7 ms·

If AI output is 60% as good as a person, but costs 1/1000th the price... what decision maker is going to justify paying 1000x more for 40% more quality? We're

If AI output is 60% as good as a person, but costs 1/1000th the price... what decision maker is going to justify paying 1000x more for 40% more quality? We're really in for it. What if there was a cheeseburger that was 50% as good as McDonalds but was only a nickel? That'd be real bad for the McD's share price.

fauxpause_ 4y ago
[flagged]
brailsafe 4y ago
I don't think I'd accept anything worse than McDonald's if they paid me to eat it.
- Gigachad 4y ago
  McDonald’s already is that product. It costs 1/4th as much as a restaurant burger. And half the time it’s just as good.
  - wnevets 4y ago
    [flagged]
  - sockaddr 4y ago
    You need to find better burger places if McDonalds is just as good half the time.
    - Gigachad 4y ago
      I have a list of known good ones but unfortunately it seems like picking any random place will quite often make something worse than McDonalds which is a pretty baseline decent burger. I've been told though that McDonalds in Australia is higher quality than other countries so this might affect things.
      - midoridensha 4y ago
        You need to try McDonald's in Japan. It's not gourmet of course, but at least it looks like the pictures on the menu and is put together exactly the way it's supposed to be.
  - anonzzzies 4y ago
    Don’t live in the US, but most cooks here don’t like making burgers or any type of ‘fast food’, so when they have burgers, it’s usually a beer bar and the burgers are much worse than McDonald’s. And personally I find McDonalds inedible, let alone most of the rest.
- maximus-decimus 4y ago
  That was my conclusion last time I ate at Harveys. The "meat" was just plain bad.
MPSimmons 4y ago
Agreed, as long as the AI crosses whatever perceived threshold for quality you need.
mjr00 4y ago
> If AI output is 60% as good as a person, but costs 1/1000th the price... what decision maker is going to justify paying 1000x more for 40% more quality? We're really in for it. Really depends on the type of output. For content marketing that most people are just going to skim over anyway (aka is just going to be used for blogspam/content farming, which is what it sounds like the linked writer was doing for this client), yes, it absolutely makes financial sense. But a lot of output is very much winner-take-all: music, movies, TV. Only the top 1%, for some definition of top, of that stuff is actually purchased in meaningful quantities. 60% of top quality of that is worth the same as 0%, which is nothing. > What if there was a cheeseburger that was 50% as good as McDonalds but was only a nickel? That'd be real bad for the McD's share price. Not so sure about this; you can look at, for example, generic store-brand colas, which taste exactly the same as Coke and Pepsi and are cheaper, yet Coke and Pepsi haven't collapsed after decades.
- chii 4y ago
  > generic store-brand colas, which taste exactly the same as Coke and Pepsi and are cheaper, yet Coke and Pepsi haven't collapsed after decades. which is because coke & pepsi uses massive advertising campaigns to ensure the brand is doing the heavy lifting. Coke is basically a marketing company that happens to sell a liquid drink.
  - no_wizard 4y ago
    I would also argue, that they in fact, do not taste the same, even in double blind tests I’ve conducted in this area for college
    - TechnicolorByte 4y ago
      Even Pepsi and Coca-Cola have obvious differences with many preferring the latter. Store-brand cola is an even bigger difference.
  - ghayes 4y ago
    Marketing aside, power-law of costs is fairly common. If a good guitar runs you $300 and a professional guitar runs you $1500, a custom guitar for $40K is still only going to be marginally better.
