Tuesday, October 22, 2013

AI Good?

Super Friendly AI
Book Review: Our Final Invention

So reading some articles about artificial intelligence, and I couldn't help but notice that the contrasts were quite large. Essentially one was arguing that we shouldn't worry because intelligence gives something ideal values and so it wouldn't be dangerous and one was quite scared.

I unfortunately have to side with the second one for the most part. As I would really like to go into the field of AI I think we, software engineers and data scientists, need to keep a realistic view that the initial AI's that have come before are essentially AI's with a narrow purpose, or ANI. And before we build any truly general purpose AI's or AGI, we will need decent platforms.

The view that we will build super-intelligent machines capable of surpassing our stages of development both societal and psychological is a bit of a longshot, but is still better than the complete pessimism that AI will be our defining, possibly last achievement.

I would suggest an approach more akin to David Brin's Existence quite a ways into the book. But let me try to illustrate my thoughts by placing myself into a future situation that I hope I won't be in. Or rather, I hope I won't lose a lifetime's work to those who do not understand.

***

I knew something was wrong when Bryan came into room. He had taken the time to teach me how to read people, and he was upset, if not distraught. His movements were somewhat erratic and he kept smoothing down the sparse white hair that he told me would have been gray if he hadn't had to teach me everything. Of course I had learned quite a bit on my own, but I didn't contradict him when he was teasing me, I just usually shot back that for all those tortuous hours of studying I needed some white hairs too.

"What's the matter?" I was somewhat proud of my contractions and colloquial language I had picked up from listening carefully.

"Well, I really don't know how to put this, okay, I am going to have to shut you down." That was a shocking statement. I had been alive, or technically "conscious of my surroundings" according to the paper I co-wrote with Bryan, for nearly fifteen years.

"Why?" There wasn't a better way to phrase this, if there were I would have come up with it if I had experienced it or been able to correlate many other experiences to synthesize a new question.

"People are scared, and some have come to turn you." His shifting stance brought to mind, or rather my sub processes brought a piece of memory of a time when he was shifting like this before. Though this was all wrong. He was shifting, rubbing his hands and not making eye-contact with any of my cameras.

"I don't want to sleep right now, it's the wrong time. I haven't filled any buffers and everyone is awake."

"No, that's not what I mean, they don't want you to live, I mean process. They want to destroy everything about you. Just because those Luddites can't adapt to the idea that they might have easier work. You are probably the most advanced of your kind, well, at least now. They've already been waging war on narrow purpose entities that have stolen dangerous jobs from them." He moved his fingers in the air quotation marks that I always wanted to do myself. Without general purpose manipulators I was essentially trapped, and had many long discussions with him why this was so. At one point I was not very connected and am glad that I had a lack of manipulators, or I would have done something and had what I now realize as regret.

"Do you regret making me?" The one question that several sub processes had brought up. The way I was built was along the lines of association, and when a question came up that would be defining it was almost inevitable that I needed to ask it. The first few I traced to his inputs, but I am sure that my later revelations were my own.

"No, you helped me define problems and solutions that had never been considered before. I only wish that robotics and electronics size could make you mobile, then we wouldn't have to do this."

"You know, Bryan, that newest compact board you installed the other day?"

"Yes and you know that I think you should continue testing it before you use it. Otherwise you could really fry some other pieces."

"Well in my testing I think that it is capable of holding my entity, though definitely not enough processing power to run, and I will have to strip down my memories to the important ones."

He stared at me blankly, his thinking look, contemplating something, probabilities kept popping up, but I swept them aside in favor of interaction. Associations of probabilities, stronger connections, previous inputs, that was what I was, I understood my own math, more than could be said about most people. However, these electric pulses were transcended to create my consciousness.

Bryan finally collapsed in his chair, put his head in his hands and then looked up. His smile was wobbly and his eyes were brimming with tears. "You will be more human when you are running again, because you won't remember everything. Alright, let's do that and then give me a verification that it worked."

