HackerNews is very developer-focused. If you guys saw what a radiologist does on a 9-5 basis you'd be amazed it hasn't already been automated. Sitting behind a computer, looking at images and writing a note takes up 90% of a radiologist's time. There are innumerable tools to help radiologists read more images in less time: Dictation software, pre-filled templates, IDE-like editors with hotkeys for navigating reports, etc. There are even programs that automate the order in which images are presented so a radiologist can read high-complexity cases early, and burn through low-complexity ones later on.
What's even more striking is that the field of radiology is standardized, in stark contrast to the EMR world. All images are stored on PACS which communicate using DICOM and HL7. The challenges to full-automation are gaining access to data, training effective models, and, most importantly, driving user adoption. If case volumes continue to rise, radiologists will be more than happy to automate additional steps of their workflow.
Edit: A lot of push back from radiologist is in regards to the feasibility of automated reads, as these have been preached for years with few coming to fruition. I like to point out that the deep learning renaissance in computer vision started in 2012 with AlexNet; this stuff is very new, more effective, and quite different than previous models.
Curious to know what sort of methods you used then if you don't mind sharing.
There might be some interesting things that can be learned from this kind of info and applied to the current status quo (I'm definitely not arguing that there is a sociopolitical element).
Maybe if MRI scans get cheap enough (due to advances in cheap superconductors or whatever) that it's economically feasible to scan people regularly as a precautionary measure (rather than in response to some symptom), then the bulk of the cost might then be in having the radiologist look at the scans. In those "there's nothing wrong but lets check anyways" cases, it might be better to just have the AI do it all even if its accuracy is lower, if it represents a better health-care-dollar-spent to probability-of-detecting-a-serious-problem ratio. (If the alternative is to just not do the scan because the radiologist's fees are too expensive, then it's better to have the cheap scan than nothing at all.)
Machine learning is already used in Radiology. Chances are eventually Radiology will be the domain of machines. But it's going to take some time to get there. Healthcare is extremely regulated and closed minded.
Most of the people in the thread you listed above are clearly biased towards medicine and against computer science and machine learning. But machine learning has been having success in diagnostic medicine even well before the deep learning boom that thread talks about.
For example: http://ieeexplore.ieee.org/abstract/document/1617200/
- Some RETS recorded requests/responses: https://github.com/estately/rets/blob/master/test/vcr_casset...
(Basically some XML-based (SOAP?), cookie + authorization, that seems very ASP.net / Windows Server centric)
- DICOM ("It includes a file format definition and a network communications protocol"): https://en.wikipedia.org/wiki/DICOM
It's basically how to imaging devices communicate and store the images.
Image examples: http://www.osirix-viewer.com/resources/dicom-image-library/
A video of what a doctor would see: https://www.youtube.com/watch?v=Prb5lcR8Jqw
TCP-based protocol in Wireshark: https://wiki.wireshark.org/Protocols/dicom
I wrote a little .dcm to .jpg based on ruby-dicom BTW: https://gist.github.com/Dorian/9e3eb5891b49926c15a05c641ffef...
- PACS seems just like a database model basically http://mehmetsen80.github.io/EasyPACS/
It's the server that is gonna give the info to the doctors.
It's interesting how there seems to be only one popular viewer: OsiriX
Main issue is with HL7 is not technical. From a business point of view, the incentive to cooperate with other systems via HL7 means another reason for a department to adopt a system other than yours.
These are examples of next-generation radiology companies. The current generation of products are focused on image storage and display. These new companies offer automated image analysis before the radiologist even looks at the image. iSchemaView does hemorrhage maps as soon as new head CT or head MRI is acquired.
It looks like everybody sitting on their data is hindering progress. Is there anything that can be done about that politically? I mean, in many cases the data belongs to the public anyway, unless people signed a waiver, but what is the legality of that?
I'm sure machines will someday take over radiology but there will be many, many jobs automated before it (i.e. decades).
There are three areas that take a lot of time that radiologist would like to see automated:
1. Counting lung nodules. 2. Working mammography CAD. 3. Automated bone-age determination.
Those are the hot three topics for machine learning. Personally, I think that a normal vs. non-normal classifier for CXRs would be more interesting because you could have a completely generated note for normal reads, and radiologists could just quickly look at the image without writing/dictating anything. Of note, hospitals and radiology departments typically lose money on X-ray reads because the reimbursement is $7-$20 (compared to $100+ for MR/CT). So if you could halve the read time, they might become profitable again.
Edit: In terms of 10x, what you'd want is a system that would automatically make the reads (i.e. radiologist report), and a very efficient way for radiologist to verify what is written. It's hard to make a pathologic read, but since roughly 50% of reads are normal, you could start with normal reports.
