Earlier quoted context omitted.
Time to sell shovels
After the first year in whatever cycle it is always about shovels.
It's why everyone wants to sell "a platform" today...
After the end of the startup era
281–290 of 304 posts
Re: After the end of the startup era
#282Earlier quoted context omitted.
I have some questions as to why you would say that - could you reach out to me at my email or add yours to your profile? Thanks.
Why do you say thanks? They may not want to say any more than they’ve already said, or they may be unfriendly to you.
No worries, I'm not an unfriendly person.
Re: After the end of the startup era
#283>IoT devices [are] hard to prototype, generally low-margin, expensive to bring to market, and very expensive to scale. Just ask Fitbit. Or Jawbone. Or Juicero. Or HTC. I feel like this is the public's idea of "iot," while I've learned after a year of working for an iot stack provider that the real "internet of things" explosion is found in places nobody looks except the procurement guys - the stadium lights, the refr…
IoT devices [are] hard to prototype Here in Shenzhen, I would say no. They are cheap as chips. I have multiple friends doing consumer IoT devices (lots of BLE) with single-dimensional backgrounds (often software) who have been able to iterate through multiple prototypes and produce concepts within months. The chief problems for IoT products IMHO are marketing and replicability (low security for initial investment). T…
kind of off topic. but, how much markup is there compared to aliexpress? it seems prototyping using parts from there is pretty inexpensive. I heard about giant multi story marketplaces in Shenzhen, but I was curious if they were much cheaper or just a lot faster since everything you need is in one building.
Re: After the end of the startup era
#284Earlier quoted context omitted.
No offense, but that's kind of the fun of linux on the desktop. Also, if you really haven't updated in three years, that means that you're still running kernel 3.16, which has 81 known vulnerabilities[0], 12 of which are critical. That's the kind of stuff I don't really want to worry about anymore. [0] https://www.cvedetails.com/vulnerability-list/vendor_id-33/p...
No idea what's your definition of customization is but running package update command isn't what i consider as customization. And updating kernel/pkgs doesn't require you to customize them.
I get both sides of this.
For 17 years I ran linux everywhere. FreeBSD as well but infrequently as a desktop. After two years of mac and seeing the breakage and perversion of the unix foundations apple inflicts I gave up on it (entirely) and work with macs only under protest.
So now? Linux and *bsd for paid work. Windows for the desktop. cygwin and mobaxterm ftw. All old is new again.
Re: After the end of the startup era
#285I look at it from another end - I feel there is a sort of an app fatigue. Maybe it is jut me. I used to be excited about Uber-like things, note taking apps, DropBox, coupon sites, review sites. Installed some of them, used them, then kind of stopped. Someone at work would say "you should try this new thing, totally cool" and I'd be inclined to do it, but now would pass it probably. Extrapolating scientifically from t…
Mobile devices are very constrained. The OSes are locked down, the screens are small, high CPU apps are out, and input is slow and cumbersome. After trying a lot of apps I have found that a smart phone has only a few functions: 1. Talking and texting. 2. Taking pictures and video. 3. Maps and directions. 4. Music, movies, ebooks. 5. Brief interactions with services like Uber, Lyft, AirBnB, Los Angeles metro, etc. 6.…
I never really got the feature-phone for anything more than data and voice either. When you add in the downsides for those mobile features: MRC,1k for a decent phone, and the moving targets all the app providers put out it's easy to see through the facade to a bubble market based on peoples low tech IQ and attention deficit: both encouraged in/by other popular media.
The only upside is it's easy to track you with these devices by those who care about you the most: big $$ and big brother.
Re: After the end of the startup era
#286Earlier quoted context omitted.
IoT devices [are] hard to prototype Here in Shenzhen, I would say no. They are cheap as chips. I have multiple friends doing consumer IoT devices (lots of BLE) with single-dimensional backgrounds (often software) who have been able to iterate through multiple prototypes and produce concepts within months. The chief problems for IoT products IMHO are marketing and replicability (low security for initial investment). T…
kind of off topic. but, how much markup is there compared to aliexpress? it seems prototyping using parts from there is pretty inexpensive. I heard about giant multi story marketplaces in Shenzhen, but I was curious if they were much cheaper or just a lot faster since everything you need is in one building.
There's more choice online but you have to pay shipping for everything. Quantities available, prices and discoverability are less attractive in the marketplaces but you can potentially save time and there are certain stores where you can get what amounts to nontrivial free in person implementation or product selection consulting. In general we buy online because we source so many things, but if you are starting out in electronics or just need some standard parts for cheap then the markets are pretty cool.
Re: After the end of the startup era
#287Earlier quoted context omitted.
Comparing current day consumption to a single year in the past reeks of bias.
What kind of bias? I copy pasted a cluster of lines from that page, but the sources themselves are well documented, mainly from the gov itself. Anyway here's another source that shows that meat consumption has dramatically increased since the 1920s, though that wasn't my central point: http://www.npr.org/sections/thesalt/2012/06/27/155527365/vis...
