and tried something like claude code or codex (because they had time to screw around).

Watching it come up with a 30 point action plan to accomplish a task, and then watching it not only attack each step but also check itself along the way is a fascinating and mildly terrifying experience. When you consider how many jobs are just 30 point action plans, things get uncomfy.

[White Out 403]

https://www.404media.co/goldman-sach...nd-unreliable/

Goldman Sachs: AI Is Overhyped, Wildly Expensive, and Unreliable

Investment giant Goldman Sachs published a research paper about the economic viability of generative AI which notes that there is “little to show for” the huge amount of spending on generative AI infrastructure and questions “whether this large spend will ever pay off in terms of AI benefits and returns.”

The paper, called “Gen AI: too much spend, too little benefit?” is based on a series of interviews with Goldman Sachs economists and researchers, MIT professor Daron Acemoglu, and infrastructure experts. The paper ultimately questions whether generative AI will ever become the transformative technology that Silicon Valley and large portions of the stock market are currently betting on, but says investors may continue to get rich anyway. “Despite these concerns and constraints, we still see room for the AI theme to run, either because AI starts to deliver on its promise, or because bubbles take a long time to burst,” the paper notes.

[PepsiFree]

Originally Posted by Russic

I do think there's a fascinating collision coming between a very fast technology and very slow moving businesses. For a lot of people, Christmas was the first time they sat down and tried something like claude code or codex (because they had time to screw around).

Watching it come up with a 30 point action plan to accomplish a task, and then watching it not only attack each step but also check itself along the way is a fascinating and mildly terrifying experience. When you consider how many jobs are just 30 point action plans, things get uncomfy.

The things it does really well is basically any job primarily based on data that can be completed online. Which sounds very general because it is. It has effectively replaced everything from technical writers to developers to accountants etc, it just now depends on adoption rates.

A lot of jobs will remain, I think, because people do like to work with people. But I also think along with pre-existing jobs shifting to AI you’ll find people never being hired for the job in the first place.

[Firebot]

Originally Posted by Fuzz

If this results in a large economic crash when the promises can't match reality, that's a pretty big failure, right? and then it will set back work, as all tech does when it doesn't live up to promises(how's that 3D thing going?).

You are really comparing AI to the 3D TV market? That is a really poor comparison. A gimmicky niche product that was always seen as a gimmicky niche and largely rejected by the public from the start versus a revolutionary defining technology that is replacing jobs and how we work.

The dot com bubble crashed in 2000 and you certainly cannot call the internet a failure. I still remember the days of free dial up, massive storage, etc. AI in its current form at its base is here to stay, regardless of if we ever achieve AGI.

Now that doesn't mean there won't be casualties or a bubble deflating. OpenAI is the most likely big domino to fall, burning cash with no prospects to profitability and companies including Microsoft and Apple moving away from them.

I think it does matter. AI implies it can substitute for human thinking. And if it is amazing and capable in some areas, it most have those same capabilities in all areas. And it is amazing at some things. Better than humans. But not universal. This leads to it being used, or demanded to be used in areas it should not, because it can not. This leads to all sorts of unnecessary problems.

Never let logic get in the way of an executive seeing an opportunity to cut and get a performance bonus and stock uptick.

There's many examples of this in the past, especially offshoring, and 5% arbitrary cut policy to lowest performers, or forcing subscription models on consumers. Some companies learn from their mistakes, while others truck along continuing because of monopolies and being stuck with the offerings.

https://www.cnbc.com/2025/01/14/meta...f-layoffs.html

When Microsoft is breaking Windows after boasting that AI is doing much of the coding now, and adding AI to Notepad of all things, it's not going to get better any time soon.

https://www.windowslatest.com/2026/0...s-rolling-out/

https://www.cnbc.com/2025/04/29/saty...ten-by-ai.html

https://futurism.com/artificial-inte...-security-flaw


In the end though, these companies will continue and force these features until the next executives get in line for their turn and the en####ifying will continue. I just don't see it disappearing, only growing and there is no avoiding it.

[opendoor]

Originally Posted by CliffFletcher

Which is compounded by the number of people per household getting lower and lower over recent decades (3.5 in 1971 vs 2.5 in 2021).

It takes a lot more housing units (including the smaller apartments and condos opendoor referenced) to house a million Canadians today than it did 50 years ago.

But by and large, those housing units exist. There were 290 houses per 1,000 people in Canada in 1971 with an average household size of 3.5. At today's average household size of 2.5, we'd need 422 houses per 1,000 to maintain the exact same ratio. As of the last census (which coincided with the peak in real housing prices), we had 411 houses per 1,000, so it's basically unchanged.

Yeah there is a very small deficit relative to 1971, but that same deficit existed in the early '00s when we had some of the lowest priced housing in modern history. It's a combination of financial demand (i.e. investors accepting terrible cap rates on the hope of price appreciation) and the lack of space for detached housing in cities that is driving unaffordability, not the number of households relative to housing stock.

[opendoor]

Originally Posted by Firebot

In the end though, these companies will continue and force these features until the next executives get in line for their turn and the en####ifying will continue. I just don't see it disappearing, only growing and there is no avoiding it.

I mean, they basically have to force them given the insane capital expenditures a lot of these companies are doing right now. Google for instance is basically committing 100% of its 2026 free cash flow to capital expenditures, primarily on AI. So they need to show a return on that by getting paying customers to use it.

Apple is a notable exception. They're spending very little, likely on the basis that either the investments will never pay off, or AI will just become a commodity that they can pay for down the road. Which is probably a decent bet for them.

[GGG]

I think people often conflate twoseparate conversations when talking about AI.

1) AI will displace significant amounts of labour in a relatively short time frame
2) AI has been over capitalized and under delivering what was promised. — AGI

Both of these can be true at the same time and likely are.

Yes there is likely a bubble that will burst and do significant economic harm but there will also be a few companies coming out of this that dominate the future and the labour make up of that future will be different.