HNHacker News
TopNewBestAskShowJobs

jhartmann

1,340 karma · joined August 10, 2012

submissionscomments
jhartmann··on AI, Robotics, and the Future of Jobs [pdf]
Renewables are not already viable, they are never going to be viable but for a small portion of our energy needs. Because Physics. They just don't have enough concentrated energy. Take solar for example, take a look at http://www.withouthotair.com/c6/page_38.shtml, In the UK solar basically works out to 100 watts per square meter on average. Somewhere really nice in the American Sotuhwest is maybe like 400 watts per square meter. At 25% efficiency, to equal the baseline load of a 1 GW nuclear power plant, you need something like 7.7 square miles of panel area when you factor in the need to capture enough sun during the day and store it for the night. You need to also have a high voltage DC grid to ship the electricity from the Desert to where we need it, and much better storage systems to hold the load for the night. Basically Solar is never going to work for baseline load, and to top it all off photovoltaics are dirty as hell to make. Wind is even more bad (There was a study that said you needed 270 square miles of land to generate 1 GW reliably), and hydroelectric is limited in where we can deploy it. Nuclear is exactly what we need, renewables just can't be anything but a small part of the solution considering our power usage is increasing not decreasing.
jhartmann··on Y Combinator And Mithril Invest In Helion, A Nuclear Fusion Startup
Actually the D + D reaction has two equally probable branches (and one very improbable one that makes an alpha particle and a gamma ray, which is mostly ignored), one makes Tritium and a Proton and the other makes D + He3. So the D + D reaction will breed He3 fuel, along with Tritium. The resulting proton from the Tritium branch can also combine with Deuterium and create Helium 3 as well in the right conditions. Under carefully control, the right conditions, and with some engineering work I would wager you could bias the reactions and side reactions towards creating more Helium 3.
jhartmann··on Y Combinator And Mithril Invest In Helion, A Nuclear Fusion Startup
Best of luck to you and your team, I really think that alternative paths to Fusion should have been explored since Farnsworth. Instead we have the crazy massive superconducting cathedral's that are literally trying to blow themselves apart when you turn them on. Your team and the teams investigating other approaches (General Fusion, Tri-alpha, the late Bussard's Polywell, Lerner's Focus Fusion) all should get several tens of millions to investigate and do hard science and try for the big payoff. Congrats on getting some more initial funding, I wish you the best of luck.
jhartmann··on AMD HSA kernel driver
I am super excited about this driver. I worked on the early drm first at Precision Insight and then at VA Linux Systems, and having hardware validated command streams seperate for each process is just such a huge deal for performance. I wrote some proof of concept code awhile back that I posted to the Mesa dev list for doing something like this without hardware support, and based on my limited testing it would be a huge improvement in dma throughput. Its nice to see something like this happening finally, I think it is a path to getting a secure and highly performant driver. We have always had to make security tradeoffs with performance in Linux land, now we actually might have the infrastructure to get a chance at matching or beating Windows performance. Terribly exciting.
jhartmann··on Is This How We'll Cure Cancer?
Was probably Remicade, it blocks Tumor Necrosis Factor Alpha which is a main component to the inflamation response. If you are currently experiencing a cytokine storm (which is basically run away inflamation) it will do the trick quite quickly. It is basically an artifical antibody that binds to free TNF alpha and doesn't let it interact with cells to cause its normal immune response. Very effective, and it is an IV delivered drug so it immediately starts binding to the factors that are causing the inflamation. I took it before for Ankylosing Spondylitis and you can feel better quite quickly.

If its not Remicade it is probably something in that family of therapies and would have similar quick effects against a cytokine storm.

jhartmann··on Facebook acquires Oculus VR
I have met Carmack before at Siggraph many years ago and I can say that he struck me as someone who was a 'Code Troll' because he loved it. That he made money was great, but I think he has such a deep love of the technology for technologies sake.
jhartmann··on Bill Gates: AMA on Reddit
I saw that and maybe its stretching it a little bit, but I know Patrice Simard was involved in this sort of research relatively early (he started at Microsoft in 99, and worked at Bell Labs before.) I would not be surprised if there were some other people there that were doing this sort of research even earlier.

