OpenAI Team Update
openai.com
openai.com
For a research organization I'd question if you'd want something more specific?
> For a research organization I'd question if you'd want something more specific?
Definitely. The goal at the moment sounds more like one of a funder, although again I'd hope for something more solid (achieve human level performance on problems X, y z for example).
I'd say that's a good problem to discuss when someone (anyone!) claims they are close to achieving it.
To be more constructive, I'd recommend a short term goal on the lines of
Identify 5 areas which could see significant improvement to the lives of people within 10 years with the concentrated effort of openai.
This kind of mid goal (work out a plan) feels fairly sensible, and should lead to research into how tech interacts with people (or doesn't yet) and how open or underfunded areas may be. Maybe you'll find a big bottleneck that would help but isn't that commercially viable to investigate.
Now with some identified areas to work in you can make more specific plans or goals for those. What's the first step?
Unsupervised Learning for Physical Interaction through Video Prediction
http://codeforces.com/blog/entry/18051
Cool.
You can use a real-world environment with a robot (noisy, more latency, computationally restrictive , more headaches, and extra development time)
Or you can utilize a 'virtual' world. Minecraft is being used as Microsoft's sandbox/playground. Video games are a good sandbox/playground : physics engine+sim+a.i hook.
You can generalize this into an observation->action loop platform.
There are many open source platforms out there for this. Take Box2d as an example which OpenAI put a custom wrapper around.. It's a 2d Physics simulation engine. 2-Dimensions = less complexity than 3-D. You have your 'physical environment' and then you hook that into your A.I .. observation/command loop.
There are public Sandboxes/playgrounds and there are private ones. Many of these engines aren't hard to tie into. I'm sure this was one of the major things to resolve at OpenAI as it is for anybody in the space : make a 'playground environment for testing your A.I'...
I'm also sure that OpenAI has a more advanced private gym with higher fidelity links to the A.I like everyone in the space does. Members only ;)
So, OpenAI has packaged a bunch of open source engines/etc into an approachable platform to make development easier. Also, OpenAI hopes to get people to upload their results and have them detail how they achieved their results... Interesting.
A.I development is at the point IMO where you don't need a name or accolades to contribute. You don't need a PHD. You don't need to be an expert in the field. You don't need to be some award winning coder. In some ways, that can harm you w.r.t to having an ingrained view on how to approach problems in a space that is begging for new paradigms.
Dedicate a solid month and you can have make a virtual A.I gym setup and be off and running.
If you have done any serious code development, you can easily break into this space.
The thing that will be the most time consuming will be wrestling with these packages, dependencies, understanding them, and figuring out how to hook/in out of them.
It seems openAI has tried to reduce this pain with the release of openAI gym.
However, you'll find, if you get into any serious A.I dev, you're going to want to start cutting through people's wrappers and add-on layers that add latency, increase complexity, and keep your A.I away from the heart of the sim.
You'll want your own custom hooks.... You'll want the code to be as low level as possible.
Hi there (gdb), cute handle =P.
I highlighted box2d as its an opensource physics/sim engine.
The sim/physics engine is 90% of the 'gym'. I'm sure if there was money involved, the sims/hooks/etc that are being sourced and packaged around could get a lot more advanced more quickly. i.e : https://github.com/erincatto/Box2D/issues
Fortunately, we'll have our first batch of papers coming out soon. (We also released OpenAI Gym a few weeks ago, but that's supporting infrastructure rather than novel research.) Hopefully we'll be able to earn our way out of the basket afterwards :).
are you batching papers instead of releasing them when they're done?
That sounds a little harsh and I don't really mean in that way. Commercial AI/ML projects, since they have a clear goal they can measure against (play Go, drive a car, get better search results) tend to make meaningful progress that we can see, whereas the more philosophical and research-oriented non-profits don't have a yardstick like that, and probably won't until we have an inkling on how to actually make "strong AI". So this organization will probably seem like vaporware for some time, but hopefully it will prove its worth in the long run.
b) Also, unfair to MIRI. They are not only researching the area but also creating it. A lot of early work want into socializing AI risk as an idea, and in that sense pretty much every future effort in this area owes to them. I've given noticeable dollars to MIRI and pretty satisfied with ROI (of course always room for improvement), would do it again.
Even with concrete plans/objectives ML/DL/AI projects rarely live up to their hype. IMO An AGI project that isn't focused around observable learning models (eg, not around "consciousness" etc...) isn't going to make much progress.