.
Profile URL | https://www.quora.com/profile/Bill-Paseman |
Question | Answer | Date |
---|---|---|
How do I find investors for my startup? | Here is what I look for in a pitch.
| 11/23/2017 |
How do I found a startup? | How to Start a Startup | 10/24/2014 |
How do I get a contract for the startup I've kind of taken over running? | Quit. If the guy asks you to come back, ask "why"? If he needs you, ask him to "make an offer". Make sure the offer has an equity provision that recognizes the contribution you have made so far. If he does not ask you back, then you are not running the company, he is. And apparently he is doing a good job if he has gotten you to work for nothing and simultaneously made you think that you are running things. If he doesn't' ask you back, what have you lost? Answer:a job that pays you nothing. In that case you can create a competing company if you have the nerve. After all, no contracts have been signed, so you will be violating nothing if you do this. | 10/28/2015 |
How do I get ideas for tech startups? | I used the process described in The Entrepreneur's Manual: Business Start-Ups, Spin-Offs, and Innovative Management: Richard M. White: 9781626548718: Amazon.com: Books to use $2K to build a $2B tech startup | 1/20/2018 |
How do I keep my organization members active and motivated without constant intervention from me? | Early on, I used the following methods at Calico (A company I took public).
| 12/10/2016 |
How do I protect an idea? | Well, first recognize that patents give you the “right to sue” and little else. If the idea is stolen, you have to defend it in court (or not). However, given this caveat, one strategy is to file a provisional patent and request that it not be made public for 1 year. You will still get patent protection from the file date forward. As such you can file a provisional patent, disclose the idea privately and NDA or not, if it is stolen, you have still patent protection. | 6/13/2016 |
How do I raise the funds for a new startup? | As discussed in some of the sources at Paseman and Associates: Pitch Checklist, I started by finding a method to correctly develop a company. Following Teknekron's lead (and analysis of the Okie on my site), I then did consulting to find the right market. Once I had the customer's trust, I told them that I was "Looking for an enterprise problem that was so bad, people would pay me to solve it and let me keep the software". After two years, I found a customer that both took me up on my offer and was in a broad market. I used their (and other early adopter) feedback and money to develop the product (another 2 years). I then switched from consulting to product sales mode. I incorporated and booked $1m in 6 months. I then went to the VCs who funded a previous failure of mine. I showed them a product, customers, positive cash flow, bookings and a competitor that grew to 32 million in sales in 3 years. I closed Mayfield, Kleiner Perkins and turned away Sequoia (they called too late). My total outlay was $2,000. I took $1m off the table at each round of funding and left when the company was worth $2B. So yeah, that was a pretty good strategy. | 10/23/2014 |
How do I request for an existing government database for my new IT project? | Most gavernment databases I work with have a developer page explaining the protocol. | 10/14/2016 |
How do I start a business in a field that I don't have much knowledge about? Is it possible to do this without going to school? | Find a customer and sell to them. What you sell, how many you can sell, how much it costs to produce and distribute and the price for which you can sell it will determine how good a business it is. Suppose you like hamburgers. If you can make good ones, you can be a fry cook. You could also find good fry cooks and wait staff, good supplies, find customers and serve them. You could then own a diner. If you get enough customers, you can own a chain of diners. How do you get customers? That's sales and marketing. Interestingly, you can make one of these products and see how much you can sell it for. etc. Paseman and Associates: Pitch Checklist has other points to consider. | 7/28/2015 |
How do I start with Artificial Intelligence? | Machine Learning - Stanford University | Coursera Also Peter norvig has more general text. Finally, deep learning is hot, here are some resources for that: Bill Paseman's answer to What are the most important deep learning algorithms? In which order should I learn them? | 3/5/2017 |
How do I stay motivated and on track as a 'lone wolf entrepreneur'? | Will Smith (A quote by Will Smith) pretty much nails my POV. But what is his motivation to stay on the treadmill? He doesn't say. I suspect that it is similar to mine. The truth is that ordinary people have enough sense to get off the treadmill and go see a movie, get a beer, eat a hamburger, etc. They'll be extraordinary. Later. So. If you are ready to admit, in your heart of hearts, in your very soul that you are ordinary, just like everyone else, why, then just get off. Will Smith? He'd rather die. | 5/1/2016 |
How do magic candles work when you blow them? | 2/2/2017 | |
How do they apply machine learning techniques to understand the patterns in financial trends? | check out Harris who reports results of the different Deep Learning approaches and Tefas who gives an overview without results. | 2/8/2017 |
How do you ask your investors to get diluted? | Tell them there is a new round of funding where the price/share has increased. | 12/5/2017 |
How do YOU challenge yourself and stay motivated in life or business? | Stay engaged and pay continual attention by Maintaining a variety of interests. In that way, most everything that occurs applies to one of them, and my only task is to figure out which interest and how it applies. | 9/26/2016 |
How do you describe how a proposed project is innovative and the rationale behind the innovation? | Metrics I use to measure a product’s innovation are its ability to meet a market need and to penetrate a market quickly. For example, by that measure Facebook, was more innovative than Myspace and Friendster although both were earlier and Facebook started by copying their features. What was the key difference? Facebook’s goal was not to help you make new friends or hookup, but rather to solidify relationships you already had. It did this by insisting initial members have an ‘edu’ suffix on their domain. | 7/10/2017 |
How do you find a side project idea? | Not sure what a “side project” is. Here is how I did initial product development for a company I created that eventually went public with 300 employees. I started as a contractor. That let me keep the IP of every idea I did on my own time. I then asked each customer what their biggest problem was (ie what would they pay the most money for). Took a while, but I eventually was able to identify a high value market. I then did product development using this scheme: Every time I begin a personal project with an idea, I fail at continuing it. What can I do so I won't change ideas fast? Hope this helps | 9/14/2017 |
How do you find all the second partial derivatives? | Find all the second partial derivatives. v=(xy)/(x-y)? has a good example. I believe that the number of derivatives grow a the square of the number of free variables. E.g. f(x) has one f'''(x), f(x,y) has 4 fxy, fxx, fyx, fyy, f(x,y,z) has 9: fxx, fxy, fxz, fyx, fyy, fyz, fzx, fzy, fzz, etc. | 3/27/2016 |