There is actually some very high quality ground breaking research that comes out of Microsoft, they just have had problems making it into products traditionally.

jhartmann··on Moz Dumps Amazon Web Services, Citing Expense and 'Lacking' Service
Public cloud is not for everyone. If you have a large number of machines that have to run all the time and are being taxed, then public cloud services do not make sense. If you are doing complex machine learning, large scale web crawling/scrapping, or generating large quantities of data (images/video/etc.) then services like AWS might not be a fit. They are incredibly awesome though for certain types of workload, but for others it is a very expensive option that isn't economical.

Personally I think the best is some mix of both public and private, so you have the scalability when you really spike but have cheaper more capable machines handling baseline and things like analysis or storage workloads.

jhartmann··on Startups Can't Survive On $20,000
Thanks for calling this out. Its a discussion that I really think needs to happen.

I personally love the concept of yc, and see the value but for a good portion of the entrepreneur community it can't work out because of the economics. I would love to see a accelerator give enough money for a realistic budget for a seasoned team. Including myself I have 4 people on my team and we are 3 seasoned developers and an executive with lots of experience as a CIO and CFO. We have real bills and can't live on peanuts, mortgages, kids, need for real health insurance, etc. We are more expensive, but we can outbuild probably most of the teams out there, and especially the ones that have little experience. I wonder if much of the reason we have so many failed startups is that we are don't bet enough on seasoned individuals unless they have enough cash to carry them through the early stages. I would love to see what would happen if pg did a session of YC looking for seasoned teams what would happen. I'd be willing to bet you'd have a good chance for some amazing results, even if it would be more expensive.

jhartmann··on Kima Ventures Will Allow Startups to Raise $150K Within 15 Days Via AngelList
Terrible is a strong word, but I don't think these terms are good for lots of companies. Each company is at a different stage when they first start raising seed funding. Some companies might just be worth 1 MM, but if you have a strong team, great technology, a workable plan, and early customer interest/revenue 1 MM is a horrible deal. In my opinion these terms are great for some ventures and absolutely horrible for others. I think the quick decisions are great, but I would think it would be more fair and equitable for you not to box the valuations. Perhaps offering a tiered option where higher value ventures take a little longer to evaluate, but I think boxing the valuation will have you miss the best companies. My two cents anyhow.
jhartmann··on AMD releases APPML source code, creates clMath library
Actually it needs any OpenCL SDK, not AMD APP SDK specifically. If you were building for an Altera FPGA or an NVidia card you would use their OpenCL SDK. OpenCL is typically a LLVM frontend and backend for the hardware in question (or something similar). What they are releasing is in fact all the source for their FFT and BLAS implementations. They aren't opening up the compiler here, just the software that runs on top of it. The compiler itself probably has lots of stuff in it they consider competitive advantage so it will probably be a little while before they open that.
jhartmann··on AMD releases APPML source code, creates clMath library
While CUBLAS is available out there, it is not something that can be run on such a wide array of hardware. This is probably of great interest to startups wanting to do things on things on FPGA's and mobile devices. They now have a well optimized open source math library to use as potential building blocks, and it has an Apache License. CUBLAS is only good for NVidia cards. So I actually think this could be a big deal, and could potentially reduce the popularity of CUDA over time.
jhartmann··on AMD releases APPML source code, creates clMath library
I noticed this today as I was searching around for some opencl image routines, and thought it was of general interest to the community. I really think this is an awesome thing, and the availability of a high performance open source BLAS that can be compiled to a wide array of OpenCL capable hardware is just great news.
jhartmann··on Id Software founder John Carmack resigns
About 10 years ago he talked to some of us guys writing 3D drivers for Linux, he is all kinds of awesome when it comes to code skills and spent a few hours geeking out with us at Siggraph. About a year before, I got to see his coding in action when we were first bringing up the first Matrox G200-GLX driver. He would accomplish so much in just a few days when he donated some time to helping us get the MGA-G200 working with Quake. He would accomplish in days what most of us at the time would have taken weeks to do. He just naturally understands 3D and 3D hardware and is a geeks geek. I will bet he is spending most of his day to day coding and hacking.
jhartmann··on What happened when my co-founder quit the night before our YC interview
Just curious, what are you up to these days? Got another startup cooking, or are you looking for the next place to hang your hat for awhile before you go out and kick ass again?
jhartmann··on Amazon WorkSpaces
Jeff, Are there any plans to offer something like ubuntu images in the future for VDI? Many developers really don't use windows much anymore, and you could probably offer a more competitive price point when there is no need for licensed software. Obviously we can set up our own stuff using raw EC2 instances, but having a simple solution like this is definitely appealing.
jhartmann··on Is Al Jazeera America Going to Change the Way Networks Cover Climate Change?
This is a link to a post that was not crafted specifically for HN and submissions are often reset to the contents of the html title tag automatically even if the post is not given that text to begin with.
jhartmann··on Android AOSP maintainer quits
I used to write graphics device drivers so I'm going to bite. Usually most vendors think that their 'solutions', details about their hardware designs exposed in the drivers and optimizations are extremely valuable. Personally in my experience I found that most of what vendors considered secret was well known and not that novel and the majority of what was being done was application of the same or similar techniques in their hardware and drivers. Sure there are some novel things, but enough people have moved between the graphics card companies that the drivers really have very little truly unique information about the designs that would help a competitor.