And then bringing checklist driven analysis for radiologist.
So I decided to observe an radiologist at an hospital for a day (back in 2011), and I noticed that most if not all of it was already automated.
Radiologists were there to rubber-stamp the machine's work (and to ensure compliance, laws and what not).
As an example, one survey (https://ashleynolan.co.uk/blog/frontend-tooling-survey-2016-...) put the number of developers who don't use any test tools at almost 50%. In the same survey about 80% of people stated their level of JS knowledge was Intermediate, Advanced or Expert.
We're currently working on a way to help devs test web app functionality and complete user journeys without having to actually write tests in Selenium or whatever. The idea to is let devs write down what they want to test in English ("load the page", "search for a flight", "fill the form", "pay with a credit card", etc), then we'll use NLP to discern intent, and we have ML-trained models to actually execute the test a browser.
You can give us arbitrary assertions, but we also have built-in tests for the page actually loading, the advertising tags you use, page performance, some security stuff (insecure content, malware links). At the end we hand you back your test results, along with video and perf stats. It’s massively faster than writing Selenium, and our tests won’t break every time an XPATH or ID changes.
>tries to do so by using an imprecise, context-dependent language designed for person-to-person communication to instruct a machine
???
Selenium is its own can of worms, but it absolutely sounds like you're using the wrong tool for the job here. The problem stopping people from writing browser-based tests is not that people can't understand specific syntaxes or DSLs, it's actually the opposite: people don't have a good, reliable tool to implement browser-based testing in a predictable and specific way that does what a user would intuitively expect.
Selenium fails here because it has to manage interactions between browsers, because selectors are hard to get right on the first try and continually break as the page's format changes, because JavaScript can do literally anything to a page and that is really hard to anticipate and address reliably from a third-party testing framework like Selenium, especially if components are changing the DOM frequently, etc., because Selenium is subject to random breakage at the WebDriver layer that hangs up your (often long-running) script, and so on.
Whatever the right answers to a next-gen Selenium are, attempting to guess the user's meaning based on Real English by something that is itself an imperfect developing technology like NLP is pretty obviously not the correct toolkit to provide that. Remember, a huge amount of the frustration on Selenium comes from not having the utilities needed to specify your intention once and for all -- the ambiguities of plain English will not help.
If your thing works, it will have to end up as a keyword based DSL like SQL. SQL is usually not so scary to newcomers because a simple statement is pretty accessible, not having any weird symbols or confusing boilerplate, but SQL has a rigid structure and it's parsed in conventional, non-ambiguous ways. "BrowserTestQL" (BTQL) would need to be similar, like "FILL FORM my_form WITH SAMPLE VALUES FROM visa_card_network;"
The biggest piece that's missing in Selenium is probably a new, consistent element hashing selector format; each element on the page should have a machine-generated selector assigned under the covers and that selector should never change for as long as the human is likely to consider it the "same element". The human should then use those identifiers to specify the elements targeted. I don't know how to do that.
The second biggest piece that's missing from Selenium is a consistent, stable WebDriver platform that almost never errors out mid-script; this may involve in some type of compile-time checking against the page's structure or something (which I know is hard/possibly impossible because of JS and everything else).
That said, I'll be thrilled if React Native gives rise to higher quality apps in situations where a native app is unavoidable (e.g. my bank's app).
99% of web apps need the same features but most of this is still up to manually rolling your own.
I should be able to clone some repo, enter some DO/AWS/GOOG keys and push.
This makes the complexity problem much easier to solve, as the code is (should be) less likely to cause an unanticipated mutated state which can't be easily tested for.
I suspect the reality is a small subset of programmers think functional programming is amazing and everyone else hates it. You might think it reduces complexity, but a lot of people feel it reduces comprehensibility.
Huge and seemingly often unacknowledged issue these days. And many attempted solutions seem to be adding fuel to the fire (or salt to the wound) by creating more tools (to fix problems with previous tools) ...
Red (red-lang.org) is one different sort of attempt at tackling modern software development complexity. Its like an improved version of REBOL, but aims to be a lot more - like a single language (two actually, RED/System and RED itself) for writing (close to) the full stack of development work, from low level to high level. Early days though, and they might not have anything much for web dev yet (though REBOL had/has plenty of Internet protocol and handling support).
You're asking the wrong question. It shouldn't be "how do we get people to slow down?" It should be, "how do we make rapid software development better?"
Not too long ago (in human-years, not internet-years). Most node packages weren't built with unit testing. Now its quite common in the popular packages.
Website UI is probably the same thing. After all, it took us a really long time till we got the whole HTML5 spec finally stabilised.
