Hey HN.
Coding sucks! serious Im mediocre at best but I love data science and AI. So we built Canopi for coding newbs like me. What makes Canopi awesome is you can share your data with people and they can use it in their projects but cant steal it. We're early days but the roadmap is awesome. Imagine safely sharing realtime data sources, connecting public data. Building your own AI and deploying to production. AI is no longer developer driven, its now democratised and by boosting diversity its now less biased.
Please have a play with the frontend proto, share your feedback and watch this space!
Hey, nah, coding is fine. But since you ask for feedback: I tried to visit your submission on mobile, please link to something that doesn't ask me to create an account and instead shows me what I'm getting for that. Without that I usually assume I'm getting weekly marketing emails.
Since you make some bold claims in your pitch: do you have technical resources that address bias and data extraction from deployed models in your application?
Hi Tasroder. Approximately 70% of the world use some form of visual thinking. However if like me, a person is a strong visual thinker and mentally atypical, coding can be quite challenging. With Canopi, I've attempted to rethink how we work with data. Conceptualising it so it is not longer tabular or nested KVP.
I'm sorry you assume I will market to you. I'm not intending to.
I didn't think they were bold claims.
'Technical resources that address bias and data extraction' is quite vague. Can you please ask your question more explicitly?
Thank you.
Hi, sure, I didn't want to dismiss the idea itself. The no code approah just isn't for everyone I guess. Most major cloud providers have some level of abstraction for ML modelling these days which is why I was curious to check yours out as well. Hence please take my main point to be the plea for something that shows what your platform is doing without exposing data. :-)
As for the other questions, your pitch text made claims like "you can share your data with people and they can use it in their projects but cant steal it." as well as, I assume, the mere existence of these tools having some impact on bias. Both data extraction from models deployed in the wild and model/data bias have been topics of discussion in ML research over the past few years and I was wondering if your tooling did anything specific to tackle those aspects.
PS: Are the icons like "view_quilt" on your home page huge? On mobile chrome I could read halfway down the page before those loaded in and was a bit confused about the placeholder text.
PPS: On your homepage in a real browser now (Firefox on Linux, recentish version), you somehow managed to disable text selection there.
tastroder - firstly, our primary aim is to democratise AI and data in general so it is no longer driven by developers. Most non-technical people I've spoken to have no idea what AWS, Azure, etc are and where 'the cloud' is. These platforms rightly so are built for developers. This is an immediate source of bias when the overwhelming majority of developers are male.(here is a really interesting article https://www.linkedin.com/pulse/give-up-gender-gap-science-cs...)
Given you understand how cognito works I'm sure you dont mind creating an account with one of the many gmail accounts you most likely control and therefore you wont be exposing any meaningful data.
Data is encrypted in transit and at rest. There are also other propriety features we are developing to limit the extent to which data can be copied with out permission.
On bias - Data is stored with meta descriptive properties and attributes, which provide the user with a guide to the underlying data composition and therefore can be used to highlight potential bias.
Our dev horizon also include similar monitoring capabilities to provide information that can assist users detect bias.
The truth here is simply - AI are the products of algorithms, data and developer
There is only one source of bias in this statement - developer
algorithms are tools; data is information without intended action; developers use these to create intended action.
The most obvious way to remove bias from technology is to improve diversity in the pool of technology creators. We are approaching this by increasing the size of the pool by enabling non-technical people to be a part of AI development. As well as increasing the ease at which collaboration occurs in AI development. This begins with the safe sharing of data, which in turn means more data becomes widely distributed for use in ML. It continues with a collaborative approach to data modelling, in which the domain expert leads the process - removing more opportunity for bias introduced by developers who may not have deep domain knowledge and who therefore may not fully grasp the consequences of modelling decisions. And similarly with collaboration on AI development, deployment and monitoring.
Imagine a future in which 10 billion people can build their own AI and not have to ask the 100 million developers to do it for them.
Thanks for this reply, not sure what the other one is supposed to be. I was just trying to give you feedback on the thing you posted, hope that didn't come across too harsh. I wish you the best of luck.
8 comments
[ 1.6 ms ] story [ 31.8 ms ] threadSince you make some bold claims in your pitch: do you have technical resources that address bias and data extraction from deployed models in your application?
I'm sorry you assume I will market to you. I'm not intending to.
I didn't think they were bold claims.
'Technical resources that address bias and data extraction' is quite vague. Can you please ask your question more explicitly? Thank you.
As for the other questions, your pitch text made claims like "you can share your data with people and they can use it in their projects but cant steal it." as well as, I assume, the mere existence of these tools having some impact on bias. Both data extraction from models deployed in the wild and model/data bias have been topics of discussion in ML research over the past few years and I was wondering if your tooling did anything specific to tackle those aspects.
PS: Are the icons like "view_quilt" on your home page huge? On mobile chrome I could read halfway down the page before those loaded in and was a bit confused about the placeholder text.
PPS: On your homepage in a real browser now (Firefox on Linux, recentish version), you somehow managed to disable text selection there.
Given you understand how cognito works I'm sure you dont mind creating an account with one of the many gmail accounts you most likely control and therefore you wont be exposing any meaningful data.
Data is encrypted in transit and at rest. There are also other propriety features we are developing to limit the extent to which data can be copied with out permission.
On bias - Data is stored with meta descriptive properties and attributes, which provide the user with a guide to the underlying data composition and therefore can be used to highlight potential bias.
Our dev horizon also include similar monitoring capabilities to provide information that can assist users detect bias.
The truth here is simply - AI are the products of algorithms, data and developer
There is only one source of bias in this statement - developer
algorithms are tools; data is information without intended action; developers use these to create intended action.
The most obvious way to remove bias from technology is to improve diversity in the pool of technology creators. We are approaching this by increasing the size of the pool by enabling non-technical people to be a part of AI development. As well as increasing the ease at which collaboration occurs in AI development. This begins with the safe sharing of data, which in turn means more data becomes widely distributed for use in ML. It continues with a collaborative approach to data modelling, in which the domain expert leads the process - removing more opportunity for bias introduced by developers who may not have deep domain knowledge and who therefore may not fully grasp the consequences of modelling decisions. And similarly with collaboration on AI development, deployment and monitoring.
Imagine a future in which 10 billion people can build their own AI and not have to ask the 100 million developers to do it for them.