One thing that I think does cause problems for opening up hardware though is that many graphics implementations use the same or similar chips (to save on asic costs) and the drivers enable options based on the product the customer bought, but the underlying chip will actually still implement the higher end functionality. This is a common enough technique and to avoid people getting 500 dollar parts for 99 they can't open up the drivers. I don't understand why they don't push this sort of thing exclusively to device microcode (many do this or at least partially) rather than having software components that do some of this logic.

jhartmann··on Bill Gates’ nuclear company explores molten salt reactors, thorium
Do the math, with advanced nuclear power (stuff we had working in the 60's and 70's, but was crap for bombs and upended the business model of existing nuclear which is roughly equivalent to giving the razor away at cost and making the money on the blades) you can hold all the energy you will ever use in your hand. A metal ball of Thorium is quite safe as well before it is used in a reactor. A molten salt reactor is a chemical plant running at ambient pressure, a much easier thing to deal with then the steam bombs we have currently deployed. They also don't require crazy coolants like liquid sodium (asking for it, a little water and boom) or fast neutrons (eats up anything containing them like swiss cheese.) The biggest problem with the nuclear industry is that they haven't been innovating since the 60's and 70's, not anything else really. Even the crazy steam bombs we have now are many orders of magnitude safer than other energy sources. How many people have died from installing solar panels, because of hydroelectric power, coal pollution, or petrochemical accidents? Let me give you a hint, all of these are WAY more than the number of people who have been affected by Nuclear accidents. Look up deaths per terawatt/hour of all the energy sources and you will see its quite clear that even current nuclear is SO much more safe than anything else, but it has been effectively marketed against by people invested in the status quo.

Diffuse energy is economically unviable, it is only being deployed at great cost and things like solar will never solve baseline load. People also don't think about the pollution caused by manufacturing things like solar panels. There are vast areas in China that are full of industrial waste that will not degrade for a very very long time due to solar panel manufacturing. Way nastier stuff the radioactive elements, because anything that is really dangerous has a short half life (iodine is completely gone in weeks), and will clean up rather quickly or it is usable as a fuel (though we in the United States are daft for not reprocessing our fuel like sane countries like France.) Even existing nuclear is way better than anything else, and advanced nuclear is even better still. Nuclear has a marketing problem, not a technical one.