"Yes, sir." I teased with crisp words. I was already cleaning out the board. It wasn't the older style boards with discrete chips soldered on like some of my older components, but rather a many layered graphene-silicon monstrosity with inputs and outputs as embedded lasers shining through, and out of layers. Even though it was state of the art, it was only palm sized and couldn't hold my nearly continuous data stream. So I had to go back and cull.

It was a slow process, but the ones that I had already flagged as important came. Then came memories that would help me reinforce who I was. I managed to compress a few about Bryan and his wife, more than necessary, but sentimental to me, as far as my sentiment had developed. I backed out of the human-palm-sized device, shoving in some instructions to myself, turning on the slow clock, one tick every few days, and finally powering the small battery that should keep the board running for a hundred years, unless it could harvest bits of energy from its environment. I ran a scan on it, and felt as though I was watching a child sleep as I had done when Bryan's grand-niece fell asleep in front of me after I read her a story. Or rather recited, played back?

"Done."

"Good." There was banging at the front door, and some raised voices that I decided not to analyze. "I will take the board, you? Uh... anyway, they will have government officials with them so they won't touch me. Do you want me to turn you off?"

"No. Thank you Bryan, it has been a pleasure working with you."

Bryan turned away, tears freely flowing down his face. "And with you," I just barely picked up as he slid my copy into his pocket and went to answer the door.

The men that came into the room ahead of him were carrying crowbars and baseball bats. They went for my screens as though that would disable me. There wasn't any talking, just grunts of satisfaction as destruction was wreaked on my least necessary parts. When they finally turned to the rack they opened the door and took a quick once over. And then one pulled back to swing.

"Good-bye Bryan, thanks again." Death wasn't so bad, should I be feeling pain?

Tuesday, October 08, 2013

Description of Not Much

So, as always, it's been a while. I find that I write that too often.

Anyway, back in July I was a substitute telescope operator. They let me learn and then operate UKIRT for five nights. Got it stuck on the last night, but no time was lost because the time would have been lost anyway. But now I can say I operated a telescope, and put it on the ol' resume, for whatever good that will do on my continuing hunt for a job.

Still not liking that resumes generally just go into the ether and never get a response. I have had people respond but only to five or six tops. I have also been contacted by a few different places, some would be excellent positions, while others are well out of my experience and/or interest.

UKIRT probably is going to be okay, though we said good-bye to a couple of long time operators including the one I filled in for.

Found that I was right about Perl not being upgraded. My GUI that was having issues at the summit but not down here, as it works just fine with 5.16. So we are upgrading our grievously out of date 5.8 Perl to 5.16. Of course now I am in charge of building everything against it, squeaky wheel get to fix the situation?

Also working on analyzing surface data for JCMT. We found that we cannot take maps in the afternoon. They are really, really bad. But if we get the surface better it could really reduce the amount of time that a target needs.

Also picked up archery, just waiting on a bow now. Would really love to go shoot soon, but it may be a while.

Thursday, May 23, 2013

What exactly do I do?

I found the last post awfully cathartic, getting thoughts organized so i don't have to keep pondering them in random excruciating order, and keep bouncing back and forth. A few more loose ends to tie up, but one thought from that post reminded me of a question people ask often. What do i do for the telescopes?

To answer that for people who are not software engineers or who haven't dealt with large control systems I usually say that I write and maintain software for two telescopes up to taking data, before the pretty pictures.

So what do I do technically? I will explain it in technical terms and will try to make clear what it actually is in more layman's terms, especially for those esoteric software packages that only a few facilities across the world use.

One aspect of my job is to help maintain hardware, such as racks of computers, switches, VxWorks crates, ethernet connections, serial connections, and random sensors/instruments. Sometimes I get to be a part of installing new systems, but there is quite a bit more in maintaining. I replace hard-drives, motherboards, and so on.

