I agree that there may be a lot of variation between models that leads to different use cases, at least today. But I’m not sure the car analogy works.
An X5 is not simply “inferior” to a CR-V, or vice versa. A Camry is not “inferior” to an F-150, or vice versa. They are optimized for different buyers, budgets, constraints, and use cases.
That may actually be the better analogy for AI models: there probably is not one universal “best” model. There are models that are better or worse for particular tasks, price points, latency requirements, deployment constraints, privacy needs, etc.
In the same way that you cannot compete against Ford at making cars without capital because Ford used their capital to automate manual labor at prices and speeds you'll never be able to reach.
"Intellectual" jobs, that require thought were yet spared, as the only way to use capital for that was just to pay for more meat: see Accenture, CapGemini, IBM, etc. You could carve yourself a spot for some people because it was hard for this capital to reach them. Now that AI, a literal machine that automates thought, is here, the costs for reaching bumfuck nowhere to take the clients away from you has dropped drastically for them, but not for you. They get economies of scale, already established patterns, tools, internal databases. You're starting with hopes, dreams and a $20 Claude sub that runs out at 10AM.
There is not infinite need for lawyers, there being more lawyers does not create more legal opportunities and cases past a certain point, and you're then dependent on other things like there being enough judges.
Robots transformed capital into manual actions and killed most manufacturing jobs save for precious few experts or things that are too expensive to automate still. AI will transform that capital into intellectual labor and crush you, take all opportunities away from you.
Actually there is effectively infinite demand for legal services, in the same way that there is effectively infinite demand for software. In highly regulated industries like healthcare we're constantly backlogged on legal capacity. It's not just "cases" but contract review, sales negotiations, customer documentation, regulatory affairs, civil enforcement activities, etc. Current LLMs are helpful for basic tasks like preliminary legal research and drafting documents but they still make frequent errors and can't carry out complex workflows lasting months that depend on high-context human relationships. Better automation would be a huge unlock to accelerate these business activities.
I will disagree with that from experience. I'm from a small (under a million) country with a surplus of lawyers... it's corrupted them in the sense that there's not enough work around to make money from being efficient, so they legendarily drag out the cases they can, and there's near-monthly trials of cases of theft from trust accounts. So, no "actually".
Again the business demand for legal services has little to do with "cases" as such. Your experience is irrelevant and you're probably ignorant about how business and the legal/regulatory system works in the USA. The vast majority of work done by corporate counsel never has anything to do with trials. The real world isn't like what you see in TV legal dramas.
The price for legal work going down might very well create more demand, including for people operating "law machines". Not sure we know where future equilibria lie.
One recent comment from my interviews was that people who use AI are using it for tasks in domains they didnt deal with before. So this would be creating dashboards or writing sql queries. Or reading and reviewing contracts.
The “easy stuff” for someone’s job, is now the AI stuff for someone else’s job. Where you would hire an intern, the potential client is using Claude instead.
The issue is that this breaks the talent / growth pipeline. You can’t have experts if they don’t go through the process of getting trained and working on incrementally harder problems.
> The “easy stuff” for someone’s job, is now the AI stuff for someone else’s job. Where you would hire an intern, the potential client is using Claude instead.
And what happens when the SQL query has some subtle error or is missing data?
With interns, there’s an implicit understanding that you will spend a bit of extra time reviewing their work and mentoring them. With “just AI it bruh,” there is not.
Demand for law related things isn't elastic. In fact, in an increasingly unemployed, AI first future, work law is a dead end, contract law is a dead end, and there will not be "AI law" jobs created.
"Price go down means more demand" applied blindly is a an economic theory so absurdly shit that even the most apeshit libertarians like Ayn Rand know it isn't true. Don't make me defend Ayn Rand.
Labor law will be hit more by widespread use of robotics, but I can envision a much larger market for contract disputes and transactional law. Having AI in both sides doesn’t mean people won’t disagree about stuff.
Which hasn't changed, because the company you would have sued that was spending 200k on a lawyer is now spending 500k worth of tokens to spend the next consecutive 300 hours without sleep to find out that you took an unauthorised 5.1 minute shit on April 23 at 3PM by reviewing every single hour of camera footage and establishing a physical movement map of MAC addresses based on proximity to the WiFi relays they put every 10 meters.