How do you get into the field of artificial Intelligence? | Sloth. I wanted to compose music but was too lazy to take the standard route. So I put composition rules in a computer and had it compose for me. One of the technologies was “constraints”, which I re-used 15 years later to create a company called calico, which I later took public. Both episodes were similar in that I started with a problem I wanted to solve, looked for an appropriate technology solution, and used that. | 2/2/2017 |
How do you go about being an entrepreneur as an Electrical Engineer? | Well, as an undergraduate, I got degrees in Electrical Engineering (Control and Communications) and ChE (nuclear and petroleum). I then went to MIT and got a (masters) degree in EECS. This background enabled me to get involved with a first class Silicon Valley Startup team. I then got 8 years of industry experience (one company went public and one failed) and started a consulting company. My background then allowed me to convince prospects that I could handle any problem they had. After working a while at each prospect, I always asked the same question "Do you have a problem that is so bad that would pay me to solve it and let me own the solution"). The answer was usually "No.". However, after 2 years, I hit a "yes". I then checked to see if the product was a market fit for other companies. It was. I built the product and I then had a developed product with one reference customer that solved a tough problem. I then sold lots and took the company public. | 9/26/2015 |
How do you implement Artificial Intelligence in an Arduino Project? | Following on to Anders Borg’s answer, First, it is important to distinguish between different types of AI: expert systems, k-means classification, Neural networks, etc. E.g. You could implement a minimal expert system loop in an Arduino. (It would look a lot like a command loop reading from Serial with a hardcoded case statement) Second you have to distinguish between development and deployment. In the expert systems example, the case statement would be created on a development system and downloaded to the Arduino. Neural networks are similar in that (in all likelihood) you will have to do training external to the part to determine the weights and then download the weights to the device for model evaluation. Let’s analyze speed requirements for a Neural Network: Basically a Neural Network is implemented as a series of multiply-adds. Look at the instruction speed for this basic operation on an Arduino and it will tell you how many you can do in a second. (Faster code Fridays: Understand division and speed of operations) Looks like it is pretty fast - microseconds fast. Let’s analyze space requirements for a Neural Network: Each two way multiply has one constant argument that you must store in firmware and one intermediate result that you fetch from ram (The “unit” output). How much space do you have? The uno and its ilk have 32–256k of flash memory and 2–8K of ram (Arduino - Memory). Due has 512K flash and 96K static. (Arduino - ArduinoBoardDue). If you use 4 byte doubles as “units”, then you have room for 2–8K/4 = 512 to 2048 units on an Uno+ and 96K/4 = 24K on a due. Let’s look at a Multi-layer Example: A trained Multi-layer neural network for MNIST (MNIST handwritten digit database, Yann LeCun, Corinna Cortes and Chris Burges) is here (Neural Net for Handwritten Digit Recognition in JavaScript) (Do a “show source”). From the page “The network has 784 input units (28 x 28 grayscale image, normalized to values ranging from [-1; 1]). These are fully connected to 200 hidden units, each having a bias parameter, giving (784 + 1) * 200 = 157.000 weights; the activations are fed through a logistic non-linearity. The hidden layer is fully connected to the output layer with 10 units, giving (200 + 1) * 10 = 2010 weights. The final output is computed with a 10-way softmax non-linearity, assigning class (0 - 9) probabilities to the input image.” So unit storage should not be a problem (784 + 200 + 10)*4. But you need a 4* (157,000 + 2010) = 640K for the weights. You wouldn’t have room enough to store this on an Uno or a Due. The obvious choices for a multi-layer architecture:
Re the last two, check out Microsoft’s Raspberry Pi work: Microsoft made its AI work on a $10 Raspberry Pi and Microsoft/ELL | 6/30/2017 |
How do you write a science fair proposal? | 9/19/2016 | |
How does a first-time founder best meet investors in Silicon Valley? | Best is by introduction. Which usually means a friend or a friend of a friend likes your idea enough to introduce you to a Vic known to them. | 3/12/2016 |
How does a fully vested founder who has quit the company differ from an active founder in terms of compensation upon getting acquired? | It depends on the company. All scenarios listed could occur. A more common one is when the founder leaves, the company is doing poorly and needs money. A round is then raised that "washes out" all who have left (e.g. suppose that there are 10,000,000 shares outstanding. 100,000,000 shares are raised at 0.01/share meaning that -all- prior owners now own 10% of the company, and the company's market cap is now .01 * 110,000,000 = $1,100,000). New stock/options are then issued to those who stayed: usually to maintain the prior percentage. E.g. a previously 1% owner (who owned 1% of 10,000,000 shares = 100,000 shares and now owns 100,000/110,000,000 = 0.09% of total shares.) is given 1,100,000 (1% of 110,000,000) shares. However in such a situation, often the founder is terminated anyway in order preserve the employee pool and raise morale for the remaining employees. | 3/15/2016 |
How does Iowa State Engineering compare with other top tier Engineering schools like Purdue and Georgia tech? | If the assumptions of the prior two contributors are true, then another thing you may be able to go to the state school for several years, and if you do very well, transfer (or graduate) to another school. Note the “may” and “do very well” part. Re “may”, check the transfer policies of the other schools. Re “do very well”, if the target does accept transfers, try to find out what the application looks like for undergrads who successfully transfer. “Doing very well” is a probably a good habit to get into regardless. I applied to MIT as an undergrad, didn’t get in, went to Rice, “did well” and then went to MIT for grad school (they paid for my RA). | 5/4/2018 |
How does one use the official batch normalization layer in TensorFlow? | I don’t know. Have you gone through the tutorial at TensorFlow? Also, you may be able to ask the question in two weeks at TensorFlow Dev Summit | 2/3/2017 |
How far into the future should a consequentialism project go? | Kinda depends on the project. Someone with a terminal cancer diagnosis will likely want to figure out the consequences of his actions until “End of Life”, regardless of relative maximums encountered along the way. | 5/3/2018 |
How far will cognitive systems develop and is the potential huge for them in the future? | Re accuracy, Below is how the imagenet recognition percentage has changed over time. Error rate moving from 26% to 3.1% over four years seems like pretty good progress to me. The application of the technology is pretty narrow (unstructured data like speech and images). Given the interest however, I suspect that application will expand. | 3/4/2017 |