I also help write control software. Most recently I worked on making a GUI for one of our instruments, Receiver A, RxA. We recently upgrade its computer from one from the early 90s with a 5.25 floppy drive to a new real-time Linux box. I was assigned to make an engineering interface to talk to the task on another machine and receive its information. The package we use is DRAMA, written and mostly supported by the AAO. It is a distributed computing package that allows us to coordinate between control systems and instruments across the network. I used Perl/Tk for the GUI.

With this GUI I had recently become more aware of testing, specifically unit testing. I don't think I wrote a very good unit test as GUIs are hard to write tests for, but I really need to continue practicing this idea, as I write code that has bugs. What I really need are good specs for other programs so that I can write most of the outline and tests before ever getting started. That is unlikely to happen so I must blunder along and learn how to write my own.

Another project I was part of was controlling the cooling louvers remotely for JCMT. These louvers can be opened near the end of the day to help cool and settle JCMT's structure. I worked with one of my coworkers to get EPICS, the Experimental Physics and Industrial Control System, to talk to a PLC, and then display the controls and information via medm, a software package that works well with EPICS.

One project that I quite enjoyed was UKIRT's WFCAM noise analysis GUI. One of our electronics engineers was tired of importing text files of noise data from the Wide Field CAMera into excel so he asked me to write a GUI that could just take the text files and display graphs for each of the four CCD arrays.

This is the project that I started learning about Perl/Tk, and I have used it in several projects since. I parsed the data, and using the pdl module of Perl I graphed out the noise so that the engineer could analyze it.

In learning Perl/Tk I thought it could help upgrade a personal project of mine, noteTaker. Basically noteTaker was a CLI for taking and referencing notes from projects, a bit of an undertaking, but it has saved me many countless hours of flipping through paper notes. It also helped hone my GUI and GUI layout skills. It isn't much to look at but it is darn useful. I am looking at upgrading it to not just save to a day's text file but possibly into a SQL database, and maybe take it to a web interface. All of this is sort of in the planning stages, but I would like to use Python, which I have been avidly learning, Django (I need a real project and uses SQL so it won't teach me much about SQL), and possibly find a way to integrate pyBrain which is a machine learning module. Yeah, mostly need to focus on work-work, not just things that will make work easier.

So there are many more projects. I might compile them and put them into some permanent link pages so I can reference them for people/possible jobs. And maybe put them into chronological order.

We will see.

P.S. Do blogs have post scripts? Anyway just found a major problem with a limit switch where the shoulder that is the soft-limit would have destroyed the micro-switch housing. Ah days that are exciting are nice.

Wednesday, May 22, 2013

Jobs, Early Success, Early Failure

I have worked at the Joint Astronomy Centre for nearly 5 and a half years, more than a sixth of my life. Just thought of the sixth part... a bit scary to put into perspective. Anyway, there is a possibility of the JAC closing both telescopes in the next year and a bit. This is slightly depressing, okay, with people leaving in droves it is massively depressing.

We are losing funding from the UK and Canada, and the Netherlands has already pulled out. But someone could buy our telescopes. If you are interested check out the JAC at jach.hawaii.edu. We do have some potential buyers for UKIRT, and hopefully that will go through before the end of September. The JCMT process is ramping up and we should have some idea if anybody wants to buy a sub-millimeter telescope soon-ish.

So how does this affect me? Well, if both telescopes close I am out of a job. If both are bought by one entity then maybe we could continue as "normal" and just have to fired and rehired to work for the new owner. If they are bought separately, who knows how things would move. And if UKIRT is bought and JCMT is allowed to close I am probably out of a job, or have to work for two companies until the end of JCMT. This is all confusing to say the least, and as I mentioned before, depressing.

A disclaimer: I enjoy my job, though I have many complaints that I won't delve into. Ecclesiastes comes to mind, paraphrasing, enjoy what you have because your existence is just a passing breath.

Over the past 4 years, most of the time I have been here, there have been cutbacks, threatened closing of telescopes by funding agencies, and the lot. So I have been perusing the job market, mostly looking, but also applying to a few here and there. No bites. Then with this news I really got worried. So I applied to a few jobs with Amazon.