You. Cannot. Win. Against. Capital.
And then you're going to sue, and realize that the system is so massively overloaded that the next available judge and jury are in 32 years. Also they're lobbying to replace those with AI judges where they entirely removed the concept of nuance.
What is an "AI first future"? Infinitely capable robots and AI? All current laws and regulations suddenly gone or changed?
Why would there be less demand for contract law or for privacy related law, for example? There is certainly some elasticity in law related things from my own experience.
> People start businesses, including law firms, all the time with little to no capital.
Those people are usually very talented and work 16/7. And they need to convince other similar people to work for them, usually at a lower-than-market rate.
The premise here is "talented humans working hard" will stop being a valuable thing due to AI.
If the rate of failure is slightly consistent then more layoffs will lead to more startups, so more will get on top. Many companies business are easier to attack than it looks. Look early 2000, early 2010 have both brought their set of amazing startups. 5 years ago people were most motivated to switch companies for more money.
> Sorry that's no realistic. People start businesses, including law firms, all the time with little to no capital.
You're clinging past: what you is true when human capital counts for something, but what happens when it doesn't? Where the party who can spend the most tokens on the case wins (or has a much greater chance of success)?
Law might not be the best example of that, but (under current trends) a lot of areas will be.
> Your asserting AI makes the economy more capital intensive, when I think you'll see in practice it's the opposite.
You're claiming AI will make the economy more labor intensive? Huh?
Sounds like Jevons Paradox to me: Amount of output per worker-hour increases(cost drops edit sorry), paradoxically worker-hours _increase_.
Mechanism? New Use Cases become viable.
Just like LED lights and Virtual Machines made light-per-watt and workloads-per-server more efficient, what did we do with that efficiency? We didn't pocket the cash, we turned every billboard and many buildings into JumboTrons, and we made millions of new cheap cloud VMs to run hundreds of thousands of new little businesses that never would have happened.
Look at coding now: People are finishing side projects that they never would have, closing out old bugs or test coverage they never would have, starting side businesses they never would have before.
This is new demand being created before our eyes as the cost of knowledge work drops.
It will never happen. Capitalism works when consumers can choose the best service provider. When there is only one universal service provide (AI), behind various "fronts", it would be catastrophic.
And all the consumers lose catastrophically.
So there should be at least some human component to differentiate between the various "AI" providers.
But if the AIs learn from the business (at some point they need to) then the differentiator get's absorbed back. If you figure out how to optimize your AI/humanoid robot, that benefit you have only lasts until the next AI model that incorporates your successes.
But LLMs does not learn/train like that. They need the same thing repeated a gazillion times in the their training data to actually use it in inference.
So unless this human element get absorbed widely into literature, then I think the scenario you present is not possible, at least in the context of current LLMs.
May be there will be true AI in future, that can read a fact and grasp it brilliance, like a human do, and use it in answering future questions. Then what you said could be possible.
I'm assuming for it to work at the level of filling rolls in businesses like this they will need to generate synthetic data to train on from onsite data, which will be the best practices of you and your competitors and hence include the corpus of what makes you better/lean compared to your competitor and remove differentiators.
I think all your points are correct except the AI making the economy less capital intensive. Its a new type of input that is suppose to make you vastly more performant not having it by definition puts you at a disadvantage. I.e you need that capital.
I don't necessarily agree with that AI is going to be as super productive as people think but if your assumption it is. It is another tool you'll need to be competitive. That last point just doesn't follow if you believe in AI being powerful
Lower savings rate, higher inflation, higher interest rates, less capital available for investment. So most capital intensive sectors wouldn't do well.
A lot more volatility, without a (or much smaller one) buffer most economic shocks would have a bigger impact on the economy.
Also there would still be a lot of inequality it would just be intra-generational.
Then again.. all the annuity money has have to go somewhere. So maybe the insurance companies would become the primary sources of investment capital (which wouldn't be great). A lot of uncertainty though i.e. buying a house if you have a family would become much riskier..
I'd need a source to back up those claims, as you note it's not trivial to understand how economy would react.