How have you overcome the barriers of entrepreneurship as a young entrepreneur? | Well. You asked about how I did it. First, I worked for several successful entrepreneurs to see how they handled the issues. During that time, I made sure I was the best employee they had. Then I (incorrectly) prioritized my lessons, raised money and failed. I then looked at other business formation processes and settled on a model developed by Harvey Wagner. P&A: Pitch Checklist I regrouped and created a consultancy that forced me to focus on customer problems for several years, and allowed me to identify a good market. I built a product, sold it several times and then raised money when I was cash flow positive. I hired a CEO, became COB and left after the IPO. | 5/6/2016 |
How important is it as an entrepreneur to surround yourself with like-minded successful people? | If by ‘like-minded’, you mean “thinks the same way”, it was better for me to have a variety of people who viewed the same situation differently but could agree on one course of action. If by ‘like-minded’ you mean “share the same goals wrt the business”, it was critical. | 6/14/2016 |
How is a linked list a dynamic data structure? | Because you can replace the thing a pointer points at with something else. See Bill Paseman's answer to What are the five most important programming concepts? Lisp creates a whole language and programming environment around this idea. | 6/4/2018 |
How is deep learning used in drug discovery? | Checkout https://arxiv.org/pdf/1502.02072... | 8/17/2017 |
How is TensorFlow different from DataFlow in terms of computational modeling? | Well, a few points. First of all, Dataflow is a technique (just as CPUs and GPUs are techniques for doing computation), Tensorflow is an application that can run using that technique. Second, I believe that Google has extended the dataflow model to fit their own needs. https://qconsf.com/sf2017/pvldb/... As an early follower of Jack Dennis’s MIT work, this is gratifying since they are actually making it work for large scale commercial problems. Third, Dataflow works best on computations that can be parallelized (conversely “aren’t serialized”). This can include fluid flow calculations (weather prediction, nuclear explosions and the like), some Analog circuit simulations and also deep learning training and execution. | 4/30/2018 |
How might I build an Arduino powered project that measures flour for cooking? | 12/11/2017 | |
How much is a 1903 penny worth? | Looks like you could get between $22 and $100 for it. | 10/2/2017 |
How much knowledge of programming do I need in order to start solving Project Euler problems? | Later problems requires bit more expertise apparently Project Euler - Wikipedia | 11/8/2017 |
How much strategic information should be revealed by Startup owners/founders to investors? | So. Point 1. Can you get the money elsewhere? If not, you have to deal with them. Point 2. They need to get comfortable with you, your experience, your team, your market, etc. If you are very good, (Theranos good), you can raise the money just on reputation. You don’t need to reveal that your machines don’t work worth a damn at all. However, if you have no money, this is your first time, the team is unknown, just what are you proposing that they invest -in-? If all you have is plan and idea, expect to get grilled on that. | 8/10/2016 |
How much time does it take to raise a pre-seed/seed round from angels in Silicon Valley? | John Sechrest’s answer is correct, but can also be used as a metric. IN 1985 I asked asked John Jarve of Menlo Ventures: “You’ve funded a lot of ventures and are very metrics oriented. What metric is most predictive of startup success?” He said “How fast you close your funding. What’s interesting about this is that this single metric encompasses a lot of different things: the market, the team, the idea, the traction, etc.”. But bottom line, if you can close the deal quickly, it must have looked like a ‘winner’ to someone. Of course, even then, this is no guarantee. | 5/7/2017 |
How much time will it take to develop a new language from C? | It depends what you want the language to do. In the 1980's, after reading the first edition of Object-Oriented Programming: An Evolutionary Approach: Brad J. Cox, Andrew J. Novobilski: 9780201548341: Amazon.com: Books. I was able to "comment extend" c to create a language with objective C semantics. One comment identified a typedef struct as a class; one identified a typedef struct as a metaclass, one identified a procedure as an instance method and one identified a procedure as a class method. I then wrote two new procedures "S"(send) and "SS"(send super). Finally, I wrote a preprocessor that created tables for S and SS. This took a week or two and allowed me to write most of the code in the original Smalltalk books. So. A few points here. The above spec is concise, complete and explains the value added by the extension. I was not interested in redefining "+", so the additional work was also minimal. If you have some new ideas and are interested in seeing how they work, again, comment extension will allow you to isolate just what you are adding and save you some time. One additional point. The extension could not be written entirely in 'C'. As Brad Cox points out in the book, the last machine instruction in the generated assembly for s and ss had to be edited to change it from a 'call' to a 'goto'. This enables the language creator to link the 'message' (function signature) to the 'method' (function implementation) while leaving the stack frame undisturbed. 'Modern languages' (after fortran) had removed the 'goto' after Dijkstra's hissy fit (Page on utexas.edu) in order to 'protect' the programmer. I suspect that this castration also makes implementing 'continuations' tougher in pure C. It certainly makes porting tougher. Bjarne Stroustrup took a more complicated but similar route. He originally wrote a preprocessor for C++ which used C as the target language. His position was that he had written a new language. Some critics argued that a preprocessor that took something that looked like C as input and produced something that was C as output, did nothing. If you are interested in writing code that targets hardware, you may not need to change the input much at all, but will spend a lot of time on code generation. In the early 80's I was part of a team that simulated a graphical schematic language. We then had to retarget the simulator from software to a piece of dedicated hardware. Then the dedicated hardware allowed you to plug in an external chip as part of the simulation. Not much input change at all, but it did require a new code generator. About 2-3 weeks work. I wrote a Lisp interpreter in PL/M in 1984 using an early version of Structure and Interpretation of Computer Programs. It took a few days to write and had a syntax that was trivial for me to parse, but difficult for users to read. This has been a perennial problem for Lisp and, IMHO, THE primary reason that the language was, has been and always will be niche. Again, making people re-understand how "+" works is a guarantee for slow adoption. My final story is where I was the user, not the implementer. I was building music synthesizers for my Masters thesis using discrete TTL in 1978. Mead and Conway came to campus with a course whose final project was a set of student produced chips. So I had a chance to build a music synthesizer on a single piece of Silicon. Trouble was, their input file format was a text description of a series of labeled rectangles. After a few days, a member of our lab (Chris Terman) announced that he had a graphic editor for the format. I was relieved, but a little apprehensive since I was proficient in AT&T's "draw" program, and not eager to learn a new editor. I then found out how Chris was able to create a new editor so quickly. He had taken the draw program and added just one command: rectangle. I could then create my chip using graphically displayed rectangles and all the other commands worked as they had: open/close, save, print, etc. So, of course the answer is "it depends", but I hope you can develop a feel from these examples. | 10/15/2015 |