gnicholas 4y ago
Also depends on how much skill/time it takes to transform AI output to something comparable to what a great copywriter could create. Maybe the output is 60% but in 10 minutes you can make it 90%.
- anonzzzies 4y ago
  Yeah, I think that’s the main point; if you have it output and just pop it online then it will be crap. However, a skilled writer will have it to great content in minutes after. That’s the difference. So if the person in question had a chance, they would offer to use gpt before the employer mentions it, to cut costs and take over the entire content and translation pipeline for that small company. Kill or be killed, so to say.
onlyrealcuzzo 4y ago
What if a flight only cost a nickel, but there was a 40% chance it blew up? Not all errors are equal. Sometimes 99% as good isn't even close to good enough. But, yes, for most things 60% for 1/100th the price, is going to cut it.
- no_wizard 4y ago
  I’d say you’re flying Peruvian air in the 1970s
- jabbany 4y ago
  > What if a flight only cost a nickel, but there was a 40% chance it blew up? Everything has failure modes though so this isn't nearly as easy of a comparison... Just as an example on the analogy, there are plenty of cases of pilot suicide bringing down planes too that one would expect would never happen with pure automation... So you can just as easily turn the analogy around and say people are the "cheap but worse" option. FWIW, we have been able to create processes to manage risks with unpredictable humans, why couldn't we with AI ¯\\_(ツ)_/¯
  - iudqnolq 4y ago
    > So you can just as easily turn the analogy around and say people are the "cheap but worse" option. No, because this is a factual question with a factual answer. You need to actually know which is more dangerous. > FWIW, we have been able to create processes to manage risks with unpredictable humans, why couldn't we with AI ¯\\_(ツ)_/¯ Because humans and AI are different? We are much better at predicting when humans will be wrong than with AI, for one.
- xbmcuser 4y ago
  Give this option to the thousands that take rickty boats from Asia and Africa for a better life. All your flights will be full.
- Ekaros 4y ago
  60% reliable car at 1/1000th of price might not be horrible deal. 13-50 vs 13000-50000 and surely you will accept quite many break downs. Even if you need to call taxi and buy new car...
a4isms 4y ago
This was the premise of McDs to begin with: They sold burgers for a dime or whatever while everyone else was charging fifty cents. The trick of drastically undercutting the market, of course, is to use innovation cut your costs, which the brothers did by serving such a limited menu that they could optimize the prep for speed.
strken 4y ago
This is already the bargain with intellectual property. What if you can get writing 50% as good as Stephen King? Doesn't matter, because the winner takes all and books are infinitely copyable. What if you can run a 400m in 80% of the world record time? Doesn't matter, because you'll only be paid a pittance. What if copilot can write code with only 50% more bugs than an engineer? Doesn't matter if each bug costs an average of tens of millions. I don't think it's good vs bad, but quality vs volume. If your output is already able to be duplicated at low-to-zero cost then quality is important.
- detaro 4y ago
  > What if you can run a 400m in 80% of the world record time? congratulations, you are now the new world record holder ;)
- robotresearcher 4y ago
  Stephen King is a weird example because there are lots of working writers in the store brand Stephen King market.
  - petre 4y ago
    Only they don't write like Stephen King. And if an AI was trained to do it (by obviously using SK books), Stephen King would probably sue.
- listenallyall 4y ago