Example:
I want to show that there is a lot more murder today than before because of guns.
I pick one year (a far edge of the bell curve type of year) to compare to today where there was the least amount of murders and compare it to day.
Without showing an average or showing even a distribution bias should be expected. There is very little reason for picking a single year to these types of arguments and it is too often done.
Re: After the end of the startup era
#288Earlier quoted context omitted.
No you cannot. A brick is also a physical product. But it doesn't communicate. There needs to be software and technology involved, and more than one object needs to talk via a network.
I think the GP's point is that software is not the hard part when you're dealing with physical products, connected or not...
Well who means what is really hard to discuss at this point.
The article's author wants to point out that the hardware part must be so cheap that you can't make any profit in that area any more. But you still need to get it right, otherwise you can't sell anything.
Komali doesn't disagree with that point, but adds another. That most of IoT is actually not in the consumer area but in the industry area where people don't see it but the scale in terms of item numbers might be much bigger. Also a good point.
Then leggomylibro points out that you could replace "IoT" with "physical product", which may or may not imply that hardware is the hard part in IoT.
In the literally interpretation I disagree with the equal sign there, because a brick is not "IoT". There shouldn't be much to say about htis. IoT _is_ about the communication, right?
Considering your interpretation of his statement, that the hardware is the hard part in IoT, makes me wonder how you define hard part. The physical important parts like chips are quite well understood I believe (not an expert though) and we churn these out in the millions each year and they mostly do their job. The point where these chips actually need to be connected to physical I/O and software, these are horribly underdeveloped areas. There is low amount of skilled people, low amount of debugging, low amount of standards. If that's the area you are talking about I agree.
However if we talk business, we also need to talk money. And while you have one or two suicidal hardware architect on such a project, you probably have 50-100 software developers, who will use open-source plugins, libraries, operating systems, which again are each developed by 50-100 developers. So the whole cost of software is margins higher than the costs of developing the hardware. And that factor is growing towards the software side on a daily basis. The hardware will grow more and more to standardized processing units (GPUs, SoCs, touchdisplays for I/O) and the main activity will happen in more and more in software.
So considering that part I have to say nope, software is the harder part (to finance, to develop, to finish) here.
Re: After the end of the startup era
#289Earlier quoted context omitted.
1) That's false. AlphaGo did not just train by self-play. It trained on millions of pre-played games, and the bot also uses hand-engineered features for their MCTS hybrid (best) model. Refer to their paper for details. 2) Read the article that you are referencing please. What you are implying is not the thesis of that article nor is it what Geoffrey Hinton is saying. What Hinton is saying is, we should throw out deep…
Sorry, but you are incorrect on both accounts. 1) AlphaGo Zero was indeed trained in the way I mention. 2) As directly quoted from the article, Hinton believes that a better way of learning doesn't require all that labeled data. If such a method is invented, as is required to push AI forward, big corporations would not have a data advantage, which is my original point.
AlphaGo, trained using supervised learning using games from KGS:
https://storage.googleapis.com/deepmind-media/alphago/AlphaG...
AlphaGo scratch: https://www.nature.com/articles/nature24270.epdf
I was specifically talking about AlphaGo, not AlphaGo scratch. Also, if you read the paper about AlphaGo scratch, the key innovation driving the self-learning is the use of MCTS as a policy improver, which couldn't have feasibly been done without AlphaGo and the supervised learning.
And I think Hinton is saying that we need fundamental breakthroughs for AI, and I don't think he is in favor of "traditional" modern neural network architectures. Anything that requires SGD, he doesn't like.
Re: After the end of the startup era
#290Earlier quoted context omitted.
I think the GP's point is that software is not the hard part when you're dealing with physical products, connected or not...
Well who means what is really hard to discuss at this point. The article's author wants to point out that the hardware part must be so cheap that you can't make any profit in that area any more. But you still need to get it right, otherwise you can't sell anything. Komali doesn't disagree with that point, but adds another. That most of IoT is actually not in the consumer area but in the industry area where people don…
What I mean by physical part being the harder here is that it's much slower and more expensive to iterate on.
You can make a lot of changes to software within a single day, for free (sans programmer time), and deploy them to all your customers. With physical items, you usually need to fab a new version for each significant change to test it, which is a time-consuming and expensive process. This is not something electronics-specific, this applies just as much to a new brick design.
This is a qualitative difference that makes physical product design an expensive process - you need to get all the things right before you start shipping; you can't just patch things after you deploy. But, as the article correctly points out, ultimately the manufacturing of a correctly designed batch of a physical item is cheap. Which means it's hard to make profit, especially if you spent all that money iterating in your lab, and then someone just reverse-engineers your final design and starts pumping out copies.
In case of IoT - common processing and communication chips are cheap and easy to get. Electrical engineering is hard and expensive, product design is hard and expensive, and - compared to that - software is cheap, because all IoT companies are doing is bog standard cloud-based CRUD.
(If they make the software part needlessly complicated for themselves, that's another problem, but it's endemic to this industry anyway.)