Net Neutrality Ends Today A vote is schedule today at the FCC to end the Obama-era policy of net neutrality. The new FCC Chairman, Ajit Pai has made no secret of the fact that he intends to get rid of net neutrality. This is hardly a surprise since Pai is a former lawyer and lobbyist for telecommunications companies that badly want to get rid of this rule. So what is net neutrality? Net neutrality means that your internet service provider cannot throttle or block content or provide fast-passes for certain companies. This means that Verizon or AT&T or Spectrum cannot charge Netflix extra money so your streaming movies appear faster, nor can they block content from companies who refuse to pay them. Pai insists that such regulation is unnecessary. He contends that the only requirement should be for transparency, meaning companies need to tell their consumers what their data policies are. Sadly, most consumers don’t have a choice of ISP’s. Meaning that even if you know your ISP is blocking or throttling content you want, you can’t switch to another one. Why does this matter? Let me make this simple: the large telecommunications and cable providers have sunk millions of dollars into lobbying against net neutrality. They are only spending this kind of money because they are intending to establish fees for content providers. They intend to throttle and block traffic. There is literally no other reason to change this policy. Your internet service will get worse and small companies who cannot afford to pay their fees will get crushed. That isn’t just my belief. That’s the whole point. Your internet starts getting worse today and will continue to get worse until a new administration changes the composition of the FCC. Small companies that innovate and drive our economy will suffer. Large companies will simply pass along fees from ISP’s to their consumers. So you will pay more for the same service. This is a disastrously stupid change based on handing over a crucial part of our economy to a lobbyist from a notoriously greedy and anti-competitive industry. Oh well. In a nutshell: Someday we’re all going to be really, really sorry we let this happen. Read More The State of AI Frank Chen of Andreessen Horowitz delivered a talk at this year’s a16z conference that should be required viewing for anyone who needs to understand the capabilities and limitations of artificial intelligence for their career. By which I mean, everyone should listen to what he has to say. (link below) Chen starts by making the point that AI will end up getting inside all aspects of our software, just like databases once did. He outlines multiple ways in which AI has surpassed human performance. But unlike a lot of AI cheerleaders, he isn’t shy about talking about those areas where AI is not performing as well as humans. AI does better on discrete tasks with a fixed set of possible outcomes. AI does worse on answering arbitrary questions based on a picture or article. For example, if you ask AI to recognize speech, it can exceed human performance. If you ask AI, who in this picture is wearing glasses, it will do quite badly. Chen also puts some of the current AI hysteria in much needed-perspective. First, he points out that according to the best AI researchers (like Hinton at Google) we are nowhere near a General AI that is capable of doing open-ended tasks as well as a person. In fact, Hinton argues that the entire industry may need a reboot in order to create General AI. If we cannot reach General AI, we are unlikely to reach a Super AI that has been promoted as a bogey-man by Elon Musk, Nick Bostrom, and Ray Kurzweil. Further, Chen makes the point that the fear that AI will take everyone’s jobs is also overblown. As he points out, technology adoption has historically led to new and different jobs. To give one example, historical data indicates that the adoption of ATM’s has actually led to more bank tellers being hired than ever before. Why does this matter? Software is changing radically. A number of innovations are coming together at the same time that enable businesses to be much more efficient and productive. Cloud computing, software as a service, narrow AI and the widespread adoption of API’s mean that companies can license or build customized software that will yield a competitive advantage. However, in order to do this, we need to stop treating this as a disturbing and potentially apocalyptic trend. The more you understand about what machine learning is good at and bad at, the more you can apply these advantages to your business. AI is more revolutionary than technology traditionalists would like to believe. But it is also less revolutionary than Nick Bostrom and Ray Kurzweil would lead you to believe. It’s awesome and yet limited. Just like the computer on your desk or the phone in your pocket. In a nutshell: AI is a tool that can serve your business right now. Read More First Person Shooters and AI Gamers like to win. Whether you’re playing against friends or strangers online, the ability to crush the competition is the only thing that matters. As Massively Multiplayer Online Role-Playing Games (MMORPG’s) and first-person shooters became more popular, they triggered an arms-race between players to have the best available technology. After all, if your opponent’s screen is refreshing 60 times a second and yours is only refreshing 30 times a second, your game experience is going to be less smooth and your reactions are going to be slower. This competition led to the improvement of graphics cards or GPU’s. These specialized processors that lie on your computer’s motherboard exist to translate information from your computer’s central processor (or CPU) to the video screen. As graphics have grown more complex and important, these GPU’s (graphics processing units), needed to develop the capability to perform huge numbers of simple calculations very quickly. Graphics cards fill in a lot of the visual information on your screen that would unnecessarily tax the central processor and allow your screen to refresh more often. All of this requires many split-second calculations. The simple way to think about this is that graphics cards can do multiple simple things very quickly and central processing units can do only a very few complex things. Nvidia became known among gamers for producing the best graphics cards, allowing the gamer to gain a crucial nano-second advantage over their competitors. And then neural networks came along. Why does this matter? Originally, neural networks were built use CPU’s. But these central processors didn’t really fit the bill. A neural network is supposed to perform loads of very simple calculations very quickly. Eventually, the big tech companies discovered that the graphics cards, so beloved by gamers, could handle the simultaneous processing much better. That conversion of machine learning to GPU’s drove up the stock price of Nvidia. This drove Nvidia to start producing graphics cards that were far too expensive for the average gamer ($3,000+), but were perfect for neural networks. Although a gamer could still use these graphics cards, the comparative advantages have started to exceed human reaction time. So it would be hard for a human being to perceive the difference. In a nutshell: Machine learning owes a lot to Doom. Read More