HNHacker News
TopNewBestAskShowJobs

romaniv

2,750 karma · joined February 2, 2011

submissionscomments
romaniv··on Going Serverless: From Common Lisp and CGI to AWS Lambda and API Gateway
I'm not proposing anything. I'm saying that training should be factored in when considering development costs. The notion that training is always a one-time investment that can be pretty much ignored if you plan to stick with some tech long-term is a fallacy.
romaniv··on Going Serverless: From Common Lisp and CGI to AWS Lambda and API Gateway
Training also takes time.
romaniv··on AWS icon quiz
It could have been AuroraDB, DynamoDB, NeptuneDB, RDBS, and RedshiftDB. At least people would be able to tell that all of those are databases.
romaniv··on AWS icon quiz
>It has terrible naming too: a lot of services are acronyms, it makes them difficult to understand for outsiders.

I almost think this was a deliberate choice. Makes the services sound more important than what they are. At any rate, many engineers I know love to smugly use this acronym soup.

romaniv··on Freeing the Web from the Browser
The simplest counter-example is Wikipedia. Most links on any page could lead to generic term definitions, or they could link to explanations of how those terms work withing the context of the page.

I.e. a link to "synthesizers" on a page about FM synthesis could lead to a generic article on synthesizers, or to a list of FM synthesizers released up to date.

And that's just the most obvious example. Having different "linking contexts" would allow to add more links without turning the original document into a mess.

Again, using Wikipedia as an example, you could add another "context" to the page by linking various paragraphs to citations. That would be much more user-friendly than what they do right now with bracketed numbers.

romaniv··on The last thing libraries need is Silicon Valley “disruption.”
Some users of this website might never go to libraries, but every time I go to our local library there are tons of people there.
romaniv··on The last thing libraries need is Silicon Valley “disruption.”
This is a very important point. Libraries do more than just lend books. They are institutions that preserve and promote knowledge and have certain values "built in". They also often serve as local community centers of sorts. Having cheap books available in Kindle store does not fully replace functionality of a library.
romaniv··on Ernst Dickmanns, a German scientist who developed self-driving cars in the 1980s
Even chess isn't tractable in terms of pure brute force search. And yet a computer won against arguably the best human player in 1996. It was mostly a PR stunt by IBM.

We had 20 years of doubling of computing power before Lee Sedol match. In those 20 years there were many other AI programs that have beaten various world champions at other board games (and no one cared). There were other good Go engines before Alpha Go. They would beat most human players in the world.

Why AlphaGo of all other programs is seen not as increment, but as some giant leap forward? It doesn't solve a new class of problems and it doesn't use any fundamentally new algorithms.

romaniv··on Ernst Dickmanns, a German scientist who developed self-driving cars in the 1980s
In 1996 IBM made a system that was better than any person at playing chess. Chess has roughly 400 moves possible at any point.

In 2016 Google made a system that was better at any person at playing Go. Go has roughly 130,000 moves possible at any point. That's approximately equal to 400 * 2^9.

Moore's law states that the number of transistors (which is a proxy for computing power) in CPUs doubles every year. 2016 - 1996 = 20 years of computing power growth.

In short, after our computing power has increased roughly by a factor of 2 ^ 20 someone made a system that plays a game that's 2 ^ 9 time more complicated than chess. Why is this seen as some giant, surprising leap forward?

romaniv··on Ernst Dickmanns, a German scientist who developed self-driving cars in the 1980s
What about AlphaGo Zero? I'm honestly not sure why a system that plays board games is seen as some kind of giant leap in AI. Seems mostly a result of Google's marketing.
romaniv··on Ernst Dickmanns, a German scientist who developed self-driving cars in the 1980s
Expert systems had industry funding in the 80s. And contrary to what some people here keep implying, they delivered results. Some of them are still used right now.
romaniv··on Ernst Dickmanns, a German scientist who developed self-driving cars in the 1980s
Another way of putting it: AI research of the 60s-80s was extremely impressive and is ridiculously underappreciated right now. Many ideas were simply way ahead of their time. It's sad that most software engineers today do not bother learning anything about the past of computing, assuming everything made in those decades is automatically "outdated". (This is not restricted to AI. For example, most OO programmers don't know what is Smalltalk.)

The guy the article about says as much.

“I’ve stopped giving general advice to other researchers,” said Dickmanns, now 82 years old. “Only this much: One should never completely lose sight of approaches that were once very successful.”

I really feel for those researchers. Seeing how your life's work is ignored by people who brute-force their ways through the problems you've already solved must be rather depressing.

romaniv··on The Mythos of Model Interpretability in Machine Learning
> If there is a very low-dimensional manifold in parameter space that accounts for all the predictive power, that would be a very easy thing to diagnose from a neural network model.

People wouldn't waste time researching neural network visualization tools, coming up with adversarial examples and developing alternative (more explainable) machine learning models if that was the case.

romaniv··on The Mythos of Model Interpretability in Machine Learning
>What do you mean by “gravitational theory” apart from “correctly accounts for predicting the gravitational effects that we observe”?

F = G (m1 * m2) / r^2

Newton's theory of gravitation identified the relevant variables and their relationships. It had no junk variables or parameters. You can do algebraic transformations on it and use the formula outside of the original context. It was not fully correct, but it made exact predictions. People were able to notice the discrepancies it had with orbits of some celestial bodies, which helped in identifying general theory of relativity.

