When you're working on a startup the hustle is real. Walking into work this morning and seeing that somebody added us here and that it hit the front page is a great way to start the day. Thanks y'all :)
As a thank you here's a code to get 30,000 credits on me to test out some of the algorithms. Here's the code: HNDec2017
very interesting. Is there any kind of forum where you can state a problem and someone can suggest what AI/ML to try and apply to the problem?
We have an IOT platform where we have thousands of sensors gathering all kinds of data for all kinds of industries, there are lots of interesting things we could do with the data. A lot would be around predicting the future, other situations is about classifying complex data streams as a certain kind of events which we can then use as alarms or information for decision making.
That's awesome. Honestly I submitted this to HN as a way to look up comments about your site. I was surprised when it came up as a new submission. This is the first time that my 'submission to find the comments' ended up on the front page. I'm glad it made your day.
There is an decentralized AI/ML startup building something in this domain on top of Ethereum called Synapse.AI you'll want to check out if you're interested in Web3 stuff. They do both the data, algorithms, and other functionality to help autonomous agents grow.
Numerai also wants to monopolize Data and ML models in a decentralized way, and I believe that is also on Ethereum.
There is also Enigma, Doc.ai, and a few others if you're interested in this space.
Yeah read the yellow paper. It's all about connecting data to models and beyond. It's next level stuff. It's been posted a lot in the groups I'm a part of and people are saying it's part of some crazy automated future of AI and robots.
I never used this platform until today, wow.
This reminds me of old Mashape and early Blockspring if they had a baby (loved those). I'm very excited to dig into this more! WOW, I can tell more developers would love this if there was more outreach and love around this product!
- Video tutorials - at least for the onboarding to show around the platform and get users adopt faster (other than the hello world). Cause when we did the api marketplace rebuild from the UX perspective we messed up the onboarding (for a while) because we didn't give users a clear explanation of what they had to do to get started. There was too much text in the docs and too many buttons - once we fixed that users started consuming APIs within a minute
- Integrations with other platforms - how can I use this with something like Zapier (where all the non techies live)
- Community <-> Have a question? Join our Community slack/discord/board whatever instead of emailing you.
Awesome. Thanks!
- In January we're going to start live webinars where users can ask questions and we can help with specific questions. We'll use those recordings to make some onboarding videos.
The second two are great ideas. I'm going to see if we can get them on the road map ~ Mike
AI-as-a-Service is a interesting space that is likely to grow over time.
Part of the problem with doing AI in-house is that it's really hard/expensive to get large corpuses of correctly labeled data to train your algorithms on, so what these folks are really selling is a set of trained model weights exposed as an API.
this is exactly what I say every single time someone shows off a project like this. Give me valuable data (and lots of it).
Instead of an algorithm store, have a dataset store. I wanna buy a dataset of 1 million tagged stackoverflow questions. I wanna buy a dataset of 1 million random Amazon reviews. I wanna buy a dataset of 1 million blog comments. i wanna buy a dataset updated monthly of billions of new urls found in the web. data, data, data.
with data I can come up with a crude algorithm that works in a day. with no data, not even the most sophisticated algorithm will work.
A related endeavor is James Simons' Flatiron Institute.
They're creating bespoke algorithms for academics in the fields of computational biology, computational astronomy, and computational quantum physics -- replacing current practice of passing down old, duct taped Fortran code from professors to grad students.
Took a deep dive yesterday comparing a few tagging/classification algorithms to the Google Cloud Natural Language and AWS Comprehend APIs, and I have to say I'm impressed.
I've been searching for the right tool to try to add tagging to a large dataset of charitable data, and while I couldn't find anything off the shelf at Algorithmia, the GUI allows for easy forking and adjustment of existing algorithms. As a primarily front-end developer, never thought machine learning would be this accessible. Great work!
If there's something that you need—please use the chat box to tell us what you need. We have engineers that are constantly helping devs get their algos dialed in. If you're doing 501c3 work you can probably ask for some bonus credits as well (we love helping good causes :) )
34 comments
[ 4.0 ms ] story [ 41.5 ms ] threadAs a thank you here's a code to get 30,000 credits on me to test out some of the algorithms. Here's the code: HNDec2017
We have an IOT platform where we have thousands of sensors gathering all kinds of data for all kinds of industries, there are lots of interesting things we could do with the data. A lot would be around predicting the future, other situations is about classifying complex data streams as a certain kind of events which we can then use as alarms or information for decision making.
Numerai also wants to monopolize Data and ML models in a decentralized way, and I believe that is also on Ethereum.
There is also Enigma, Doc.ai, and a few others if you're interested in this space.
- Integrations with other platforms - how can I use this with something like Zapier (where all the non techies live)
- Community <-> Have a question? Join our Community slack/discord/board whatever instead of emailing you.
* Azure Cognitive Services (REST-based)
https://azure.microsoft.com/en-us/services/cognitive-service...
* Amazon ML (REST-based), including Polly, Rekognition etc.
https://aws.amazon.com/machine-learning/?nc2=h_l3_ai
AI-as-a-Service is a interesting space that is likely to grow over time.
Part of the problem with doing AI in-house is that it's really hard/expensive to get large corpuses of correctly labeled data to train your algorithms on, so what these folks are really selling is a set of trained model weights exposed as an API.
Disclaimer: I work at http://www.monkeylearn.com/
The algorithm is one very important building block but a good data set is what finally allows you to materialize a solution.
Instead of an algorithm store, have a dataset store. I wanna buy a dataset of 1 million tagged stackoverflow questions. I wanna buy a dataset of 1 million random Amazon reviews. I wanna buy a dataset of 1 million blog comments. i wanna buy a dataset updated monthly of billions of new urls found in the web. data, data, data.
with data I can come up with a crude algorithm that works in a day. with no data, not even the most sophisticated algorithm will work.
They're creating bespoke algorithms for academics in the fields of computational biology, computational astronomy, and computational quantum physics -- replacing current practice of passing down old, duct taped Fortran code from professors to grad students.
There was an interesting profile in the New Yorker a few days ago: https://www.newyorker.com/magazine/2017/12/18/jim-simons-the...
I've been searching for the right tool to try to add tagging to a large dataset of charitable data, and while I couldn't find anything off the shelf at Algorithmia, the GUI allows for easy forking and adjustment of existing algorithms. As a primarily front-end developer, never thought machine learning would be this accessible. Great work!