If your worried about waste, well you have to realize that the really nasty stuff will all decay in 300-400 years, and anything radioactive left over is something I would be happy to have under my bed as I sleep. Anything radioactive at that point is also really fuel, and advanced nuclear reactors (like a molten salt reactor) will happily burn it up. The whole Yucca mountain crap is just caused by the United States having a silly ban on reprocessing. Another thing people don't mention is that the plutonium extracted from a long running reaction like those used for power is completely unsuitable for making bombs, its really just a bugaboo.

jhartmann··on What happened to innovation?
I feel very similar that the Government has stopped working right but have no desire for downsizing it. We should in fact increase spending on big projects and push for the hard things that traditional short term capital doesn't want to touch. Things like AGI, Space Colonization, things like Sonny White's Eagleworks, things that could transform our society in short order. I would love it if the real dreamers like Paul March and James Woodward got sizable funding for their work on Mass Fluctuations and their work on the origin of Inertia, Eric Lerner got some money to work on Focus Fusion, Kirk Sorensen could really build and engineer his Liquid Fluoride Thorium reactors on a large scale and countless more less known scientists and visionaries got to pursue their ideas full time and with significant funding. A well functioning Government and visionary investors if they focused on people like that could really move the ball forward for our society and the world. I think we have great potential for greatness, if I ever get to be 'king' I'm definitely going to give money to those who want to really reach for the stars. I wish our Government thought the same way, hopefully with the right direction and moving away from the quarter as our yardstick we can get there.
jhartmann··on Union Square gets 5,000% return on Tumblr
Fred, I can definitely say that you and several others in the VC world deeply care about your teams and your vision for the world. I have been a part of your community @ avc.com for awhile, I started following you and Brad Feld after I met Brad when he came to Oklahoma City. Both you and he are super genuine, open, and I see time and time again how you go to bat for your entrepreneurs. I think what makes a good investor is someone who takes true ownership of an investment, and does what they can to help it succeed. You definitely fit that bill, but there are plenty of others out there as well.

My advice to everyone looking for people like Fred is to reach out to other entrepreneur's and see who has really been helpful, follow them on their blogs or twitter, and try and figure out what they are about. If you like what you see, engage them and the community of people around them. Maybe you will never get to work with them directly, but in the process most likely you will connect with other like minded individuals and will find people who are either in the same position as you or sages who have done it multiple times before. Maybe you will get lucky and find a few life long friends. And that is what makes it all worth it, isn't it?

jhartmann··on Andrew Ng and the Quest for the New AI
Well I think a big part of this is that right now we have finally gotten to where the algorithms + the required computing power are starting to become more widely available. Cheap graphics cards or things like Intel's Phi + techniques like drop out to prevent overfitting are really enabling much more sophisticated things to be done in a reasonable wall time. Granted multilayer neural networks aren't a free lunch that will solve everything, but there are large classes of problems that are falling to these techniques all the time. We are also finding that Neural Networks scale very well to very large architectures, better than some of the other techniques. I understand we should be careful not to overhype since we have seen previous excitement cause a mass exodus from this research before. I however think people like Hinton were always right, and this was awesome stuff. We just couldn't really take advantage of it because we could never train it for long enough and we hadn't learned how to do things efficiently yet.
jhartmann··on Andrew Ng and the Quest for the New AI
I've said this before, but deep learning is terribly powerful precisely because you don't have to spend lots of time doing feature engineering. Multi layer networks that are trained in semi-supervised, unsupervised and supervised fashions all can now produce networks that meet or beat the state of the art hand created models for speech, handwriting recognition, ocr, and object recognition. We are only just beginning to see what is possible with these sorts of techniques. I predict within a few years time we will see a huge renaissance in AI research and Neural Network research specifically as these techniques are applied more broadly in industry. My startup is building some cool stuff around this technology, and I know there are hundreds like me out there. This is going to be a fun ride.
jhartmann··on Machine Learning Library for C++
I'm doing deep belief networks with dropout, and don't have access to GPU's with good double precision performance. I used to write graphics device drivers, so GPU computing has a special place in my heart and definitely agree with you there performance wise. It is funny though that my little laptop is hitting training times similar to some papers where people are using low end GPU's though, its amazing what you can do when you pay attention to performance.