  Stephen King writes faster than ChatGPT can.
- UncleEntity 4y ago
  > What if copilot can write code with only 50% more bugs than an engineer? Is it only in the “AI gonna take my job” discussions that programmers insist on calling themselves “engineers” or is this just a recent phenomena? A bit off topic (but not really) and I know I’ll get downvoted (because, yeah…) but somehow the tide has turned from “programmer” to “software engineer” to just “engineer”. Just think it’s strange.
famouswaffles 4y ago
Creative output from GPT-4 like everything else is a huge jump from 3.5. People say 60% as good but don't be surprised if it's already much closer or even surpassing the human baseline in question.
ulfw 4y ago
What if surgeons get replaced by better robotics hardware (more accurate cutting techniques than a human hand) and better AI software. The premise that a more experience surgeon is a lot better than a newbie surgeon holds true. How do you become an experienced surgeon? By seeing a lot of cases and knowing what to do when things don't go as planned. What does AI do? It learns over time by seeing more and more cases. Will it replaced a true star surgeon within the next decades? Likely not. But how many of us have access to the world's best surgeon? Or the world's top 10 or top 1000? Vast majority see one who is local and good enough (or have no choice if you land in an ER with an accident). Those are highly skilled highly trained positions earning high pay. If you asked me a year ago I would have recommended kids to go into studying medicine and become a surgeon as that surely wouldn't be replaced by computers. Today I am not so sure this will be a profession for the majority of cases (that don't get to see the highest of skilled) even in 10 or 15 years from now.
- xyzzy123 4y ago
  Doctors have an unusually strong professional monopoly. In much of the world they regulate to maintain perpetual under-supply and to ensure they are inserted as bottlenecks in strategic/lucrative places. When AI comes for health care you can bet that doctors will ensure they have a key role in service delivery. (BTW: doctors are great! I just wish we had more of them and that on the whole they were less over-worked). Personally I would not be worrying too much for the doctors.
  - bufferoverflow 4y ago
    > Doctors have an unusually strong professional monopoly. That won't matter. Once people find out AI/robotic solutions are better than the best human doctors, most will choose AI/robots, they will demand them. People aren't stupid.
  - midoridensha 4y ago
    The OP is talking about surgeons, not doctors. The two are entirely different professions; it's like conflating doctors with nurses, or doctors with medical researchers.
jwie 4y ago
Small objective differences create vast differences in outcomes. Those things where that isn't true have already been automated away or commodified so there's very little margin left anyway. It might be true in a misleading sense that the AI {product} is some large fraction of a genuine {product}, but that doesn't mean anything because the value is contained in those differences above the AI produced baseline. The value floor, so to speak, is a lot higher, but the ceiling hasn't risen an inch. Those that move down in quality to preserve margin will soon discover there's not a there there.
pnt12 4y ago
Where I live, lots of burger restaurants opened up in the last years. Emphasis on the restaurant part, not fast food like McDonald's or Burger King. I very much prefer them, as the food is better and the places are cozier, even though a bit more expensive. Maybe we'll get a counter movement soon: "made by humans", as in, the premium experience. Anti AI sentiment may also grow and make that even more viable, as peoplr may boycot heavy-AI shops.