How realistic is the dream of successful entrepreneurship? | It helps to have a strategy covering Market, Team, Product and Financing and refine it. I raised money with the wrong team, product and market for my first venture (Atherton Technology) and failed. I then analyzed my mistakes and corrected them in the second venture (Calico Commerce) and succeeded by my metrics. (In particular, my total investment was $2,000. When I left, the company was valued at $2,000,000,000). At Calico, I didn’t raise capital up front since I didn’t want to re-live Atherton. At Atherton, I had wasted so much money I could have burned a BMW in the parking lot every day for four years. Whats more, at Atherton I had to drag around the wrong team after I hired them with all that money. So after leaving there, I read Harvey Wagner’s work (see P&A: Pitch Checklist ) and started a consultancy (Paseman & Associates) to get market insight. I then looked for a problem that was so bad that people would pay me to solve it and let me keep the software. This mantra proved to be a high bar but a good strategy since it covered 3 of the 4 listed points: determine market need, get initial financing and create an initial product. Once I found the market, earned the money and created the initial product, I formed Calico and booked $1M in my first 9 months of business. But Calico still needed the “right” team. Atherton proved that I was unable to determine team fit via a standard interview process. So I altered it by doing a standard interview and then bringing on candidates as consultants. This let me determine fit. I went through 3 VPs of Marketing/Sales (One industry analyst, one Harvard MBA and one top Salesguy at a High tech company) using this strategy before I found one that could do the job as I saw fit. This is one example strategy. There are others. And like military plans, they change on the field as they are executed. However if you approach the endeavor with a clear head, a good map and well thought out contingencies, your chance of success is greater. I also found this book helpful: The Entrepreneur's Manual: Business Start-Ups, Spin-Offs, and Innovative Management: Richard M. White: 9780801964541: Amazon.com: Books | 6/2/2018 |
How should I get ready for a Masters in Computer Engineering? | Thomas Wong is correct, but with a math background, you might might be able to bridge into hardware by looking at some control theory and communications (e.g. DSP) books. Coursera has some good ones: | 8/31/2016 |
How should I prepare for a machine learning Ph.D with a BS in computer science and work experience? | What is you ultimate goal? What will you do there that you can not do in industry? | 6/24/2015 |
How to label a million images per day to train an ML model for visual recognition? | A drone mounted video camera operating at 24 frames/second camera can take just over 1 million images in 12 hours. If it is programmed to fly around an object and photograph it at different altitudes throughout the day as the light changes,, you would have a million images with 1 label/drone-day. | 3/2/2017 |
How will you fund your own project? | Are they hobbies? Then you’d probably be funding them yourself. If they are for a product you hope to sell, I started a consulting practice that I used to gather requirements and create a prototype. Teknekron was expert at this. Checkout the Harvey Wagner references here: P&A: Pitch Checklist | 2/5/2017 |
How would you know when is time to stop practicing and learning, and is the "right time" to take action, in a situation when you only have one chance? | If you can recast this as the Secretary problem, meaning you know that you only have n tries; then you decide after n/e (37%) of the encounters. As discussed in Knowing When to Stop there is even a special solution for the case of only one try. An example of this is finding the lowest price gas station in an area. If there are about 10 stations, first get the prices of the first four. Then stop at either the 10th one or the one that has a price less than or equal to the lowest of the first four. However, a more helpful question (in my experience) is "How do you structure the decision for maximum return for minimum cost". I.e. work to make sure that no action toward the goal is fatal. In this approach, you are acting toward the goal all the time. „The Gambler´s Ruin“ und die kritische Wahrscheinlichkeit- Geeignete Risikomaße bei Anlagen zur Alterssicherung? („The Gambler´s Ruin“ and the Critical Probability of Gain) provides a formal framework for this in investing. The question here is "How certain do I need to be about the outcome of a decision to invest a fraction of my portfolio?". It has a "Modern Portfolio Theory" style motivation in that the goal is not to maximize profit but minimize the possibility that you will lose all of you bankroll in a series of bets. The probability (as I recall) needs to be about 60%. | 11/11/2014 |
How would you make a suit designed to keep you from being detected by infrared? | Borrowing from Stealth technology I would concentrate on creating triangles tuned to infrared | 10/15/2015 |
I am 18-years-old, and I make 100K USD per year, what do you recommend me to do to become a millionaire before I am 22-years-old? | Expanding on Lucy Pulver’s point, If you can get 10 people to work for you and get them to do whatever you are doing and pay them $90K, you have doubled your earning power. Give them less or hire more people. | 3/31/2018 |
I am a final year telecommunication engineering student.I'll need suggestions on very innovative projects I can work ok? | Radiation only travels a few meters through water. How would you create a sonic system to communicate with submersible drones? What are its limits (bandwidth vs distance) and what are the system components? Can you another carrier besides light and sound to communicate? We would like to keep a drone vertically overhead indefinitely. Can you design a system where the power source is on the ground? (e.g. firing lasers at a solar panel on the drone). What are its limits (height and power received)? What wavelengths would you use? (Note, this is a power dissipation problem). A flying aircraft carrier dispatches drones. The carrier pilot is on the ground as are all the drone pilots. The carrier is powerful enough to carry a receiver of any weight up to 200 lbs. This is wartime and the communications may be jammed. You can use lasers to carry the comm channels, but that limits distance for line of sight to carrier, and from there, line of sight to the drones. Can you do better? If not, what are the limits? | 10/17/2016 |
I am good at math, but I don't like it. It is important to me that I enjoy it. How can I learn to love math? | Tough to answer not knowing a) what you do enjoy and b) What Math you are good at (or how old you are). (e.g. Algebraic Topology is different from Trig). c) Why you want to enjoy it. Frankly, I’ve studied several mathematical fields, but have no ‘love’ of math. I do like how math lets me do stuff: computer music composition (Markov matrices, 12 tone rows, generative grammars, etc.) , Companies (version management, constraint systems for hardware configuration). One book long ago that gave me a different perspective on ‘math’ was Hofstadter’s “godel escher bach”. It helped me understand ‘formal systems’ in a straightforward way, and how to think about systems that actually ‘did math’, e.g. my music systems. So instead of working “inside the system”, looking at problem statements, understanding axioms and rules of deduction, etc., and doing the work myself, it made clear that I could create (using a very tired term) a ‘meta’ system that solved the problems for me. Since I am lazy at heart (I HATE to practice music), that idea was VERY appealing. | 12/21/2017 |
