3 ms·

The intention is for them to do things that are more like projects. E.g. build a website or even a business (e.g. an ecommerce shop with marketing, etc). AutoGP

by jasfi 4y ago

The intention is for them to do things that are more like projects. E.g. build a website or even a business (e.g. an ecommerce shop with marketing, etc). AutoGPT and BabyAGI are only a few weeks old, so they need time to build. So they are being criticized too early in my opinion.

saurik 4y ago
They are also being hyped too early. It is 100% fair--hell: I'd even say, more correct--to criticize something for what it is, instead of trying to predict what it might one day become. If the stuff gets better, it can come back around then and soak up its deserved praise. (Though--and this is an unrelated issue--in the case of some of this AGI work, if it actually works we are probably going to collectively wish it hadn't been built, as no good is going to come from rapidly empowering an AI to become in any way autonomous, even momentarily.)
- jasfi 4y ago
  I think a lot of the hype is because it's close, but not quite usable. Nobody's really sure what the weak points or limitations are. This sort of dev isn't easily slowed or stopped, it's Open Source except for GPT-4. Even the Open Source LLMs are getting there.
  - ratg13 4y ago
    Right but all you have to do is learn to engineer prompts better. You should be able to start with a fresh session and feed it what you need to get the result you want. If you’re always relying on the entire history of your sporadic conversations you’re often going to get a mess. IMO these “improvements” are just crutches for people that don’t want to understand AIs, and want to see them as they are portrayed in movies. It’s an attempt to make them seem intelligent, but it’s just a gimmick, not a tool.
  - jondwillis 4y ago
    Also, anyone who has looked at Auto-GPT and used it will notice how FAST features and improvements are being added. It boggles the mind.
    - jasfi 4y ago
      I think they're being inundated with issues and pull requests. So, the dust really has to settle.

by jasfi 4y ago

saurik 4y ago

They are also being hyped too early. It is 100% fair--hell: I'd even say, more correct--to criticize something for what it is, instead of trying to predict what it might one day become. If the stuff gets better, it can come back around then and soak up its deserved praise. (Though--and this is an unrelated issue--in the case of some of this AGI work, if it actually works we are probably going to collectively wish it hadn't been built, as no good is going to come from rapidly empowering an AI to become in any way autonomous, even momentarily.)

jasfi 4y ago
I think a lot of the hype is because it's close, but not quite usable. Nobody's really sure what the weak points or limitations are. This sort of dev isn't easily slowed or stopped, it's Open Source except for GPT-4. Even the Open Source LLMs are getting there.
- ratg13 4y ago
  Right but all you have to do is learn to engineer prompts better. You should be able to start with a fresh session and feed it what you need to get the result you want. If you’re always relying on the entire history of your sporadic conversations you’re often going to get a mess. IMO these “improvements” are just crutches for people that don’t want to understand AIs, and want to see them as they are portrayed in movies. It’s an attempt to make them seem intelligent, but it’s just a gimmick, not a tool.
- jondwillis 4y ago
  Also, anyone who has looked at Auto-GPT and used it will notice how FAST features and improvements are being added. It boggles the mind.
  - jasfi 4y ago
    I think they're being inundated with issues and pull requests. So, the dust really has to settle.