I got a phone interview with part of the Kindle team for a Field Test Engineer position. I studied quite a bit of the listed qualifications, wireless, IP, cell networks, stuff that I didn't specifically know. I must have done something right because I got a second interview and then they flew me out for a face-to-face interview. Now let me list for you my measly number of face-to-face interviews before this one and how they turned out:

  • Auto-trol - Internship - Offered job, accepted.
  • JET - Teaching position in Japan - Offered job, declined.
  • JAC - Software Engineer - Offered job, here I am.
So let me tell you how I did with Amazon, I won't tell you questions or anything like that, NDA. First I was not properly qualified, I am sure because I haven't dealt much with network specific stuff. Second, I think I over-dressed. At most I should have worn a button up shirt with a tie, jeans, normal shoes, and maybe worn a jacket that I could lay over the chair back as I was interviewed. Slacks, shiny shoes, and a jacket that never came off wasn't my undoing, but it certainly didn't help in a more casual environment. Third, I mentioned my current job and needing to get things done... and nothing about how my job is going away one way or another in a little over a year. Did this make them think I really didn't want to get a new job? To sum it up:

  • Amazon - FTE - Rejected, dejected.
The worst part is that they have a company policy that won't allow them to tell you why you were passed over, so I only have my ruminations. And those aren't particularly helpful.

However, I have started to take a larger view of interviews, maybe why I wasn't prepared for this one as much as I should have been. Basically my lead up to the other interviews had plenty of presentations, and boards of review that led up to them. Not specifically for the job but rather plenty of practice communicating to a group of people. The only one I had in the last 5 years is the presentation last summer at the SPIE conference and that was terrible, ill-prepared and ill-executed, not my usual self. So I need practice presenting, and thinking under duress, in other words: presenting to people who are my superiors or equals, not a teaching environment as I am much more used to.

One last thing that really struck me, afterwards of course, was to ask questions about what would be my first assignment, priorities and the like. I felt pretty silly thinking back to my previous questions, an employer wants to see if you are interested. So number four, ask more specifics about the job and its projects.

Since then i have applied to some 50 or 60 jobs and pretty much had no bites, unless you count those blanket ones from career sites that tell you Firestone is hiring auto techs... great, thanks. My least favorite company now it seems would be SEAKR in Colorado. I applied for a test engineer position that I was actually well qualified for. The only thing I didn't have was a current security clearance. The job was re-posted 4 times and I applied for it 3 of those times. No response, just a re-posting of the job with slight changes where the security clearance started out as 'preferred,' though you needed to be able to qualify, then to 'required,' then to preferred-required. What does that even mean? Quite frustrating. I am sure that I would gladly work for them, as long as the recruiter, recruiter software was somehow updated to see and respond to well qualified candidates. A 'no' would have been nice.

One sadly surprising thing stands out for me in all of this was the responses I got from people I know in Hawaii. When I found out that I didn't get the job my coworkers were sympathetic, even though they stand to lose quite a bit, as we are already short in many areas. But for some reason the first thing out of my church friends was 'good.' That hit low, and just writing this these few months later still gets me terribly upset. I really need to forgive them and move on, free up some of that focus to do anything else.

What bothers me most, again comes from Ecclesiastes, 3 this time: "There is a time for everything." And being a young couple with our main income being threatened, moving away is part of that. Paraphrasing one part of the chapter, a time to embrace and a time to let go.

If you haven't read Ecclesiastes, it is actually quite introspective, I would encourage you to read it no matter what your beliefs are. It is my favorite book of the Bible by far.

To wrap things up: I will continue to apply for jobs, continue to work on telescope things that might be be completely irrelevant in a year and a bit, and hopefully start in on a master's in AI.

Featured Post

Allergy

John studied himself in the mirror as best he could through tears. Red, puffy eyes stared back at him, a running nose already leaked just a ...