3 ms·

Add these technologies to your toolbox. It is not easy to do "prompt engineering". Keep in mind that most generated material looks awful -- there is a huge sel

Add these technologies to your toolbox. It is not easy to do "prompt engineering". Keep in mind that most generated material looks awful -- there is a huge selection bias going on. The outputted stuff needs to made production ready also, distribution, etc. If anything, the emergence of these technologies should greatly expand. You will have a new customer base that was not available before, and you will be able to deliver new products.

adamredwoods 4y ago
Agreed, we're still decades away from solid results. I compare this to automated driving: Truck drivers and cab drivers are still in business. Another example is the uncanny valley of CGI people: at what point was it acceptable to audiences, and at what point was it excellent?
- OJFord 4y ago
  > I compare this to automated driving: Truck drivers and cab drivers are still in business. It's pretty early to declare that a victory for human drivers isn't it? I'd be worried, if I were them.
- bulbosaur123 4y ago
  > Agreed, we're still decades away from solid results You spelled "two years" wrong.
TillE 4y ago
> Keep in mind that most generated material looks awful -- there is a huge selection bias going on. Yeah. Look at machine translation - it's gotten pretty good, in some limited cases it's perfect, but it's also frequently just clueless, and for anything serious you absolutely need a real human translator. I think these AI tools will improve, but I'm extremely skeptical that it's going to improve by like an order of magnitude any time soon. It will still require extensive tuning, heavy selection, and accepting that you probably can almost never get it to make exactly what you want.
- earthboundkid 4y ago
  I have a childcare provider whose first language is Spanish but also speaks English well. In texts, she’ll frequently use Spanish, and then I can just push a button to translate it. The translations are usually good, but sometimes bad. I think overall, I wouldn’t use machine translation in a scenario where I would use a translator before, but I would be more likely to work with someone who doesn’t speak English perfectly. So, it’s net increasing the demand for translation because the transition costs for small translation work is lower.

adamredwoods 4y ago

Agreed, we're still decades away from solid results. I compare this to automated driving: Truck drivers and cab drivers are still in business. Another example is the uncanny valley of CGI people: at what point was it acceptable to audiences, and at what point was it excellent?

OJFord 4y ago
> I compare this to automated driving: Truck drivers and cab drivers are still in business. It's pretty early to declare that a victory for human drivers isn't it? I'd be worried, if I were them.
bulbosaur123 4y ago
> Agreed, we're still decades away from solid results You spelled "two years" wrong.

TillE 4y ago

> Keep in mind that most generated material looks awful -- there is a huge selection bias going on. Yeah. Look at machine translation - it's gotten pretty good, in some limited cases it's perfect, but it's also frequently just clueless, and for anything serious you absolutely need a real human translator. I think these AI tools will improve, but I'm extremely skeptical that it's going to improve by like an order of magnitude any time soon. It will still require extensive tuning, heavy selection, and accepting that you probably can almost never get it to make exactly what you want.

earthboundkid 4y ago
I have a childcare provider whose first language is Spanish but also speaks English well. In texts, she’ll frequently use Spanish, and then I can just push a button to translate it. The translations are usually good, but sometimes bad. I think overall, I wouldn’t use machine translation in a scenario where I would use a translator before, but I would be more likely to work with someone who doesn’t speak English perfectly. So, it’s net increasing the demand for translation because the transition costs for small translation work is lower.