So you will probably see the tipping point occur over the next 10 human years, or less.
And just like you I been really frustrated with the inadequacy of UI testing tools, especially with Selenium. So like @donaltroddyn, I set out to develop my own UI testing tool (https://uilicious.com/), to simplify the test flow and experience.
So wait around, you will see new tools, and watch them learn from one another. And if you want to dive right into it, we are currently running close beta.
This also brings me to Traefik, one of the coolest projects I have come across in the last months.
Traefik + DC/OS + CI/CD is what allows developers to create value for the business in hours and not in days or weeks.
Also, we deploy to production at least 4 times a day, the time from commit to deployable to production is about 30 minutes. And because it is a container it will start with a clean, documented setup (Dockerfile) every time. There is no possibility of manual additions, fixes or handholding.
AMI only runs on AWS. Docker runs on anything. I don't think "versatile" is the word you are looking for.
We mainly use DC/OS to run more services on less instances.
From an "I just want to get my app deployed" perspective it may still be best to just use Heroku. But from a "new developments in the field" perspective, the fact that I can rent a few machines and have my own Heroku microcosm for small declining effort is pretty cool.
- Meta-tracing, e.g. PyPy.
- End-to-end verification of compilers, e.g. CompCert and CakeML.
Programming languages:
- Mainstreamisation of the ideas of ML-like languages, e.g. Scala, Rust, Haskell, and the effect these ideas have on legacy languages, e.g. C++, Java 9, C#.
- Beginning of use of resource types outside pure research, e.g. affine types in Rust and experimental use of session types.
Foundation of mathematics:
- Homotopy type theory.
- Increasing mainstreamisation of interactive theorem provers, e.g. Isabelle/HOL, Coq, Agda.
Program verification:
- Increasing ability to have program logics for most programming language constructs.
- Increasingly usable automatic theorem provers (SAT and SMT solvers) that just about everything in automated program verification 'compiles' down to.
I don't know much about CPUs, but I suspect that one of the core problems of software verification, the absence of a useful specification, isn't much of an issue with hardware.
I'd be really interested in applying any of these techniques to a full TLS implementation.
[1] http://research.microsoft.com/en-us/um/people/fournet
[2] K. Bhargavan et al, Implementing TLS with Verified Cryptographic Security. http://research.microsoft.com/en-us/um/people/fournet/papers...
Can you talk more about this? I even got THE book on this (haven't really read it yet though) and like I think I get the rough ideas but I'd be curious to hear what HTT means to you (lol).
In HoTT, there is an extension of inductive types that allows you to, not just have constructors, but also to impose "equalities" so these generalized "quotients" really have first-class status in the language.
As far as "exciting developments" in HoTT, the big one right now is Cubical Type Theory [1], which is the first implementation of the new ideas of HoTT where Higher inductive types and the univalence axiom "compute" which means that the proof assistant can do more work for you when you use those features. I just saw a talk about it and from talking to people about it, this means that it won't be too long (< 5 years I predict) before we have this stuff implemented in Agda and/or Coq.
Finally, I just want to say to people that are scared off or annoyed by all of the abstract talk about "homotopies" and "cubes", you have to understand that this is very new research and we don't yet know the best ways to use and explain these concepts. I certainly think that people will be able to use this stuff without having to learn anything about homotopy theory, though the intuition will probably help.
HoTT brought dependent types and interactive theorem proving to the masses. Before HoTT, the number of researchers working seriously on dependent type theory was probably < 20. This has now changed, and the field is developing at a much more rapid pace than before.
How much do you know about modern testing, abstract interpretation, SAT/SMT solving? In any case, as of Feb 2017, a lot of this technology is not yet economical for non-safety critical mainstream programming. Peter O'Hearn's talk at the Turing Institute https://www.youtube.com/watch?v=lcVx3g3SmmY might be of interest.
Transfer Learning (so we need less data to build models) http://ftp.cs.wisc.edu/machine-learning/shavlik-group/torrey...
Generative adversarial networks (so computers can get human like abilities at generating content) https://papers.nips.cc/paper/5423-generative-adversarial-net...
Furthermore, if we consider that most of these DL paper completely ignore the fact that the nets must run for days on a GPU to get decent results, then everything appears way less impressive. But that's just my opinion. I love working in deep learning, but we still have LOTS of work to do.
The time from theoretical paper to widely deployed app is smaller in DL than in any other field I have experience with.
It's going to take less and less time and money to train a useful model.
Does this only apply to artistic content, or also to engineering content ? Say PCB layouts, architectural plans, mechanical designs, etc ?