All of this would have been lost of the formula was a state vector of a large neural network.

romaniv··on The Mythos of Model Interpretability in Machine Learning
>I’m saying the development of a model that yields scientific progress doesn’t have to have a connection to the parent comment’s proposed definition of interpretability

Let's say someone used a neural network to predict planetary positions as a function of time, mass and so on. They made a network with several hundred parameters, trained until it stopped converging and published their final state vector.

Would that ever lead to gravitational theory?

romaniv··on The Mythos of Model Interpretability in Machine Learning
>But then in a practical setting, the model doesn’t correspond to something physical or to the inference goal of a practitioner.

How does this invalidate the notion of interpretability?

A model can be both wrong/misapplied and fully interpretable.

romaniv··on NLP's ImageNet Moment: From Shallow to Deep Pre-Training
>It is now possible to grab a pretrained model and start producing state-of-the-art NLP results in a wide range of tasks with relatively little effort.

Are there any applications/websites where this can be seen in action? It's increasingly hard to judge how good state-of-the-art really is from research papers.

romaniv··on Fuzzy Logic in Agent-Based Game Design [pdf]
Complex system simulations (look up Net Logo). Distributed, fault-tolerant data processing systems. The latter is harder to back up, because all real-life "agent" systems I've worked with are rough approximations of "agents" made with commonplace technologies and don't use the concept to its fullest extent.
romaniv··on Fuzzy Logic in Agent-Based Game Design [pdf]
I think that's a different kind of "agent".

http://alumni.media.mit.edu/~mt/thesis/mt-thesis.html

romaniv··on Bias detectives: the researchers striving to make algorithms fair
The problem here is that modern machine learning is rarely more than a computerized way of building statistical models based on observed data. And the bias mitigation techniques I've read about all revolve around (indirectly) manipulating those models based on what results they produce. This can lead to various problems in the long run.
romaniv··on Fuzzy Logic in Agent-Based Game Design [pdf]
I wish agent-based programming was more popular today. It's an extremely powerful way of thinking about complex problems and distributed solutions. And it has much more actual theory behind it than things like SOA (which is mostly a re-branding of good OOP practices with a lot of buzzwords added in).
romaniv··on Chatbots were the next big thing: what happened?
That's a great way of putting it. Also, GUIs attempt to use human natural reasoning (about objects, space, movement, etc.) and have some kinesthetic components (the mouse, pointing, dragging).
romaniv··on Chatbots were the next big thing: what happened?
It's interesting to think about. In some ways GUIs are more primitive than text interfaces. The scathing characterization of GUI as "caveman's point and grunt interface" isn't entirely wrong. But at the same time, chat bots are spoken word. Text is a later invention, which was created for a reason. It has much higher information density, allows you to look at multiple tings at the same time and skip irrelevant detail, etc.
romaniv··on Ask HN: Are you working on interesting technical problems?
I think answers here will be skewed, since people who work on trivial programming tasks often don't visit websites like this.

I've worked on moderately interesting problems in some positions, and I've also worked in positions where everyone with experience was sucked into fixing infinite streams of fuck-ups from junior developers and contractors, who end up doing all the actual development.

IMO, it mostly boils down to control. If you have control and you are knowledgeable, you can solve "boring" problems in interesting and fulfilling ways. If you don't have control over how you do your job, programming can quickly turn into glorified data entry.

romaniv··on Microsoft acquires Github
>I judge companies on current behavior and Microsoft seems to be doing far more right than wrong.

This isn't about behavior. This is about product management. As a reminder, Microsoft's current products include TFS (and I don't mean just TFSVC) and Teams.

romaniv··on The US military is funding an effort to catch deepfakes and other AI trickery
Right now the visuals for these fakes seem to fall apart at higher resolutions or when there is any significant head rotation. This is actually better than what we have in the realm of images, where almost anyone can create a "realistic" Photoshop of the wast majority of scenes.
romaniv··on The US military is funding an effort to catch deepfakes and other AI trickery
This doesn't automatically mean GANs can outpace any detection method.
romaniv··on Big tech products are mostly useless
> automated system integration is the combination of existing technologies.

Which ones? As far as I can see, it's all manual right now. Service integration is probably the largest area of IT at the moment and it sucks in more and more engineers every year. So maybe the notion of "web services" and "APIs" simply aren't the right ways to go about it.

romaniv··on Big tech products are mostly useless
> I do think though that AI is the last "fundamental" change information technologies will bring though.

There are wast continents of mostly unexplored or abandoned territory in IT. Moreover, some of them are well documented. Some of them were documented in 1970s. The fact that IT professionals today don't know about them and can't imagine anything more significant than "AI" for pizza ordering, is unsettling. Just to name a few things Alan Kay often talks about:

- Automated system integration or tools for generalized interconnection of applications available to end users.

- WYSIWYG for the Web.

- Constraint-based problem-solvers accessible to mere mortals.

- Agent oriented programming with UI simple enough for normal people to use.

- Dynamic simulations in "normal" software (not CADs)

I am absolutely sickened by the fact that the most common response to this list is "nobody needs this". As far as I'm concerned, each of those things is needed orders of magnitude more than a haircut-scheduling AI.

romaniv··on Show HN: Whiteboardfree – Developer Jobs at Companies That Don't Whiteboard
>I think you have got the wrong lesson off those guys. They are about engaging the spatio-visual systems. The whiteboard is the first step away from text towards that.

Not when you are literally writing code on it.

← PreviousPage 5 of 30Next →