Machine learning APIs are too expensive: that’s why we built Machine Box
The hottest shit around right now is blockchain… I mean Machine Learning. It’s what powers self-driving bitcoin, right?
The truth is, developers only really care about something when it’s useful. And we can only figure out what’s useful if we can take it for granted as a capability. Otherwise, we’ll never have the true freedom to innovate.
The early internet (when it was expensive to take part) was pretty rubbish by today’s standards; dancing Jesuses, the <blink> element and repeating backgrounds where any old hex code could be thrown up on the webpage until you’re left dry retching into your mouse-that-had-a-ball-in-it-for-some-reason. It wasn’t until basically everybody had access to it, that the floor was raised and the online revolution could properly begin.
“we had a bill of $40k in one month — for testing”
If you try to innovate on top of Machine Learning APIs, and try to test it, using cloud hosted Machine Learning APIs or heavy enterprise hosted solutions, you’ll quickly run up against either extreme costs (at a previous company, we had a bill of $40k in one month — for testing) or impossible setup with complicated licensing, all of which will stop you dead in your tracks.
Enter Machine Box
At Machine Box, we want to do something different. Machine Box aims to give developers what they need, right now, for free so they can start their innovation, and get this bloody Machine Learning revolution off the ground once and for all.
For example, Facebox unlocks the ability to teach and recognize faces in images and videos, with two or three simple APIs. You don’t need to read pages of documentation or learn complicated workflows to incorporate start-of-the-art face detection and recognition into your app today. Tagbox does the same thing for image recognition. And there are more boxes too, and more coming.
Neat and tidy: less cognitive noise for devs and ops
All Machine Box technology runs inside a neat Docker container, that can easily be spun up and down in one command. Once the box is running, each developer gets their own private console (hosted from inside the container, Morty!) containing the API docs, a web based Try it now panel, and even interactive apps that let them explore the possibilities of this new capability. Once the box has fulfilled its request, you can spin it down and it stops existing.
Your data remains your data
If you believe in privacy, this approach becomes invaluable
Since everything runs inside a Docker container, you are free to run them in your own environment; one that you control whether on-prem or not. So you don’t need to send sensitive information over the web to the cloud. Google and Microsoft never even see your data, and neither does Machine Box. If you believe in privacy, this approach becomes invaluable.
As you scale, the price… doesn’t?
One of the defining differences between Machine Box and its competitors is the pricing. To use the boxes for commercial use, you just need to buy a simple flat-rate monthly subscription. And you can use the boxes as much as you like.
Most companies so far are on the $99/month subscription, and the most your company can ever pay right now is $499/month, whether you’re running a million instances or more
We think this is the first step towards making Machine Learning capabilities truly accessible to developers.
We’re small and new, so you can talk to us
Machine Box is currently a very young company and small team of dedicated people, which we think makes us ideal for a partnership or integration.
With “impressive support” and “the best developer experience of a product ever” we are ready to be entrusted to always deliver cutting edge tech and service.
For product owners…
If you own the product and have a little technical understanding, then you should check out the next section; because it’s actually really quite simple once you know how to run commands in a terminal and use a web browser.
If typing commands into a black rectangle isn’t your bag but are still interested in ways Machine Box can help improve your app, feel free to contact us and we’ll give you a quick demo, and chat about what you’re up to.
For developers…
The hardest thing developers have to do to get started is go to https://machinebox.io/account and enter their email address in exchange for an API key.
You don’t even really need to create an account right now.
You’ll be emailed a magic-link which you can use to access your key.
Once you’ve set your key (in an environment variable as recommended), you just need to use the docker command to run the box.
docker run -p 8080:8080 -e “MB_KEY=$MB_KEY” machinebox/faceboxMost people already have Docker installed, but if not it’s pretty easy to do
The box will download and run locally. Head over to http://localhost:8080 and check out under the hood:
From here, you can either play with the Try it now sections to get a feel for what it can do, or start to write code — or curl commands — that consume the APIs with which you will incorporate machine learning capabilities into your app.
Support Machine Box
If you believe Machine Learning needs to become more affordable before it can be truly useful, then please support Machine Box by playing with it, using it, tweeting about it, and subscribing to our STARTUP plan so we can continue to solve hard problems with exciting new technology.
Got demos?
Aside from what you get from running the boxes locally (most boxes have a demo app that we built to show it off), we’ve also put together this stunning Visual Search application:
- See a video of Visual Search in action, or
- Play with it yourself at https://apps.machinebox.io/visualsearch
Visual Search uses Tagbox to automatically find similar images from a pre-trained public image set, and you can read more about it in this blog post:
Visual Search is just a starting point, you can do many things like:
- Do face recognition on video with Video Face Analysis
- Use Web Face ID to identify people in your Web
- Increase the power of your search engine with Textbox
- Stream and recognize people from a Webcam
