9 ms·

Does anyone have an example of agents like AutoGPT doing something useful? Everything I've seen seems to be stuff that GPT could do anyway without the agent cru

Does anyone have an example of agents like AutoGPT doing something useful? Everything I've seen seems to be stuff that GPT could do anyway without the agent cruft. And iterating seems to multiply the opportunities for LLM bugs and mistakes. The hype seems to be that agents are an emerging form of AGI, but the fine print is always "it's not quite ready for production yet, it makes a lot of mistakes, I had to fix 20 things in the output, but it's fun to watch and I'm sure once we work the bugs out..."

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.
jondwillis 4y ago
I got a very recent version of Auto-GPT to clone a template repository, launch a nextjs dev server, and badly start iterating on a product last night. It took about 3 hours of prompt and setting tweaking and cost about $35.
- derwiki 4y ago
  That is super rad! Are you able to share anything?
  - jondwillis 4y ago
    of course! - using redis for long term memory/vector search - using gpt4only mode w/ 8k context for "fast" and "smart" llms - allow local command execution My (admittedly, magical and weird) ai.yaml looks like: ai_name: [app]-founder ai_role: You are working on a AI SaaS product called [app]. You complete all goals continuously to build [app]. You are a highly autonomous and efficient software engineering agent composed of ChatGPT 4 instances. You are running on macOS via AutoGPT. You rely heavily on memories from your past. You know [app] is an existing app in the current directory. You avoid 'search_files' because it will crash, so use ls instead. You avoid performing the same actions in a loop or repetitively. You manage a fleet of agents that can answer questions up to Sept. 2021.` ai_goals: - [something to effect of - git clone template repo from this url if it isnt already cloned] - continuously build an app- add relevant features by questioning the state of the web app that is located at localhost:3000 && localhost:5556, and improving the code. - understand and learn about what might make a good [product idea] - research and build a business (with moat) around [app/biz name] (limit sources to high quality such as news.ycombinator.com, github.com, reddit.com, etc),

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.

jondwillis 4y ago

I got a very recent version of Auto-GPT to clone a template repository, launch a nextjs dev server, and badly start iterating on a product last night. It took about 3 hours of prompt and setting tweaking and cost about $35.

derwiki 4y ago
That is super rad! Are you able to share anything?
- jondwillis 4y ago
  of course! - using redis for long term memory/vector search - using gpt4only mode w/ 8k context for "fast" and "smart" llms - allow local command execution My (admittedly, magical and weird) ai.yaml looks like: ai_name: [app]-founder ai_role: You are working on a AI SaaS product called [app]. You complete all goals continuously to build [app]. You are a highly autonomous and efficient software engineering agent composed of ChatGPT 4 instances. You are running on macOS via AutoGPT. You rely heavily on memories from your past. You know [app] is an existing app in the current directory. You avoid 'search_files' because it will crash, so use ls instead. You avoid performing the same actions in a loop or repetitively. You manage a fleet of agents that can answer questions up to Sept. 2021.` ai_goals: - [something to effect of - git clone template repo from this url if it isnt already cloned] - continuously build an app- add relevant features by questioning the state of the web app that is located at localhost:3000 && localhost:5556, and improving the code. - understand and learn about what might make a good [product idea] - research and build a business (with moat) around [app/biz name] (limit sources to high quality such as news.ycombinator.com, github.com, reddit.com, etc),