Hacker News Reader

Hacker News

Comment by 2024throwaway | Hacker News Reader

Parent Full thread

2024throwaway

Looking at the source code [0] it's a bare bones wrapper around the OpenAI API, and not much else.

[0] https://github.com/Deeptechia/geppetto/blob/main/geppetto/op...

View on HN

Simplicity doesn't mean lack of value. Evaluate on value.

I was responding to the question in the comment

> How do you improve upon/customize their models for your use cases?

It does not do this.

Geppetto is useful for us internally, we think others will also find it useful. Geppetto is not the first integration of ChatGPT with Slack, there are several and it seems like we are not the only ones finding it useful.

If you are looking to an AI/AGI/ML innovation, this is not what Geppetto is. It is clear from the beginning.

BTW, all this does not imply we are not working on the ML field in cybersecurity static analysis though.

> If you are looking to an AI/AGI/ML innovation, this is not what Geppetto is. It is clear from the beginning

I think part of the confusion, and why it isn't clear, is that releasing an "A.I. assistant" implies that you've put work into the A.I. or assistant aspects of it. The A.I. assistants in this case seem to be ChatGPT and DALL-E, and this seems to be a connector between existing A.I. assistants and Slack.

Adding to the confusion is the human-sounding name, which makes it seem like an A.I. assistant, rather than an integration.

As an aside, show H.N. is about being open and collaborative, and your replies to several people on this thread seem quite secretive and adversarial. Would you consider sharing more information, rather than less? Let us in :)

From what I can see the only thing it does above and beyond Slack's own integration is generate images. To me that's the most low-value part of the value proposition.

https://www.salesforce.com/news/stories/slack-news-dreamforc...

OpenAI also has a teams product:

https://openai.com/blog/introducing-chatgpt-team

I mean, I understand, it's an internal tool and all, you're not building a business here, out of this tool, but I'm going to guess that you end up ripping this out within months as Slack invests big bucks into integrating AI into their platform.

Has no value considering that Salesforce/Slack already built it into Slack.

https://www.salesforce.com/news/stories/slack-news-dreamforc...

OpenAI also has a teams product that allows collaboration in a team:

https://openai.com/blog/introducing-chatgpt-team

A little wrapper like this is going to get eclipsed fast.

Are they open source?

Does that matter? Some open source code that wraps a proprietary service and presents it to another proprietary service is basically entirely proprietary anyway.

Clearly yes, you can modify our software to use Ollama and will save you hours of work. Full non-proprietary.

Look at the video please. We are using it at work. For example, we have a #learning-and-teaching Slack channel when the team members ask and answer questions. Now, you can ask questions to @geppetto which makes all the discussion more appealing and everybody in the channel can continue the thread. This includes generating images.

Also, have you look at the size and quality of the code? It is enough small so that other people can fork and play with it without requiring a lot of knowledge.

One of the code reviewers is a top Python community participant though.

I did watch the video.

> which makes all the discussion more appealing

I have never experienced AI making anything more appealing. YMMV. I also have no need to generate images.

> have you look at the size and quality of the code? It is enough small

I agree it is a very small code base, in that it does basically nothing. Not sure how I'm supposed to view this as a positive. As far as quality, it looks like pretty much all the other python code I've seen.

> One of the code reviewers is a top Python community participant though.

Cool.

It seems you don't need Geppetto. Which is fine.

We need it and for several reasons (e.g. security) we write it from scratch.

The use case for integrating Slack with ChatGPT is well known, so it is useful for several organizations. Enough.

I have actually built a tool very similar, slack <-> openai interface.

Except it does things like augment the input prompts, and can perform additional action based on the outputs.

So no, I don't have a need for something as simple as a ChatGPT pass through.

You might get a better reaction if you weren't so defensive over a sub 500 LOC project.

The sub 500 LOC is a positive thing. If someone is looking for code to fork, learn and/or copy 500 is ideal instead of knowing a lot of details of more complex piece of code.

I receive a good reaction though.

> makes all the discussion more appealing

Yuk.

> One of the code reviewers is a top Python community participant though.

Double yuk.

Something to add?

Ahaha, the video is a GIF one can't fast-forward through, at least it simulates in real time the long wait for the generated image. I've seen enough marketing stuff that I'd like to tell you, that long wait in your advertisement will bore the hell out of people.

Ahaha, the video (on the github readme) is a GIF one can't fast-forward through, at least it simulates in real time the long wait for the generated image (and on your website it's 20 seconds, 33% of the whole video!). I've seen enough marketing stuff that I'd like to tell you, that long wait in your advertisement will bore the hell out of people.

Reply on news.ycombinator.com