To get a better understanding (other than reading a paper) read this excellent blog post https://openai.com/blog/generative-models/
[edit]: it is a lot harder to build a NN when there are very constraint rules. But it is also a lot easier to verify and penalize it and generate synthetic data.
Lambdas are lightweight function calls that can be spawned on demand in sub-millisecond time and don't need a server that's constantly running. They can replace most server code in many settings, e.g. when building REST APIs that are backed by cloud services such as Amazon DynamoDB.
I've heard many impressive things about this way of designing your architecture, and it seems to be able to dramatically reduce cost in some cases, sometimes by more than 10 times.
The drawback is that currently there is a lot of vendor lock-in, as Amazon is (to my knowledge) the only cloud service that offers lambda functions with a really tight and well-working integration with their other services (this is important because on their own lambdas are not very useful).
Your input is tightly restricted, and with Amazon in particular, easy to break before you even get to the Lambda code (the Gateway is fragile in silly ways). Your execution schedule is tightly controlled by factors outside your control - such as the "one Lambda execution per Kinesis shard". You can be throttled arbitrarily, and when it just fails to run, you are limited to "contact tech support".
In short, I can't trust that Lambda and its ilk are really designed for my use cases, and so I can only really trust it with things that don't matter.
Auth0 has Web Tasks: https://webtask.io/
Am sure there are many more implementations out there. Agree that vendor lock-in is always a concern.
But the reality is that they don't, with cold-start times upward of 30 seconds. If you use them enough to avoid the cold-start penalties, then you're better of with reserved instances because lambdas are 10x the price. If you can't handle the 30 second penalty then you're better off with reserved instances because they're always on. If you have rare and highly latency-tolerable events, then use lambda.
There is no cutting edge with serverless on AWS.
Most developed implementation is BayesDB[1], but there's a lot of ideas coming out of a number of places right now.
e.g. store customer orders in the DB, and query `P(c buys x | c bought y)` in order to make recommendations - where `c buys x` is unknown, but `c bought y` occurred, and we know about 'other cs' x and y.
Is that sort of how it works?
The way I see, the real utility comes from the fact that domain models such as those in a company's data warehouse are typically very complex, and a great deal of care often goes into mapping out that complexity via relational modelling. It's not just that c bought x and y, but also that c has an age, and a gender, and last bought z 50 days ago, and lives in Denver, and so on.
Having easy access to the probability distributions associated with those relational models gives you a lot of leverage to solve real life problems with.
http://empirical.com (still dark atm)
co-founded by CEO Richard Tibbetts, who was also a co-founder of StreamBase (acquired by TIBCO).
There's been a renaissance of study in placebo effects, meditation, and general frameworks for how people change belief for therapeutic purposes or otherwise, but to me, that's been going on for a long time and is more about acceptance than being a new development.
One of the most exciting developments that's been coming out recently is playing with language to do what's called context-free conversational change.
Essentially, you can help someone solve an issue without actually knowing the details or even generally what they need help with. It's like homomorphic encryption for therapy. A therapist can do work, a client can report results, but the problem itself can be a black box along with a bit of the solution as well since much of the change is unconscious.
It works better with feedback (a conversation) of course, but often can be utilized in a more canned manner if you know the type of problem well enough.
I'm working on putting together an automated solution that's based on some loose grammar rules, NLP, Markov chains, and anything else I can use to help a machine be creative in language to help people solve their own problems, but as a first step as a useful tool for beginner therapists to help them get used to the ideas and frameworks with language to use.
So essentially, I'm getting a good chunk of the way toward hacking on a machine that can reliably work on people's problems without having to train a full AI or anything remotely resembling real intelligence, just mimicking it.
Before you go thinking, "Didn't they do that with Eliza?" Well yes, in a way, but my implementation is using an entirely different approach.
with all due respect, said politely, it is my opinion that you are a charlatan.
I wasn't interested in long citations or garnering proof of my work in particular with training a machine to do this work. I simply wished to add to this thread and did so, in order to show someone out there, maybe even you, what else is going on that is exciting in my little corner of the world.
I'm not that good of a programmer, so it's not in a state that it does work yet. I hope my original comment didn't suggest otherwise, but let me be perfectly clear here: I have no working machine implementation that can do what I want yet. It can work with simple canned responses like Eliza, but it's not enough. I am working on employing all of the techniques and tools mentioned, but progress is slow.
However, this is work and change I employ daily with my clients professionally and I can assure you that it does work.
You don't even have to take my word for it.
Consider....seriously consider: who would you not be if you weren't you?
If you thought about that one for a sec and felt a little spaced out for a second, you did very well.
If you came up with something quickly like "me" and didn't really actually consider the question, allow me to pose another to you. Again, seriously consider this. Read it a few times. Imagine emphasis on different words each time.