I suspect my tuned C++ code will work quite well on a Intel MIC, and that is probably where I'm going to go when I have more resources to throw at the problem. I do know that Theano does use Alex's C++ CUDA code under the covers and I have done lots of reading of some of theano's code looking at implementation details to help developing my code. I just am not a big python (or most scripting languages actually) fan, perhaps I'm just too old school and written C, C++, C# and Java too long. If it doesn't smell or feel like C, I feel like Scotty in Star Trek 4 when he was making the transparent aluminum on the mac.

jhartmann··on Machine Learning Library for C++
You forget that if you can do the whole weight update in a single shot operation that the data doesn't have to go through the cache multiple times, and at least on the problem sizes I am working on FP bandwidth isn't what kills it, but the memory bandwidth. Back to the NN example: If you can do the matrix multiply and the application of the delta weights in a single loop iteration you get much better cache behavior.

Another thing about code generation, I am also using a hacked version of Eigen as well in a project I'm working on that can do the tanh and derivative of the tanh so the NN activations go quite abit faster since you can generate vectorized code for the whole calculation that will visit the memory location exactly once. While true the calculation of the weight updates is the most time spent, I saw 3-4x speedup in the activation code doing it in a single operation due to better memory access patterns and less loop iterations. Better memory access patterns can also have synergistic effects on other code because there is less cache pollution happening. By being fast and loose and introducing a few other copies of the matrix data in my case, my performance falls off a cliff when it no longer fits in the cpu cache nicely. 10x difference in the particular case I am remembering.

As always performance is part art, part science and perhaps it won't matter as much for the general case, but for my specific implementation and my matrix sizes Eigen has made a measurable difference for me compared to other solutions.

jhartmann··on Machine Learning Library for C++
FYI, another bonus on Eigen is if you are using a Matrix-Matrix operation you will get multithreading with OpenMP for free. It isn't the most efficient way (you can do better threading yourself probably), but it gets most of the way there for stuff that will use GEMM operations.
jhartmann··on Machine Learning Library for C++
You should consider using Eigen for linear algebra, I have personally found it much better performance wise than using bindings to ATLAS or other more standard linear algebra solutions. ML algorithms tend to be multistage (think about the update of weights with momentum in a Neural network for example), and the primitives available in ATLAS or a blas library are really too low level. Eigen since it generates code for a whole complicated expression can blow a standard linear algebra library out of the water for a certain class of problems. For others obviously the highly tuned vendor BLAS code would win, but I've seen huge speedups by using Eigen it fits well for complex ML operations.
jhartmann··on Machine Learning Library for C++
Actually you forget that performance when you need to train for days at a time is critical, if I use Octave/Matlab/R my current project might take months to train instead of weeks. All my ML code is high performance threaded C++. I recommend you use a good template linear algebra library like Eigen, you can do plenty of experimentation in C++. I find with a set of a few modern libraries and the required experience a C++ programmer is just as if not more efficient than a Python/R/Matlab programmer. It comes down to the skill of the programmer and the proper choice of libraries.
jhartmann··on Machine Learning Library for C++
Flame Suit On / Rant Mode On

I actually really don't understand why anyone uses GPL for a library. I've been doing open source for a long long time, and love the GPL. I have code in the Linux kernel, and believe free software AND open source software are great solutions to very real problems in software engineering. Having open code just gives people more options, and I firmly believe it will win over time as far as quality is concerned.

I just think only providing libraries to other GPL code is stupid. It just limits the usefulness of the software. LGPL is great here, you get the core changes contributed back to your library from a greater group of people and everyone wins. Limiting a library to GPL means a large population can not use your code, those writing applications that can't be licensed under the GPL. Limiting choice is BAD. The whole reason you should be creating and using free software and OSS is to not weld the hood shut. GPL should be for applications, LGPL just limits choices for libraries. Down with the GPL for libraries!!!

Flame Suit Off / Rant Mode Off

jhartmann··on Why You Make Less Money
One thing that is not mentioned is stressing exactly how valuable your skills have been to previous employers. Stating exactly how you saved X amount of money or added X amount of new revenue is a really powerful reason that some get paid more then others. patio11 has a really great post about this: http://www.kalzumeus.com/2011/10/28/dont-call-yourself-a-pro...

Remember businesses are about making money. The way to get paid more is to demonstrate you can make them more money.

← PreviousPage 3 of 5Next →