I am looking to invest in a computers startup a 100k$,from year up entil three from now, how do I find a startup? | The other guys here answered your question directly. I would however like to point out a couple of other issues. 1) Sector Diversification: Angel investment should be one part of a general asset allocation strategy. The means, among other thing, that you need to consider the percentage allocation. E.g. A 5% (not atypical ) target means you have $1,900,000 (excluding your home) that you are investing elsewhere. 2) Time diversification: The idea is to use the initial investment as an "option" to buy more later. That means the initial investment is the smallest amount you will invest. Dave McClure's 500 startups understands this well and gives the odds of finding a unicorn in his fund here ( http://www.slideshare.net/dmc500hats/vc-101-500-startups) | 9/11/2015 |
I am not a developer but I would like to learn how to program a neural network. Where should I start? | 1/6/2017 | |
I am not able to decide the topic on my project related to civil engineering? | What is the optimal process for creating a balsa wood bridge that can break world records in the shortest period of time? E.g. see www.SisterZone.com for a starting point and some tips. | 10/18/2016 |
I am totally lost. How would I begin to start to create code for a registration desktop program? | Several free "out of the box" programs that include registration as a subfunction are on Drupal - Open Source CMS In particular, look at Conference Organizing Distribution (Drush make files and install profile). If you are unfamilar with Drupal in general, I did a presentation on using it: Drupal 6 my experience. Although for Drupal 6 (COD is in Drupal 7), it should give you an idea of the prerequisites required to host and deploy such a site. | 11/11/2014 |
I am very worried about the future because what would happen if artificial intelligence would take over all the professions? | I would like to rephrase your question. I would argue that the basis of your concern interrelates technology, society and economics. In particular, I believe your concern is the social implication of fast economic change fueled by technology (AI in this case). If the change is slow enough, then Valerio Cietto’s answer is on point. An example is the change from agrarian to industrial society in the US, where farmers left the fields for the city to find easier work. There, big changes happened, but society “adjusted” and stayed largely intact. The US saw no revolution after 1776 since people were largely hopeful for their future. They knew that if they worked, saved and invested, they would have had a good future for themselves and their family. If Valerio is right, then some people may have difficulty, and the face of society may change, but by and large, the change will be orderly. For example, I suspect one change is the current trend of “kids staying with their families well past school” in Italy and the US. Multi-generational homes were common in the past due primarily to economic necessity, not familial harmony. If automation-fueled economic division continues to accelerate, I believe that a hidden fact will become quickly apparent. A (Capitalist) society cannot be sustained if all capital is concentrated in the hands of a few people and no (compensated) labor is required. If people are given no way to improve their lot, then we will probably have something like the Revolutions of 1848 - Wikipedia, where monarchies fell and democracies rose. Unfortunately, this is not 1848. Now a lot of automated military power is also concentrated in the top. This lets minorities rule majorities: North Korea, Iraq, Syria, etc. One solution adopted by these countries is just to kill all opposition. North Korea’s solution is to brainwash the populace. Some less charitable critics of the US and Russia also make that claim, although the brainwashing is done more subtlety. So, to your point, there is a lot to be concerned about. If the goal is to minimize casualties, the trick is to find out a new system that works for everyone and transition to it quickly. | 3/13/2018 |
I get overwhelmed when I start doing electronics or programming related projects as I think I may not be capable enough, how should I get over it? | 8/23/2017 | |
I got a B in a college course I took. How will the MIT undergraduate admissions see this? | I recently had an opportunity to speak with an mit dept head about what they are looking for. He said he was not looking for members of clubs or presidents of clubs but for people who created clubs to help reach a goal. Similarly, he said he liked people who explained how mit would help them accomplish something. I didn't know this 40 years ago when I applied, but at the time I was interested in computer music composition. I coopted the computer graphics class to write these algorithms and convinced a classmate to help me build a synthesizer to play the music in real time. I discussed the work in my mit application and said I wanted to work w Marvin Minsky. They let me in. | 12/30/2016 |
I have almost no work experience, no hobby, no college or professional education. Then not wise to start a business? | Sounds like good analysis to me……. Spend some time making mistakes as an employee first. | 8/5/2016 |
I have an idea for an invention but, I have no money, limited college education and I know this could change how we do things today. What should I do? | Figure out a way to implement it using the customer’s money. here is an example: Accidental Start-Up | 6/30/2016 |
I have no business idea. What happened in your life that you experienced that made you realize and execute your successful million dollar business? | I started a consulting practice in order to discover what problems rich companies had and were willing to pay a great deal of money to solve. | 8/3/2016 |
I have to do a project on carbohydrates. Can you give some ideas? | Get some dogs dying of cancer. Give them a ketogenic diet and see if their tumors shrink. | 5/15/2017 |
I have to make a huge project in order to get my certificate in high school. I've studied computer sciences. In what field of CS should I work? | What have you studied. What have you done. What do you like and what are your goals (beyond graduating)? | 1/20/2015 |
I like designing and I'm good at math but not a special genius. Is starting a software business the right career for me? | Well how good are you and how much nerve do you have? Napolean said “On s’engage et puis ... on voit.” - Basically "leap then look". 30 years ago, I knew a navy guy who fixed hardware, got a programming contract, read the manuals before he showed up the next day, worked his way into a programming career and retired a millionaire in his early 30's. Key (to me) was that he got that first contract. Try selling your services. If you want to start a business, you have to learn how to sell eventually anyway. It will require understanding the customers problem, coming up with a solution and charging a price they will pay. You will have to "read up" as you go. Make sure the job is good. Hire a programmer for the first job if you have to, and learn from him/her. Finally, once you go through the exercise, ask yourself if that is what you want to do. If "yes", then do it. | 7/16/2015 |