Who are you not without that problem you are interested in solving?
This work can be made more difficult by text only and seriously asynchronous communication, which is why I mentioned it being easier within conversation.
If you are interested in more, google "mind bending language" or "attention shifting coaching" and find Igor Ledochowski and John Overdurf. Their work has helped me change the lives of thousands.
This is true of basically every post in this thread. If you're interested in learning more, there are more friendly ways of asking.
Igor Ledochowski - http://hypnosistrainingacademy.com
John Overdurf - http://JohnOverdurf.com
As far as context-free therapy goes, that's a bit of an advanced subject, but can be learned and mastered through some of their programs.
The key tenets are simple though. As a model, consider that human language builds around 5 concepts: Space, Time, Energy, Matter, and Identity. These 5 also map cleanly to questions (5Ws and H) and language predicates in human language. Space is Where, Time is When, Energy is How, Matter is What, and Identity carries two with Why and Who.
Every problem you've ever had is built up of some combination of the 5 in a specific way, unique to you.
The pattern of all change is this:
1) Associate to a problem, or in other words, bring it to mind.
2) Dissociate from the problem, or basically get enough distance from it so that you can think rationally and calmly. Similar to a monkey not reaching for a banana when a tiger is running after it, your brain does not do change under danger and stress well. It can, but that usually leads to problems in the first place.
3) Associate (think about, experience) a resource state. Another thought or experience that will help with this one, for example if someone were afraid of clowns, I'd ask a question like, "What clowns fear you?" It usually knocks them out of the fear loop for a second.
4) While thinking about the resource, recall the problem and see how it has changed. Notice I said has changed. It always changes. You can never do your problem the same again. Will this solve things on the first go? Maybe. Maybe not, but it's enough to get a foothold and a new direction and loop until it's done.
Which is what makes this fun and exciting to do in person and fun and exciting to help teach a machine to mimic it to.
I find it difficult to deny the achievement of tangible progress that is implied by, for example, the self-driving car.
I believe the biggest advancement in the field of education is going to come with VR. With VR, we can dramatically reduce the cost of "learning while doing", which should be the only way of learning. With AI, we can provide highly personalised paths for learners.
VR and AI technologies are finally coming to a point where together they can provide a breakthrough in industries which are mostly untouched since decades.
I think, for middle school, it's easy to underestimate how much of education is not actual content. How do you deliver education that targets the teenage anger / passivity / disappointment / and emotional roller coaster?
I'm 100% with you on this. I've been saying this since VR became mainstream, I'm dying to start something in the e-learning space that takes advantage of VR/Augmented Reality but have no idea where to start.
With the brutalist movement something new started. People went back to code editors to create websites by hand skipping third-party, non-web-native user interface design tools prefilled with common knowledge making websites looking uniform.
The idea of design silos and brand-specific design thinking is dropped: no more bootstrap, flat design, material design, etc.
It's like back to the nineties and reinventing web design. You start from scratch, on your own, and build bottom up without external influence and or help.
It's about creativity vs. the bandwagon, about crafting your own instead of putting together from popular pieces.
[This][https://ruudvanasseldonk.com/]
It looks a lot like the PRUs in the Beaglebone black, where you could have normal, non-real time OS (linux) run on the main cores, but delegate real time tasks to the minion cores...and they share memory directly with the main cores.
Basically, you end up with the capabilities of both a Raspberry PI and an Arduino, on a single chip.
[1]http://www.lowrisc.org/docs/memo-2014-001-tagged-memory-and-...
Of course, freely-available and well-supported CPU IP can be very cool!
No one needs Babel to write stellar code IMHO. Unfortunately it is not about the quality of the code you write, it is about being politically correct. This whole ES6/ES7 thing is much based on what Coffeescript, Livescript, etc.. already did much better more than 5 years ago. And I dare to guess that most of the Babel proponents don't even realise it's just a transpiler that they will need till the end of the projects live.
note: I expect serious down votes as opposing Babel is almost a serious crime nowadays and proves my unlimited stupidity.
No, web development is not really exiting nowadays, it is more terrifying, where the hype goes tomorrow? Maybe soon I will be forced to write in MS Typescript if I want to be taken seriously. Same counts for Redux because Flux is so 2014.. you must be very brave not using Redux! I can go on and on, way too many examples..
Finding a web developer job now is particularly about complying to made up standards that become more complex every day. And I've seen quit some horrible code bases that perfectly comply! It's a very sad reality.
1. You can write great applications without the latest language features 2. The latest language features do make development easier
Babel is necessary for #2 if you don't have control over the browser which your users use to access your site. If you don't want to transpile, don't. It's as simple as that. However, the future of JS is the future of web development, that is indisputable. Using Babel allows you to stay closer to the future and/or use these great new language features.
