Of course there are a tons of different Frameworks but I keras

I came to I love keras as the meta language is able to work on top of TensorFlow, theano and also Microsoft CTNK is in development of supporting it.

Even for some, and I believe only a few, use cases the pure frameworks are more powerful – in most cases development time, optimisation, collaboration and education with it is amazingly efficient.

Read about keras: https://keras.io

Additionally, also Lasagne is such a good meta language approach. As I didn’t implement it yet in any projects, I can’t compare it much.

I will give a deeper insight in my following speeches and webinar.

The post I ♥ keras for Deep Learning appeared first on ai-claudio Artificial Intelligence.

It’s already a bit older, from 2010. In the clip Hassabis states, that five years old neuroscience news – are mostly outdated already – but still I consider this lecture very valuable.

I’ll write more about AGI soon, either here or at Mobile Geeks. -> I’ll keep you updated.

The post Demis Hassabis on AGI & neuroscience a while ago appeared first on ai-claudio Artificial Intelligence.

As we at ai-claudio missed on being there – and now online stream oder video available currently, here are some cites from: OfficeLeaseCenter:

On Wednesday, November 19th 2014, OLC attended Cognitive System of the Near Future at RocketLabs in San Francisco where a Jordan Novet moderated a panel discussion.

Jordan Novet –  Staff Writer, VentureBeat also Moderator

Adam Gibson – Founder, Skymind.io

Dave Sullivan – CEO, Ersatz Labs

Subutai Ahmad – VP of Research, Numenta

Joshua Bloom – CTO, Wise.io

JN – What is deep learning and what is cognitive computing?

http://venturebeat.com/

JB – I like to think that cognitive computing is marketing jargon.

http://www.wise.io/

DS – Machine learning is thinking about a program that takes an input (any kind) and you expect it to have some output. What we are trying to do is have these programs write themselves instead of by a developer.

http://www.ersatzlabs.com/

AG – Cognitive computing is marketing. Machine learning is taking what we already know from statistics and building something with it. It can come down to big data and thinking we have this data, what can we do with it.

http://www.skymind.io/

JB – Machine learning can help you keep up with things as the world keeps moving. Data is coming from all sorts of different sources and dimensions. It can be hard to pull the correct data out you need.

JN – You mentioned text and images, what’s the most interesting case you’ve heard about a go-to-market strategy for a new company?

DS – there are a lot of companies out there that are already doing this perhaps using Mechanical Turk to do these tasks for us where we are trying to create a program to do this. With the data collected from accelerometers, with this movement, what is the person doing?

SA – Nemo baby, is a little onesie for an infant that has a heart rate monitor, and all these other sensors connected to wifi to the cloud to an app that allows you to monitor your babies health real time, all the time.

http://numenta.com/

JN – Running a company in this industry, how do you find great talent? Where do you look?

AG – You teach them! I happen to work with some schools, so I’m able to meet potential hires.

SA – We have a good internship program. We get students – they come to us actually and we work with them and train them. We don’t advertise anywhere for new hires and people just naturally come to us.

AG – There are lots of classes out there to learn this stuff. Online academies, master programs. It comes down to wanting to work on something cool and enabling people to do it.

DS – At our company, myself and our CTO are the only ones that really understand this, the rest of the folks are web developers.

JB – I get lucky because I’m a professor at Berkeley and I’m exposed to students all the time. The thing with this is you can’t always hire these people right out of school. You need your senior devs out there to help out the newcomers.

JN – I believe a lot of these good devs in deep learning are working down in Mountain View working for Google. How do you try and get these folks.

DS – We try and avoid battling with them because we can’t compete with them. We try and look for people who want to work for a battling startup.

JN – Does it make sense to keep humans in the loop..should machines replace or augment human work?

JB – I wrote a blog post about this a while ago robots where they looked real and fake and can creep people out, then you have things like Pandora that are doing machine learning. What if Google Glass told you one day you should be dating someone of the opposite sex by the data they are collecting. The machine can get to a point where it knows you better than it knows yourself.

You’ll get to a point where these services will supercharge our lives and empower us.

SA – I don’t worry about these consequences with this type of tech, I’m more worried about computer viruses and can get far worse than they are today. I read about people 3D printing DNA and open sources, so people could create some new organisms and I’m more worried about that.

AG – We’re already seeing some of it like these activity trackers such as Fitbit. We need to make these apps better to understand. If these things take off, we’ll get the data to allow people to creative better automation tools.

We’re not going to have something like Skynet viewing on us and spying on us.

JN – Google has such a talent pool of deep learning employees, and some people say they are a deep learning company. What happens if Google releases an API or created machine learning as a service.

DS – They’ll shut it down in two years after they realize they aren’t generating enough ad revenue from it.

SA – There are only a small amount of business models that work for Google. One day Google might figure something out.

Dave Sullivan

CEO, Ersatz Labs

Dave is the primary architect of Ersatz and an expert in deep learning and neural networks.

Adam Gibson

Founder, Skymind.io

Adam is the founder of deep-learning-as-a-service company Skymind.io and a machine learning expert at Zipfian Academy.

Joshua Bloom

CTO, Wise.io

Josh is a self-described “data-driven scientist” who leveraged the power of machine learning to enable landmark astronomy discoveries.

Subutai Ahmad

VP of Research, Numenta

Subutai brings experience across real time systems, computer vision and learning to Numenta.

Jordan Novet [ MODERATOR ]

Staff Writer, VentureBeat

Jordan writes for VentureBeat about big data, cloud computing, and other technology for business.

The post Cognitive Systems of the Near Future – Canon Panel – Splash appeared first on ai-claudio Artificial Intelligence.

www.cs.stanford.edu/people/ang//papers/icml07-selftaughtlearning.pdf 

The post Self-taught Learning: Transfer Learning from Unlabeled Data appeared first on ai-claudio Artificial Intelligence.

The post DeepMind appeared first on ai-claudio Artificial Intelligence.