fauxpause_ 4y ago

[flagged]

brailsafe 4y ago

I don't think I'd accept anything worse than McDonald's if they paid me to eat it.

Gigachad 4y ago
McDonald’s already is that product. It costs 1/4th as much as a restaurant burger. And half the time it’s just as good.
- wnevets 4y ago
  [flagged]
- sockaddr 4y ago
  You need to find better burger places if McDonalds is just as good half the time.
  - Gigachad 4y ago
    I have a list of known good ones but unfortunately it seems like picking any random place will quite often make something worse than McDonalds which is a pretty baseline decent burger. I've been told though that McDonalds in Australia is higher quality than other countries so this might affect things.
    - midoridensha 4y ago
      You need to try McDonald's in Japan. It's not gourmet of course, but at least it looks like the pictures on the menu and is put together exactly the way it's supposed to be.
- anonzzzies 4y ago
  Don’t live in the US, but most cooks here don’t like making burgers or any type of ‘fast food’, so when they have burgers, it’s usually a beer bar and the burgers are much worse than McDonald’s. And personally I find McDonalds inedible, let alone most of the rest.
maximus-decimus 4y ago
That was my conclusion last time I ate at Harveys. The "meat" was just plain bad.

MPSimmons 4y ago

Agreed, as long as the AI crosses whatever perceived threshold for quality you need.

mjr00 4y ago

> If AI output is 60% as good as a person, but costs 1/1000th the price... what decision maker is going to justify paying 1000x more for 40% more quality? We're really in for it. Really depends on the type of output. For content marketing that most people are just going to skim over anyway (aka is just going to be used for blogspam/content farming, which is what it sounds like the linked writer was doing for this client), yes, it absolutely makes financial sense. But a lot of output is very much winner-take-all: music, movies, TV. Only the top 1%, for some definition of top, of that stuff is actually purchased in meaningful quantities. 60% of top quality of that is worth the same as 0%, which is nothing. > What if there was a cheeseburger that was 50% as good as McDonalds but was only a nickel? That'd be real bad for the McD's share price. Not so sure about this; you can look at, for example, generic store-brand colas, which taste exactly the same as Coke and Pepsi and are cheaper, yet Coke and Pepsi haven't collapsed after decades.

chii 4y ago
> generic store-brand colas, which taste exactly the same as Coke and Pepsi and are cheaper, yet Coke and Pepsi haven't collapsed after decades. which is because coke & pepsi uses massive advertising campaigns to ensure the brand is doing the heavy lifting. Coke is basically a marketing company that happens to sell a liquid drink.
- no_wizard 4y ago
  I would also argue, that they in fact, do not taste the same, even in double blind tests I’ve conducted in this area for college
  - TechnicolorByte 4y ago
    Even Pepsi and Coca-Cola have obvious differences with many preferring the latter. Store-brand cola is an even bigger difference.
- ghayes 4y ago
  Marketing aside, power-law of costs is fairly common. If a good guitar runs you $300 and a professional guitar runs you $1500, a custom guitar for $40K is still only going to be marginally better.

gnicholas 4y ago

Also depends on how much skill/time it takes to transform AI output to something comparable to what a great copywriter could create. Maybe the output is 60% but in 10 minutes you can make it 90%.

anonzzzies 4y ago
Yeah, I think that’s the main point; if you have it output and just pop it online then it will be crap. However, a skilled writer will have it to great content in minutes after. That’s the difference. So if the person in question had a chance, they would offer to use gpt before the employer mentions it, to cut costs and take over the entire content and translation pipeline for that small company. Kill or be killed, so to say.

onlyrealcuzzo 4y ago

What if a flight only cost a nickel, but there was a 40% chance it blew up? Not all errors are equal. Sometimes 99% as good isn't even close to good enough. But, yes, for most things 60% for 1/100th the price, is going to cut it.

no_wizard 4y ago
I’d say you’re flying Peruvian air in the 1970s
jabbany 4y ago
> What if a flight only cost a nickel, but there was a 40% chance it blew up? Everything has failure modes though so this isn't nearly as easy of a comparison... Just as an example on the analogy, there are plenty of cases of pilot suicide bringing down planes too that one would expect would never happen with pure automation... So you can just as easily turn the analogy around and say people are the "cheap but worse" option. FWIW, we have been able to create processes to manage risks with unpredictable humans, why couldn't we with AI ¯\\_(ツ)_/¯
- iudqnolq 4y ago
  > So you can just as easily turn the analogy around and say people are the "cheap but worse" option. No, because this is a factual question with a factual answer. You need to actually know which is more dangerous. > FWIW, we have been able to create processes to manage risks with unpredictable humans, why couldn't we with AI ¯\\_(ツ)_/¯ Because humans and AI are different? We are much better at predicting when humans will be wrong than with AI, for one.
xbmcuser 4y ago
Give this option to the thousands that take rickty boats from Asia and Africa for a better life. All your flights will be full.
Ekaros 4y ago
60% reliable car at 1/1000th of price might not be horrible deal. 13-50 vs 13000-50000 and surely you will accept quite many break downs. Even if you need to call taxi and buy new car...

a4isms 4y ago

This was the premise of McDs to begin with: They sold burgers for a dime or whatever while everyone else was charging fifty cents. The trick of drastically undercutting the market, of course, is to use innovation cut your costs, which the brothers did by serving such a limited menu that they could optimize the prep for speed.