I need an video idea about hope for a school project. ideas please? | This might be “out” enough to be “in” at this point. Catholic Theology formalizes this. I’d start by interviewing a priest about the Catholic Church’s basic teaching on hope, segue to St. Jude and then see how particular catholics (e.g. nuns such as Mother Theresa) make it work in their lives. | 12/17/2016 |
I need help in creating a good title for my science fair project. Can you help me? | How hot is pink? How cool is blue? | 9/25/2016 |
I need ideas for a cool and unique art project. Does anybody have any suggestions? | Art works (crafts perhaps) that I like include 40 Incredible Examples of Optical Illusions in Photos as well as works by Escher and Scott Kim. In addition to symmetry, I appreciate the fact that different perspectives produce different pictures. Most of these works however are two dimensional. Can you create a 3D sculpture that transforms as the viewer walks around it? | 11/28/2016 |
I tend to have great projects but always stop at some point. What should I do? | 12/11/2017 | |
I want to learn to code, what steps can I take to become capable of coding my own programs? | It depends on what you want to do. That said, Python (Welcome to Python.org) is not a bad language to start with. You can enter the python interpreter and type in examples a line at a time to learn it. It has libraries (prepackaged programs) that you can combine with what you write to create games and websites. It's key limitation os that it is not fast enough to do some tasks; however it is certainly a fine place to start. | 10/19/2015 |
I would like to build a customized AI for my regulatory consulting practice. This would be game changing for my work - How do I accomplish this task? | Adding to Robby Goetschalckx point, you might also determine what stuff you have that might give you a competitive advantage. For example, if all the knowledge about regulatory consulting is in your head, you might want to create an expert system, which usually requires a long process to get the info out of your head into the machine. If you have a large database of regulatory filings which no one else has, you may decide to start with a database to make then queryable. If they are are already queryable, then the number of them that you have may determine what you can do with them. The place to start might be to hire a consultant in your geographic area to do an assessment of what you have, how it might be applied and costs associated with each. | 12/20/2016 |
I'm 24 and just now going to college, is it too late to become highly academic and brilliant like people who go to MIT or change the world? | No. As proof I offer the fact that most brilliant people who changed the world did not go to MIT. That said, most extraordinary people I know had (have) drive, a “sense of urgency” and realized that time was their enemy. So if you are serious, I’d get started now. (By the way, I did go to MIT). | 7/25/2016 |
I'm an aerospace engineer who loves deep learning. Can I educate myself using online courses, or do I need a PhD in order to start an AI company? | Depends on what the company does and if you are the one doing the coding. | 1/24/2017 |
I'm attending high school (junior year) without a goal. Should I drop out? | If you truely have no goal, why not just stick around? | 3/13/2018 |
I'm becoming terribly cynical, how can I create a change in mindset? | I think failure -does- mean something different if you have a lot of money. That is not cynicism, it is common sense. Consider a casino where you can play poker. There are different tables with different ante's. You could play your entire savings in one hand at a high stakes table, or learn how to play at lower stakes table and work your way up. Of course, you might not have an aptitude for poker. or patience, or skill. In which case it is best to find this out as quickly and cheaply as possible. | 5/9/2016 |
I'm interested in investing in Facebook regarding its development and investment in AI. If I buy a Facebook share, does that involve the future income from AI research and results? | That is certainly what an investment advisor would counsel. Keep in mind though that people leave, investments split off and large companies (eventually) move slowly. There is no guarantee that Facebook will stay on track, especially of the CEO is more interested in running for public office than making FB a leader in that field. | 8/7/2017 |
I'm not a naturally brilliant computer programmer. How can I compete with brilliance in the job market, and can I become a great programmer? | Programming is simply another method of human expression. By analogy, some prose may be brilliant, but most useful prose is clear, precise and short. So I would suggest that you practice writing for readability (as you would when writing a good essay). Paraphrasing Samuel Johnson 'Read over your code, and where ever you meet with a line which you think is particularly fine, strike it out.' | 12/31/2014 |
I've built an algorithm that accurately predicts the movement of the price of gold. What should I do with this? | Check out The Marketplace For Algorithmic Trading Systems | Quantiacs. If you have no better option and you believe their pitch, you might consider entering their contest. | 7/10/2016 |
If a person were thrown back in time to Europe prior to the Renaissance, and arrived with nothing from the current day, what are some modern scientific advancements he or she could make use of with the materials and technology available at the time? | "A Connecticut Yankee in King Arthur's Court" started by building Military inventions and then he branched out to peaceful ones. Following that tack, Military inventions buildable at that time would include the Welsh Longbow, which had enough force to penetrate armor or the chinese Repeating crossbow. Even more simply, Hans Zinsser points out in "Rats, Lice and History" that the key factor that determined whether a siege would succeed is .... wait for it ... how much wine you brought. At the time, they knew that if you brought large groups of people together and they drank water, they would get cholera. As such, when laying siege to a city, they brought wine since they knew you wouldn't get cholera from that. So you could introduce the idea of boiling water before you drank it to a general and literally change the course of history. | 5/22/2015 |
If I have 20% of a company, and the other has 80% and an investor comes in for 30%, is 15% taken from both or is the 20 and 80 diluted by 30%? | Your problem is not completely specified. Does the investor own 30% before or after investment? These calculations are less confusing if you do them in terms of Pre-money valuation and Post-money valuation | 3/15/2016 |
If I have an idea, that would be a great solution for some global problems. How could I reach the right people or investor to give the plan to them? | Why not just publish it? There are many venues (e.g. letters to the editor in various magazines) where "gatekeepers" will pass the idea along if it has merit. Keep in mind that this approach may solve the problem but not attribute the solution to you. As Harry Truman said: "“It is amazing what you can accomplish if you do not care who gets the credit.” | 4/29/2016 |
