Not sure what your particular area of expertise is but every time I have browsed Wikipedia for articles related to technical topics including math, programming, AI/ML as well as science, I have found their articles accurate and informative. I also recall a few studies comparing the accuracy of articles on Wikipedia with Encyclopedia Brittanica and journals, which conclude that Wikipedia compares favorably with both (easily found via a google search).
We run a few high traffic websites, and would be great to monetize our US traffic at $10 cpm. But your profile link to your LinkedIn page returns a Page not available error. Please get in touch - would be great to chat.
Would be interested to know if you can do all this by using their open-source conmmunity edition, or would you need to use their paid / Pro version. Since the pricing for the paid version is not given, not sure but assuming it would be fairly expensive to use / deploy? My preference is for an open source tool so we are not locked in to a propreitary platform for the long term.
I was wondering what would be the requirement for self-hosting Jitsi - in terms of server resources - bandwidth, CPU etc? Is the Video/VOIP transmitted client to client or does it get routed via the server? Thanks.
Do you mean for more than 3 - the call/video is routed through the server and would consume server bandwidth? If so how is jitsi paying for all this if it is free?
Interesting that this is Intel's second acquisition of an AI chipmaker. They had acquired Nervana for $400M in 2016 - but now seem to be abandoning it in favor of Habana's chips. They would need to redo the Nervana software stack to work with Habana. As per the article, one of the reasons is:
"Nervana’s Neural Network Processor (NNP-T) uses a proprietary interconnect for scaling, while Habana’s Gaudi can scale to thousands of nodes over standard 100Gb Ethernet."
Prediction:'Bus ticket collecting' will enter our lexicon as a new way to describe obsessive pursuits with little or no perceived impact - or at least it would appear to have no foreseeable impact or future to you at that point in time. The opposite of what you should be doing with your life if you want to make a difference. As in : "Although I enjoyed my new research area, at times I wondered if I was merely bus ticket collecting.."
The paradox though is that you cannot know in advance if you are merely bus ticket collecting or pursuing something that will have a real impact. Just like Darwin had no way of knowing if his excessive obsession with natural history would go on to change the world, and lead to the theory of evolution. In the essay, pg does give heuristics on how to figure out the difference - but the greater the potential for impact the less likely it is that you will be able to tell the difference. In other words, the most transformative discoveries will likely seem most like bus ticket collecting and a wase of time, at least initially.
Having watched Shark Tank for a long time, I always wondered if Kevin O'Leary just played a ruthless shark on the show or if he really was one. The article seems to confirm he is definitely the latter: "He was the soulless businessperson who just came in and bought a bunch of companies and scaled them back and laid off all the good people.”
Moreover, he manages to sell TLC for $3.5 billion to Mattel, which is forced to sell it for just $27 million 18 months later. Of course, that was a poor decision by Mattel's then CEO, considering she was warned not to do the deal: "TLC was a “house of cards.” The business was boosting its revenue by purchasing new companies, but the significant acquisition costs weren’t reflected in their top-line earnings."
Interesting that another shark on the show - Mark Cuban - also became a billionaire by selling his company to Yahoo at its peak, and Yahoo then essentially shutting it down.
Great - can't wait to try out the new version! I am running fast.ai in a virtual environment. How does one upgrade to the latest version for both Fast.ai and PyTorch. Are the new versions backwards comaptible with the earlier version - would it break any code? If not, would you recommend I clone the current install and upgrade that to be on the safe side. Thanks!
This is great news! Here has been my experience using Fast.ai : I had been training a deep learning network using Keras with Tensorflow, for diagnosing medical images - and it took me several months of hard work - tweaking parameters, training and testing to get acceptable levels of accuracy for our models. And then last month, I switched to Fast.ai (their pre-release version) and I was blown away - my models trained faster, and I matched and finally exceeded accuracy levels acheived with my earlier models. And I accomplished what had taken several months in Keras, in just a few days! And the biggest reasons for it were in my view, fast.ai's learning rate finder, the differential learning rates, and Test Time augmentation - all which are advanced features built into fast.ai. And the other great thing is that fast.ai uses the best defaults automatically, and it trains much, much faster than Keras / TF for some reason.
So I can't wait to try the new release out. I think Fast.ai has set a new bar for deep learning frameworks in terms of speed and ease of use. Thank you for all your great work!
The fashion industry is another example of an area that has thrived because of a lack of copyrights, as argued is this TED talk :
"Copyright law's grip on film, music and software barely touches the fashion industry ... and fashion benefits in both innovation and sales, says Johanna Blakley. In her talk, she talks about what all creative industries can learn from fashion's free culture.."
Yes, RNNs and autoencoders, etc. do seem to be encoding concepts via embedding vectors in higher dimensions. But see the argument made by Doug Hofstadter about language translation using RNNs [2]. We have also seen how CNN layers can "visualize" higher levels of features in images as we go deeper in the network. And many think that
some variation of this basic approach done repeatedly, with more data, computing power etc. will be sufficient to lead us to AGI in the future.
However, many of the leading figures in AI - including Geoffrey Hinton - the father of deep learning, is very skeptical of the approach to AI that he pioneered. He recently stated - "My view is throw it all away and start again." [0]
Francois Chollet - the author of the deep learning framework Keras, has said: "For all the progress made, it seems like almost all important questions in AI remain unanswered. Many have not even been properly asked yet." [1]
And of course, Doug Hofstadter, who thinks it is going to take a lot more to come close to human level intelligence & understanding, even when you consider the most advanced RNNs of the day - those that run Google Translate .
>maybe what we think of as conceptual understanding is just as much simple interpolation of our experiences as our neural networks...?
Possibly - we don't know. Sure if we could train one of these imagenet type classifiers with a million or billion images - close to all known objects in the universe, it may well be able to "recognize" everything. But that still doesn't solve the problem of abstraction or meaning, much less of intuition and generalization. Humans are able to generalize to new domains based on internal models of the world around us. The models used in RNNs for encoding word embeddings, seem to a bit closer to representing meaning. I agree though that NNs are evolving - we are in very early days, and who knows the NNs of the future may reveal that what we consider as understanding is no more than simple interpolation, as you suggest.
This example, and others like it point to the central weakness of neural networks for image recognition:
No matter how much data you feed it, they never really develop concepts or abstractions of what the objects it is classifying really represent or mean. The weight and biases that get fine tuned by gradient descent, are no more than a highly complex function mapping the input pixels to discrete classes. While this may well represent how the visual cortex works at the lowest level, what appears to be missing are higher levels of abstraction and meaning. Perhaps machine learning needs to be coupled with some of the older paradigms of AI which included modeling, logic, reasoning, to achieve understanding. As of right now, a well trained convolutional neural network is no more than a mechanical pattern matching algorithm on steroids.