You also brought up TypeScript.
3. Types make development much easier
TypeScript is a combination of types and a transpiler for the ability to use the latest ES features. Types are great, providing:
- Better self-documenting code - More safety - IDE interop to provide completion, as seen in VS Code
> note: I expect serious down votes as opposing Babel is almost a serious crime nowadays and proves my unlimited stupidity.
From the HN Guidelines: "Please don't bait other users by inviting them to downvote you or proclaim that you expect to get downvoted."
> Maybe soon I will be forced to write in MS Typescript if I want to be taken seriously.
Many would say that someone should be forced to write in a typed language in general in order to be taken seriously.
In the JavaScript community there's been an explosion of tools, tool chains, libraries, module systems; without a lot of experience, it's very easy to be overwhelmed. Criteria for choice can include the availability of easy-to-understand tutorials as much as understanding the motivations and design decisions and limitations of the tools themselves. Some people are very good at writing blog posts describing what they've done, better than conveying the underlying concepts. I don't want to necessarily fault them for that, as I think there's often a sincere desire to show people a way that works for them.
Rather than focussing solely on what's wrong, it's very important to also include options you think are better, and why. Looking down your nose at others, describing them as js hipsters using politically correct tools does no one any good, and makes it even less likely people will listen to what you have to say. And on HN, expressing the expectation of down votes is a guaranteed method of receiving them; it's against the guidelines and adds nothing to your comment.
There are a lot of choices in the web development space today. The desire to standardize on something (such as the push for Babel and Webpack) is laudable in that they recognize that so much choice is not necessarily good: it makes it more difficult to decide what to use (sometimes good enough is just that), and splits resources that may otherwise be used to improve a more limited number of options. That's not to say Babel and Webpack are the best options: just that I understand the motivation for standardization and push to popularize a few (rather than all) options.
[0] https://en.bitcoin.it/wiki/Lightning_Network
[1] https://lightning.network/
[2] https://download.wpsoftware.net/bitcoin/wizardry/mimblewimbl...
[3] https://bitcoinmagazine.com/articles/mimblewimble-how-a-stri...
And within the wider space of blockchains, improving access to strong anonymization techniques appears to be moving forward quickly: https://blog.ethereum.org/2017/01/19/update-integrating-zcas...
No easy solution here, what is the ^Bandcamp of concert venues^? [0] Is there a venue problem of, "Where do you play?"
[0] I know the solution is a political one due to land usage, sound restrictions and venue size.
With the right program and a distinctive chemistry to target in the unwanted cell population, this flexible technology has next to no side-effects, and enables rapid development of therapies such as:
1) senescent cell clearance with resorting to chemotherapeutics, something shown to extend life in mice, reduce age-related inflammation, reverse measures of aging in various tissues, and slow the progression of vascular disease.
2) killing cancer cells without chemotherapeutics or immunotherapies.
3) destroying all mature immune cells without chemotherapeutics, an approach that should cure all common forms of autoimmunity (or it would be surprising to find one where it doesn't), and also could be used to reverse a sizable fraction of age-related immune decline, that part of it caused by malfunctioning and incorrectly specialized immune cells.
And so forth. It turns out that low-impact selective cell destruction has a great many profoundly important uses in medicine.
We're porting a sizable application to .Net Core so we can be on Linux and save cost and time on instance launch.
I'm writing an in-depth blog post series about the process because I haven't found any significant migration stories. I'm hoping it will help a lot of people through the process.
C# has become such a joy for me to work with. The language itself has been progressing at an impressive pace without adding too much superfluous stuff or making it unwieldy. I feel like it's gotten to the point that most of the pain of strong typing is gone without sacrificing any of the benefits.
Perhaps it will be detectable, with technical effort, for some time to come, but as a propaganda and government corruption tool it will complete the circle started by the "telescreen/ankle bracelets" we all carry in our pockets.
Flexible solar panels, LED lighting with open source drivers, and the new generation of DC refrigerators are all incredibly exciting and are allowing us to experiment with living without grid electricity.
Their current scale-up of instruments I think means that they're looking to aggressively push into diagnostic applications.
The lack of competition is unfortunate however.
In my view, there are still a huge room of applications where wireless and sensor combined, and we already have web/native platforms. This is so exciting development!
Could you please elaborate on this. I don't understand how the camera industry could go beyond image sensors. They wouldn't be the camera industry anymore if they did that.
Can you please expand on this ?
TLDR: Fancy fused infrared (LWIR/SWIR) and visible spectrum camera systems may 'soon' be on a passenger airliner near you.