If I have no machine learning or computational research experience but have statistical background, is it possible to do research in this area at graduate school or do an internship? | I think it is possible to learn the necessary skills "on the job" if you come in with something to offer. One way to go about this is to look at the research interests of professors at some target universities/institutions, come up with an approach that addresses them and go pitch the approach. E.g. If some are interested in reducing the time it takes to train a particular model on a particular problem, then see if you have a way to reduce the number of samples required to do the training or test steps based on some cheap apriori statistical tests on the corpus. If you do (or even if you don't but have some ideas), ask the responsible party for an interview to discuss your idea. In sales, this is called "cold calling". It is a bit un-nerving and the success rate is not high, but it is possible. One tip is to pick out several universities and start with the one you want to work for -the least-. That way, if you blow the interview, you don't feel so bad, and have learned something you can apply to the next interview. In addition to a good idea, success here requires the ability to function while feeling embarrassed. However if you go in the the honest desire to help, you often find that people want to help you as well. Good luck! | 2/28/2015 |
If I raised $100 from VC for my startup, and I pocketed $20 for my own use. Would I be charged with anything? | Though David Rose is right, in essence the answer is “yes” unless you negotiated something beforehand. The crime would be embezzlement. However, you can also negotiate a deal BEFORE you take the investment. When I raised $4M in a round for one of my companies, I sold a portion of my stake for $1M as part of the round. So $1M for me and $3M for the company. | 7/25/2016 |
If someone is starting their way as an entrepreneur, what would be a good book to recommend? | more stuff at | 6/12/2017 |
If we use all our knowledge on AI and Machine/Deep Learning, could we employ it to model the way human neurons function? | We could. Keep in mind we don't know much yet. Check this out The possibility of processing with nuclear spins in the brain If anything remotely like this is true, we don't yet have the science to even understand it, much less reproduce it. | 3/7/2017 |
If you could go back to when you started programming, what would you suggest to yourself? | Well. I started at age 13 with BASIC in 1968. I'd have suggested that I think through the two implications of what I had realized immediately: that programming was akin to magic, and ultimately, it was just a tool that let me accomplish goals faster. Re Magic: One implication was that the more spells you know, the more powerful you become. Another is that you must learn the languages that allow you to access the spells. Another is that in order for the spell to effective, it's interface to the real world matters. An example of where I "got" this was in 1975, when Howie Johnson (Home) and I wrote music composition algorithms in PL-I and APL on an IBM-360 which drove a Tektronix 4013 terminal over a 300 baud line which drove a homemade digital music synthesizer in real time. This multi-disciplinary task required understanding how to write device drivers driven by some primitive AI (Markov chains at the time) in two languages in real time. Re real world interface: I could have had the algorithms print out sheet music; hand the music to an orchestra and then have them play it. Doesn't demo the same way at all. Re Goals: Beyond the pleasure of doing it, the above project helped me get into MIT but did little else. However, if I had thought through the implications, I would have not only realized Why Software Is Eating The World , but the order that the consumption would probably occur, where it would have the biggest impact and how I could personally leverage it. E.g. Finance is pretty much a pure thought construct anyway. My uncle, Hellmut D Scholtz, realized this in 1978. Had I fully realized the implications, I might have partnered with him earlier and done more quant stuff when there was a LOT less competition. Re Teams: One thing I was able to figure out quickly however was the value of small teams (minimally me and one domain expert). Too big and you spend a lot of time dragging around dead weight. If it is just you, you have to spend a lot of time educating yourself not only to the material, but to the gotchas that occur when you start doing things in the domain that have never been done before. Jobs did that with Wozniak, and Jobs didn't even need to learn to code. Anyway, hope this helps. | 4/19/2016 |
If you could, what book would you make everyone read? | The Bible, the Koran and the Constitution. Everyone seems to have an opinion on them, but few seem to have actually read them. | 10/23/2017 |
If you had control of 1000s of people for data collection, what would you make them collect? | I guess I'd do what you are doing now: ask them for ideas on what data is worth collecting. | 1/24/2017 |
If you receive 20k euros, in what kind of technology start up idea would you invest? | Bill Paseman's answer to What are the steps to be followed for creating a successful startup? I invested in myself. You could potentially do this investing in someone else. | 7/9/2017 |
If you were given 24 hours to learn one skill that would be your main expertise for the rest of your life, what would it be? | And you only have 24 hours to learn it? Negotiation. | 8/12/2016 |
If you were to do a research project with an endless budget, what would you do? | Find out how they managed to create an endless budget. | 5/5/2018 |
In general, are deep learn algorithms more mathematically sophisticated than classic machine learning algorithms? | Expanding a bit on Michal Illich's answer and taking from Bill Paseman's answer to What are the most important deep learning algorithms? In which order should I learn them? The current consensus regarding deep learning algorithms is that more data produces better results and processing this data in parallel is faster than processing it serially. Baidu's Bryan Catanzaro deals with this by segmenting problems across Nvidia GPUs and Google's Yangqing Jia and his team created the Caffe | Deep Learning Framework to address this using both functional and data parallelism. As such, the additional sophistication comes from addressing the scaling issues inherent in these type of networks. | 1/18/2015 |
In mathematical optimization problems, the first derivative is often used. Why not the second, or higher order derivatives? | This is a little out of left field, but in a lot of physical systems, the second derivative is a constant. (E.g. few things naturally experience continuous 'jerk' (x -> dx/dt 'velocity' -> d2x/dt2 'acceleration' -> d3x/dt3 'jerk')) | 3/22/2015 |
In research, what is the difference between insight, intuitive, and heuristic? | Re heuristic, I believe it is simply a ‘rule’ that doesn’t always work. E.g. ‘if person a is taller than person b, then a weighs more than b’ Lots of exceptions. A less flattering word in the social realm would be ‘prejudice’, ie you ‘pre judge’ the conclusion by using little (insufficient) analysis. Heuristics have degrees of goodness, eg ‘all objects propelled into the air eventually come back to earth’. Worked 100% of the time until devices were created that reached escape velocity. | 11/1/2017 |