strken 4y ago

This is already the bargain with intellectual property. What if you can get writing 50% as good as Stephen King? Doesn't matter, because the winner takes all and books are infinitely copyable. What if you can run a 400m in 80% of the world record time? Doesn't matter, because you'll only be paid a pittance. What if copilot can write code with only 50% more bugs than an engineer? Doesn't matter if each bug costs an average of tens of millions. I don't think it's good vs bad, but quality vs volume. If your output is already able to be duplicated at low-to-zero cost then quality is important.

detaro 4y ago
> What if you can run a 400m in 80% of the world record time? congratulations, you are now the new world record holder ;)
robotresearcher 4y ago
Stephen King is a weird example because there are lots of working writers in the store brand Stephen King market.
- petre 4y ago
  Only they don't write like Stephen King. And if an AI was trained to do it (by obviously using SK books), Stephen King would probably sue.
listenallyall 4y ago
Stephen King writes faster than ChatGPT can.
UncleEntity 4y ago
> What if copilot can write code with only 50% more bugs than an engineer? Is it only in the “AI gonna take my job” discussions that programmers insist on calling themselves “engineers” or is this just a recent phenomena? A bit off topic (but not really) and I know I’ll get downvoted (because, yeah…) but somehow the tide has turned from “programmer” to “software engineer” to just “engineer”. Just think it’s strange.

famouswaffles 4y ago

Creative output from GPT-4 like everything else is a huge jump from 3.5. People say 60% as good but don't be surprised if it's already much closer or even surpassing the human baseline in question.

ulfw 4y ago

What if surgeons get replaced by better robotics hardware (more accurate cutting techniques than a human hand) and better AI software. The premise that a more experience surgeon is a lot better than a newbie surgeon holds true. How do you become an experienced surgeon? By seeing a lot of cases and knowing what to do when things don't go as planned. What does AI do? It learns over time by seeing more and more cases. Will it replaced a true star surgeon within the next decades? Likely not. But how many of us have access to the world's best surgeon? Or the world's top 10 or top 1000? Vast majority see one who is local and good enough (or have no choice if you land in an ER with an accident). Those are highly skilled highly trained positions earning high pay. If you asked me a year ago I would have recommended kids to go into studying medicine and become a surgeon as that surely wouldn't be replaced by computers. Today I am not so sure this will be a profession for the majority of cases (that don't get to see the highest of skilled) even in 10 or 15 years from now.

xyzzy123 4y ago
Doctors have an unusually strong professional monopoly. In much of the world they regulate to maintain perpetual under-supply and to ensure they are inserted as bottlenecks in strategic/lucrative places. When AI comes for health care you can bet that doctors will ensure they have a key role in service delivery. (BTW: doctors are great! I just wish we had more of them and that on the whole they were less over-worked). Personally I would not be worrying too much for the doctors.
- bufferoverflow 4y ago
  > Doctors have an unusually strong professional monopoly. That won't matter. Once people find out AI/robotic solutions are better than the best human doctors, most will choose AI/robots, they will demand them. People aren't stupid.
- midoridensha 4y ago
  The OP is talking about surgeons, not doctors. The two are entirely different professions; it's like conflating doctors with nurses, or doctors with medical researchers.

jwie 4y ago

Small objective differences create vast differences in outcomes. Those things where that isn't true have already been automated away or commodified so there's very little margin left anyway. It might be true in a misleading sense that the AI {product} is some large fraction of a genuine {product}, but that doesn't mean anything because the value is contained in those differences above the AI produced baseline. The value floor, so to speak, is a lot higher, but the ceiling hasn't risen an inch. Those that move down in quality to preserve margin will soon discover there's not a there there.

pnt12 4y ago

Where I live, lots of burger restaurants opened up in the last years. Emphasis on the restaurant part, not fast food like McDonald's or Burger King. I very much prefer them, as the food is better and the places are cozier, even though a bit more expensive. Maybe we'll get a counter movement soon: "made by humans", as in, the premium experience. Anti AI sentiment may also grow and make that even more viable, as peoplr may boycot heavy-AI shops.