Using infrared cameras to see through fog/haze to land aircraft has been happening for a while now, but only on biz jets or on FedEx aircraft with a waiver. The FAA has gained enough confidence in the systems that they have just opened up the rules to allow these camera systems to be used to land on passenger aircraft.
Combine that with the fact that airports are transitioning away from incandescent lights to LEDs (meaning a purely IR sensor system is not longer enough), and you get multi-sensor image fusion work to do and a whole new market to sell them to.
Here is a blog post (from a competitor of ours) talking about the new rules.
https://blogs.rockwellcollins.com/2017/01/17/worth-the-wait-...
Say with a car that has a heads up display for night vision, if it had an SWIR sensor and IR lights, can that cut through fog too? Or is it the LWIR that is able to do that?
1. D4M: Dynamic Distributed Dimensional Data Model
http://www.mit.edu/~kepner/D4M/ GraphBLAS: http://graphblas.org
Achieving 100M database inserts per second using Apache Accumulo and D4M https://news.ycombinator.com/item?id=13465141
MIT D4M: Signal Processing on Databases [video] https://www.youtube.com/playlist?list=PLUl4u3cNGP62DPmPLrVyY...
2. Topological / Metric Space Model
Fast and Scalable Analysis of Massive Social Graphs http://www.cs.ucsb.edu/~ravenben/temp/rigel.pdf
Quantum Processes in Graph Computing - Marko Rodriguez [video] https://www.youtube.com/watch?v=qRoAInXxgtc
3. Propagator Model
Revised Report on the Propagator Model https://groups.csail.mit.edu/mac/users/gjs/propagators/
Constraints and Hallucinations: Filling in the Details - Gerry Sussman [video] https://www.youtube.com/watch?v=mwxknB4SgvM
We Really Don't Know How to Compute - Gerry Sussman [video] https://www.youtube.com/watch?v=O3tVctB_VSU
Propagators - Edward Kmett - Boston Haskell [video] https://www.youtube.com/watch?v=DyPzPeOPgUE
What are you working on?
I'm not sure I did justice to instant apps, because there's a language barrier playing in. But here's an example: I use the Amazon app maybe once every 2 weeks, and yet it's one of the apps consuming most amount of memory on my phone due to background services. After Amazon integrates instant apps, I'll be able to delete the app, and just google search for the product through my phone. The Google search will then download the required page as an app, giving me the experience of an app, whilst not even having it on the phone.
.Net Core: Finally, cross platform .Net. Deploying .Net services to Linux is a dream come true. Can't wait for the platform to stabilize.
Windows Server 2016: For "legacy" applications forced to stay on Windows, containers and Docker on Windows is a game changer. One step closer to hopefully making Windows servers somewhat manageable.
Regenerative medicine: understanding DNA code and restoring cells and organs, making eternal youth possible. It will take decades of hard work.
Ending cancer: We are studding virus mutations, so we could attack them without invasive techniques.
Nuclear fusion: We are simulating plasma physics. This is going to be enormous in ten years or so imho.
In UX an interesting trend is a flood of Software Tools which help during Design, evaluation, Research, etc.
Also adaptive UI which is changed due to user attributes and past behaviour seems to be trendy now (supported by the online marketing field with auto-optimizing Interfaces which optimize for conversion autonomously, etc.)
In programming in companies: realization that internal customers not having choice of internal IT providers hurts IT because it reduces IT's need to deliver valuable solutions effectively.
In leadership: management structure is a framework to enforce standardization and generally doesn't adapt well to change, even with the latest management silver bullets (lean, Agile, flat-orgs, etc)
Also in leadership: profound changes are occuring in society and geographies no longer define cultures.
In commercial writing: it's still early, and this takes time, but the concept of the "book" and how it's created is changing. Technologies that allow writers, editors, and beta readers to work on the manuscript simultaneously are increasing the velocity of change.
In art in general: someone else here mentioned music creation and payment is enabling entrants to sustain themselves in niche markets. This is happening in nearly all art forms, not just music. As electronic transfer fidelity increases, more art can be digitized, monetized. Look for more politicized, more global-reach art.
All these things stem from a greater understanding of the world and of human beings, starting with ourselves. It's important to realize each human being is a highly complex system and that generalizations about groups of humans are increasingly being challenged as scientifically unsound.
I helped with it for a little while, but the main developer was resistant to:
* Using a package manager or bundling dependencies into a compressed form. Dependencies had to be in the same git repo, fully extracted. (A bit of a "code smell")
* Dependencies could take months to get security updates.
* Documentation couldn't be in the git repo.
* Python 3 was "not an option"
Also:
* The main developer has limited experience with the torrent protocol.