In the case of bonus shares, cost of acquisition shall be nil, but if bonus shares are issued before 1.04.1981 the cost of acquisition shall be market value in 1.04.1981, why so? | Could be a variety of things. E.g. one may be that some shares were bought, expensed and put in a pool on 1/4/81. Those who were issued after participate in the pool. Those before, didn’t. Generally, stuff like this is done when a company is “up and coming”. In which case the size of the bonus plays a part as well. If the stock has been steadily rising, maybe the early guys got a lot more (at a higher price). Either way, you out check out the language with your tax atty. In the 90’s, some guys exercised options at a high price, the stock fell and they then sold at a lower price. They were then taxed at the high price, and the stock proceeds did not even cover the cost of the taxes. They then owed millions to the government. Not the way it was supposed to work. Why not just ask the guy who wrote the contract? | 10/1/2017 |
In your own words, what is computer science? | One interesting facet is that the idea historically was an an outgrowth of mathematics and mathematical considerations. E.g. "Under what conditions can you determine if an algorithm will complete its task". Various models of computers were proposed on paper that were used to answer this and other questions. Post's Tag system, the Turing machine and the Von Neumann architecture, which is what modern computers are based upon. Once computers were actually built, it was then determined that even if the algorithm can theoretically stop, it may not be practical. This lead to the field of computational complexity, where you could make statements like the following "Suppose all atoms in the universe were computers, and each was able to evaluates a chess move in one nanosecond and all had been working since the beginning of time, then they would still not have finished evaluating all chess moves". Len Adleman once said that this idea was non-obvious even to Gauss, since all the algorithms he proposes in his Principia Mathematica are exponential time (i.e. take an impossible amount of time to solve). Through time, computer science became less mathematically and more practically focused (since there are more engineers and businessmen than mathematicians). This lead to layers of software abstraction being developed to make programming easier (assembly language -> Fortran/Lisp -> Object oriented Languages). Practice in turn lead to new theoretical questions: How is it best to run an algorithm in parallel? How is it best to structure data? How can one determine if an algorithm (e.g. Google search) actually returns the results you want? And there is even a lot more after this.... | 11/18/2017 |
Is "entrepreneurship" just another word for "personal greed"? | Under capitalism, man exploits man. Under communism, it's just the opposite. John Kenneth Galbraith I remember working at ARCO in the 70's late nights, alone. I was in the top tier of raises that year: 8%. Profits were greater and several co-workers died in rig accidents. I made good money, but had no desire to die for Arco's balance sheet. And 8%/year seemed small payment for that kind of dedication. I remember working for Nixdorf in Germany in the 80's. The air conditioners turned off at 5:15 and the place was plunged into darkness by 5:30. I was told that in a social democracy, I better get used to participating in social groups outside of work. I left. In Silicon valley, I was told I could work as much as I wanted, the lights would stay on and we might have some extra money in our pockets after a couple of years. My boss made more, but I did fine. I then started my own company, and struck the same deal with my co-workers. Many of us did well. So I guess "group greed" worked better for me than "personal greed" and that worked better than "It's 5:00, screw it, let's get a beer". | 3/24/2016 |
Is a random text generator a form of artificial Intelligence? | It was in the 1960’s. In language, it became what is now called n-gram - Wikipedia analysis. It is mostly used for analysis, not generation. In fact, one of the earliest descriptions of unigram generation was in Gulliver’s travels (written in 1726!) (See Lagado’s The Engine - Wikipedia), where using it for generation was satirized as being one of many useless government funded projects. That, in turn seems to have come from Ars Magna of Ramon Llull (1275). I used markov chains for music composition in my teens based on this Scientific American Article (Pinkerton, R. C., 1956. Information theory and melody. Scientific American 194 (2), 77–86.). Note it was called ‘Information Theory’ because Minsky only coined the term ‘AI’ in 1958. The markov chain approach (for music) has been talked about as recently as a year ago (http://antor.uantwerpen.be/wordp...). As a side point, I wrote about my teenage work in my graduate application to MIT and they accepted me. So apparently it was non-nonsensical enough to be taken seriously in the admissions process in the 1970s. However, traditionally, AI sits a the boundary of what is well characterized and what is not. As new computational approaches are discovered, characterized, taught and commercially applied, the definition of AI ‘floats’ over time. ngram analysis is now one simple tool in a vast set used for language analysis and generation. | 10/25/2016 |
Is implementing deep learning on FPGAs a natural next step after the success with GPUs? | 2017: What’s Next In Neural Networking? and Shane Ryoo's answer to Can GPU architecture be scaled for emerging AI/ML workloads? 2016: In 9/2016 the 11th BAMMF on Sep 26 (Mon), 1:30pm - 4:30pm - BAMMF had a talk on nervanna’s Neon work. Here that are making a Deep learning chip with Intel (Doesn’t look like FPGA to me). 2015: See December 2015 ACM SIGAI Bay Area Chapter "Machine Learning on FPGAs: acceleration of CNNs and other large-scale algorithms" The comment stream includes:
2014: I asked Bryan Catanzaro this at his November 4, 2014 talk at Baidu (ACM SIGAI Bay Area Chapter). Expanding on Eric Jang's last point, he questioned whether special purpose hardware can get over the inherent advantage that GPU's have, namely a "day job" which finances their further development. This response mirrors my experience with other special purpose hardware (e.g. database accelerators). | 8/31/2017 |
Is it a good idea to make a project titled "The effects of depression in the economy”? | 12/16/2017 | |
Is it a sign of insecurity to not have confidence in one’s ideas or could it be that the idea is nonsense? | Many ideas require time to prove out. See the discussion at Bill Paseman's answer to Every time I begin a personal project with an idea, I fail at continuing it. What can I do so I won't change ideas fast? | 8/2/2017 |
Is it always better to find an investor who may do follow-ups at later stages or are new investors each round also OK? | Marquee investors are reluctant to do a round following a non-marquee investor. So for my startups, I worked to get a marquee investor the first round. This is especially helpful if you get in trouble at some point during corporate development, since non-marquee investors are more likely to pull out in later rounds. | 10/8/2017 |
Is it hard to build a power plant simulator using a huge artificial neural network for simulation and education purposes? | Does the dataset cover extreme cases? (E.g. Varying the controls to get the plant to shut down or explode?). Probably not. As such, creating a simulation with "objects" representing the various parts and loops would provide a better educational experience. | 3/22/2016 |