I also don't see why buying a house would be much riskier? If you buy a house for your family it's because you either prefer the lifestyle or think it provides economic advantages over renting. Given you only need housing when your alive, I think what happens after you pass is not as major a concern as presented.
A source like what? I don't think there are many studies refuting bizarre not well thought out policies. Also it's pretty hard for me to argue against a suggestion that's so ill defined.
Albeit a massive increase in consumption and a reduction in savings would be the most obvious outcome (with all the implications of that).
> after you pass is not as major a concern as presented
Therefore there is no point for you to own your house. When you get older you either get a reverse mortgage or don't buy property in the first place. There would be no rational reason to own property beyond a certain age.
If no sources exist, then you must accept making claims such as this would 'lower savings rates' are simply not backed up. Maybe it will... maybe it won't.
So what if there is no reason to own property beyond a certain age? Even if we take this claim as true... that doesn't explain if this is a good or bad thing.
I don't, because this argument is nonsensical (I mean your point about source specifically). Unless you disagree with some of the core principles of modern economics (not saying that you have to agree with them..) that would be the most obvious outcome.
> So what if there is no reason to own property beyond a certain age?
Well that would mean that the savings rate would go down (for better or for worse).
You've not established that your suggestions are core principles of modern economics or derived from them.
For example, you are asserting there would be 'no reason to own property beyond a certain age'... which isn't supported, and then jumping to the conclusion that that would lower savigns rates.
The general consensus amongst most economists is that humans behave in a rational way? Not spending all the wealth you before you die would be irrational if your children won't inherit it.
Of course in reality that's often not the case especially these days so it might not be sufficient enough.
> clearly true, just supposition.
Well by such standards every discussion on any policy that hasn't been tried is meaningless because there isn't any empirical evidence.
I was demonstrating the Apples to Oranges comparison. If they were both free no one would pick DLSS. It shows Rasterizing is preferable. So comparing Rasterizing performance to DLSS performance is dishonest.
Yeah but they had pretty meagre memory bandwidth until now, aside from the parts they made exclusively for Xbox and Playstation. AMD didn't seem to be interested in bringing fast unified memory to real computers until Apple did it. Now they need to double up the bus again to make an M4 Max-alike...
In part because it's an odd compromise. With the exception of LLM's which are a decent development... there wasn't a lot of need to high memory, but moderate GPU compute parts. You'd either have a lot of memory and a CPU, or a lot of memory and a beefy GPU.
It isn't what the market wanted, and by market I mean OEMs, because I'm sure consumers would have loved it.
The OEMs buying APUs to use in laptops and SFF desktops were more interested in cutting costs than boosting graphics performance. Users who want better 3D performance can buy a higher end laptop with a discrete GPU and juicier profit margin.
True, but apple's the benchmark in this space and have managed thin laptops with good battery life and decent (but not class leading) GPU performance.
Doubling the memory width (and tripling the bandwidth) helped Apple's GPU performance substantially and should do the same for AMD. Which means that a larger fraction of the laptop market should consider it "good enough" and still have a reasonable TDP to avoid the 2" think laptop that last for less than an hour on battery while sounding like a hair dryer.
Those agreements would be >15 years old at this point, I doubt AMD would agree to sandbag their entire APU lineup going forward just to make the base model PS4 and Xbone look good, and I especially doubt that they would do that again with the PS5 and Xbox Series when they actually had money and room to negotiate. Likewise, it doesn't make sense as a restriction Microsoft or Sony would impose: the console business is one of convenience, not power. They aren't trying to beat PCs and they don't care if PCs are a better deal. They care if they can get you to buy a box that locks you into their DRM scheme.
Furthermore, during the chip shortages of the last few years, AMD was actually selling broken PS5 silicon for use as a normal Windows PC[0]. If there were restrictions on selling APUs above a certain performance level, then this PC wouldn't exist.
Historically PC OEMs cared about price more than graphics performance so that's what they got. If you look at mobile SoCs or the eDRAM-equipped SKUs Intel made for Apple you see more emphasis on memory & graphics performance similar to consoles.
One oil well can produce an amount of free energy (24/7) that a 100 acre wind farm can only produce sporadically, assuming the well is a reasonably high volume producer. It depends on the specific well/geology.