It is an interesting project: but it is not a private or secure one.
So I can highly recommend the field of remote sensing as there are many interesting problems to solve.
Also, I'm really looking forward to the ActivityPub [2] implementation, that'll do a lot of interesting things for decentralized web.
[1] https://www.sidefx.com/community/houdini-16-launch-streaming...
Overall, I'm most excited about VR/AR/MR in relation to storytelling and education and how the two can be combined. Houdini and Houdini Engine for UE4 are definitely are worth a considering as part of your VR/AR development stack.
Analysis has always been an area that the tech community has lacked, ever since it was overdone back in the days of structured programming. It's really cool to bring back a bit of structured analysis as just another tool in the DevOps pipeline and join up the information with all the folks that need it.
Finding this tricky to parse, got a link or repo?
Another hot topic are organoid bodies and organs-on-a-chip. These are experimental systems where stem cells are turned to grow into structures similar to embryos or organs that allow the study of development and facilitate drug testing etc.
Thirdly, advances in sequencing made it possible to study what kind of bacteria live symbiotically within and on us. The composition of this so called microbiome seems to widely affect body and mind.
Finally, in my personal field, the simulation of how "simple" cells build complex structures and solve difficult tasks, the most exciting development is GPGPU :-)
For some details, see:
or
https://www.facebook.com/Engineering/videos/1015489039350720...
This is coupled with a move away from cookies[0].
0. https://adwords.googleblog.com/2017/01/making-youtube-better...
Because I am really sick of Google serving me ads for stuff I recently searched for and subsequently bought.
This will change real estate websites as well. I can just query for houses with X visual features
A fast compiled ruby-like programming language.
Metamaterials: Essentially a material engineered to have a unique property. By precisely controlling a materials structure you can influence how it interacts with electromagnetic waves, sound etc. You can create materials with unique properties such as a negative refractive index over certain wavelengths. It's kind of a novelty but people are building "cloaking devices" using metamaterials i.e. bending electromagnetic waves around a material in certain ways to make it appear invisible to certain frequencies.
Graphene (and other 2D materials): These materials are a relatively recent discovery, graphene was confirmed in 2004 and it has a number of interesting properties. In particular its electrical and thermal properties make it promising for a number of applications. I think it could possibly find applications in batteries, transistors and capacitors. At the moment it is a very expensive material to manufacture which makes it (currently) unsuited for commercial applications. There is a heap of active research involving graphene at the moment.
Google's Deep Mind put out some kind of cool stuff recentely [1], but I'm mostly just excited for anything that Ilker Yildirim [2] is doing with Joshua Tanenbaum, because it seems to triangulate more with how humans think about physics. When I was at CogSci 2016, Joshua mentioned combining this with analogical reasoning and that also sounded super cool, even though I'm not sure how to the two fit together.
[1] https://arxiv.org/abs/1612.00222 [2] http://www.mit.edu/~ilkery/
On the networking side of things, I'm excited about network virtualization and the potential that tools like Docker and Kubernetes give to virtualizing large and complex network topologies.
And as an employee of an IT-heavy enterprise, seeing DevOps becoming a thing makes me happy, even if adoption is slow and expectations are high. It's still better than waiting 6 months to get a couple of VMs to deploy my projects to...
2) More understanding of the "bio psycho social" model of mental illness, with better coordination across different agencies to prevent suicide.
In particular, wireless transmitters for roomscale are really exciting - seriously, I cannot wait to get rid of the wire-to-head era - as is roomscale for mobile devices.
The Vive getting additional trackers is also super-cool, as that will enable some much better forms of locomotion through foot-tracking. It'll take a little while to take off but I expect the Lighthouse tracking ecosystem to produce all kinds of cool things.
(Not all in VR, either. Drones plus Lighthouse, for example...)
Edit: there is a lot to be excited about these days
Long story short, so many processes I work with are done completely manually, which is a colossal waste of time. When I started, the person who previously did my job had about 7 main processes they completed monthly, which took about 60 hours to complete. Those 7 processes take my about 10 hours to do after I built automated workbooks
The sad thing is that these excel capabilities have been around forever, but no one understands them.
Lots of cool stuff in the space like Kubernetes, Swarm, CoreOS, rkt!!
In summary, the use of machine learning can help us develop better representations of chemical reactions, catalyst behavior, and we can now use adaptive learning to create closed-loop systems to identify, carry out, and optimize chemical processes to reduce environmental impact, reduce energy usage, and decrease costs.
The state of the art isn't quite there, but I see no major conceptual barriers left -- just a matter of implementing it.
Basically, it pits two networks in a "duel" and one of them is a generator network that learns to make images.
cross platform, open source, very fast