The issue with fossil fuels is that they liberate fossil carbon, which has larger macro effects on the global environment. (It injects a lot of ‘new’ carbon into the carbon cycle)
They also do sometimes have some medium sized local effects from spills or contamination. But those can usually be controlled.
Geothermal is also usually ‘low footprint/high value’, but is only viable in specific limited locations.
Solar, wind, hydropower, tidal energy all have large physical footprints for the amount of energy they produce. Aka ‘low density’. All are also somewhat tied to specific, and often limited geology.
For solar for instance, areas with a lot of desert or other open ‘non productive’ land nearby, it’s great (assuming decent insolation). In areas where land is at a premium for other uses, or is very rugged/high maintenance, it definitely is a problem. Aka cities, certain types of high intensity farmland, heavily forested areas, high snow load/storm areas, etc.
Solar is not an awesome economic choice in Siberia, for example. It is an awesome economic choice in Southern California, Arizona, Nevada, etc.
For areas with geography that supports it (typically the right kind of mountain ranges) and rainfall, hydropower is awesome, though has serious side effects on wildlife and river health. For a place that doesn’t have the right geography (say England), it’s a non starter.
> One oil well can produce an amount of free energy (24/7) that a 100 acre wind farm can only produce sporadically, assuming the well is a reasonably high volume producer. It depends on the specific well/geology.
Except it can't, a 100 acre wind farm can produce energy indefinitely while a oil well will eventually run dry.
The idea that fossil fuels are more ecologically favorable because it's 'dense' needs to address not only external factors, but that fossil fuels are non-renewable.
> Except it can't, a 100 acre wind farm can produce energy indefinitely while an oil well will eventually run dry.
Perhaps more true in that the wind (as far as we know) won’t run out but wind turbines do have a limited lifespan. After 20-30 years they usually need to be replaced. Some of the components are recycled but a significant portion - including the blades - are either not recyclable or not economically recyclable. Work is being done on this but there’s no guarantee it’ll produce dividends.
It depends entirely on what scope one considers for impact. Do we count maintenance roads? Total land area disturbed? Windmill foundation pads? Global co2 levels? Abandoned equipment in general? Noise levels over how much area?
What is ‘leftover’ from an abandoned well can be as simple as a buried 6” ground level plug, or as messy as an acre of abandoned equipment and a giant oil spill/hazmat area. Plus a billion tons of atmospheric co2 - which is invisible.
Oil is so widely used because it is incredibly cheap and easy to use at large scale, with minimal obviously visible consequence.
Because co2 is invisible. And as long as we don’t spill large quantities of it, it doesn’t seem to cause any visible problems.
The effect of the low density from wind, solar, etc. isn’t visible until you go to areas it is widely deployed and then do the math on how much energy they are actually producing, which is a small fraction of what would be produced if the same area was impacted to produce oil or nuclear.
"The movies that play in virtual screens are native to the films’ aspect ratios, which can vary movie to movie, eliminating the black bars of “letterboxing” and “pillarboxing” you typically have on iPads, iPhones, or MacBooks."
Then proceeds to show screenshots where more than 50% of the screen is background. Just because you decided to show cloud instead of black doesn't mean it's not letterboxing.
The screenshots are misleading. "Inside" an AVP watching a movie the corner to corner angle can exceed THX recommendation of 40 degrees.
Doesn't seem like a lot until you hold a protractor up to your eyeball. And then you realize how much area falls outside that angle, surroundings you don't think about at the theater.
The AVP has to render that unwatchable area for you as well, and it ends up in screenshots.
Lol, adding a hue is usually a crutch for a too small screen with lifted blacks... which doesn't paint the pro in a very positive light!
Ironically by adding a gue to a nomral screen, you're effectively 'adding' pixels to the display... unlike the pro which wastes pixels on the 'hue' because many of them are not in the ideal viewing angle.
An X5 is not simply “inferior” to a CR-V, or vice versa. A Camry is not “inferior” to an F-150, or vice versa. They are optimized for different buyers, budgets, constraints, and use cases.
That may actually be the better analogy for AI models: there probably is not one universal “best” model. There are models that are better or worse for particular tasks, price points, latency requirements, deployment constraints, privacy needs, etc.
reply