2014-05-31 at

Darwin Award

Unknown email address:

x: So u go studying at US after SPM?

me: yup

x: How,u just took the SAT test?
x: How about the financial matter,u are self-sponsored?

me: Usapps.org ; good luck

x: Yea,I know USAPPS and have read almost all the things there.But I still hope that u answer me like are u self-sponsored?

me: What is the point of getting answers to new questions, when you have not yet fully acted on the answers you already have?

x: Come on guys,I just need some advice.

me: Do you home work, fool. The advice has been on each college's admissions office page (along with respective contact details) for time immemorable.

x: How much do u spend per year in US?

me: This is now going ot my spam folder.

Data Modelling

Quick one - I was looking for a free/open-source tool to quickly render JS objects as flow-charts. Honed in on these.

http://www.jointjs.com
http://jsplumbtoolkit.com
http://arborjs.org/atlas

Not yet sure if this will do, but they seem like good visualisation tools for doing solutions architecture. (Working on a robot personality model myself... picking tools before I get started.)

This won on simplicity: http://adrai.github.io/flowchart.js/
I'm a technical guy. I'm looking for more technical co-founders.

Objective: work on an open-sourced solution to permanently break the Turing test (some resources of interest: http://en.wikipedia.org/wiki/List_of_artificial_intelligence_projects)

Possible progression:
- (6 months) chit-chat and work on a GitHub repository in your free time
- Q1-2015 : apply for Y-Combinator 2015, or other funding.

Not going to ask for anything else, but you're free to offer.

Saturday, nibbling away at administrative duties at the office. 4pm. Time for office social.

That leaves a mountain of paper work and a few dozen internship resumes for tomorrow. :)

Hopefully I can squeeze in some time for technical studies this evening.
x: Hi.

me: Hello!
Can I help you?>

x: hi.. was responding to your artificial intelligence
about an hour ago

me: Cool. Are you a programmer?

x: ya
u?

me:Yes. What sort of related work have you done?

x: I build mobile phones

me: what about u?
Do you do personality modelling while building mobile phones?
I've studied some of the information architecture of brains, and also looked at the WordNet project.

x: ok.. so how do u monetise
what is your qualifications btw?
have u built anyyhing in the market that sells?
any product already in the market?

me: forget monetisation. if you're more interested in monetisation than in making humans redundant, I think you've got the wrong project.

x: I have to make sure I dont talk to an idiot

me: same here.

x: I have completed a dew projects
have 11 usa patents
published a book
if u have not a single patent
forget it
u probably an idiot that use big words

me: ok. bye

x: ok.. bye

me:Nice resume you have, by the way. Keep it up.

x: thanks
Cannot brain some of these people. I want to cry.



Someone asks what I mean by breaking the TT.
I think the moonshot (which is near enough that it's not a Mars shot) is to basically get past milestones where you can't tell a human from a bot, say first in a customer service phone call, then in a sales phone call, then in a new product development discussion on instant messenger, etc. etc. There's no limit to the number of tests that are harder and harder - it's only permanently broken when it's trivial to try and come up with newer arbitrarily hard tests. Then the Turing test is "permanently" broken.

2014-05-30 at

Katas

Optimise and reiterate.

And do absolutely boring stuff like hang out with old friends. So off to Bangsar I go...

Heading home. Bring on the C++ for Friday night... What a life...
(hmm... what happens if i do this...)

Serious question.

Anyone looking for a TECHNICAL CO-FOUNDER? Post your idea / product / company below.

(evil laugh)

This is followed on the forum by banter about perpetual motion machines, a startup-cofounder job board, and the archetypical co-founder ads there.

I think it's high time someone pitched a sentient machine to Ycombinator. 3 months and no distractions. Hmm. Since last weekend's hackathon I've been thinking to upload some open-sourced architecture proposals... just for discussion's sake, of course.
Human expectation is depressing.

Must find business case for startup to end it.

2014-05-29 at

#yesAllWomen

Struggling to find ways to use #yesAllMen tastefully, in a joke. Seriously - because I care about weird shit like that.
Eh? Sorry, I can't tell the cryers from the strong. Live in fear of being hurt all the time. #YesAllMen ehehe.



Meanwhile I'm just reminded that the issue is:
- I can agree on opposition to violence
- I can't seem to agree on making gender stereotypes an issue... for example, let's assume it's fact that men are responsible for most of the violence committed on earth; so now explain to me, what the fuck that has to do with me if I'm not a perpetrator of said violence... regardless of my gender
- QED: #yesAllWomen can go fuck themselves, it doesn't make me angry, in fact i'll be over here following my own advice ;P

Fear

If you're not living in fear, you're not doing it right. #forSparta

2014-05-27 at

I disagree with this. Emotions are simple. It is people who can't manage them, who are stupid. Maybe this remains an area to be proven.

But by the time I find time for it, they'll have resolved it too... at the rate I typically get around to these things.

---

My take on this ... [30,000 feet perspective, mind you ;)]

Most of the problems with AI to-date have been bad data-structures. For example, trying to get a machine to converse in common English without giving it a model of human emotion and motivation, and a model of social mores, is plain stupid... the dumbest approaches use pure text as the raw input data - that's like saying that a machine should be expected to learn how to write PHP by reading PHP scripts that work, and having no other knowledge of computers.

"Deep learning" is on the right track, in this respect, but it's not a new approach (see Princeton's WordNet, for example). [Expect Deep Learning to soon be the next Big Data in terms of business fads - go register a few domains now... ;);) ] But the comprehensive model is really about having a complete model of human sensation. emotion, and cognition... usually AI nerds are over-focused on the cognition without emotion, social consciousness, etc. So sure fail lah.

I already know when I'm getting bored of this problem:

1. they get to the point where passing the Turing test with open-sourced implementations is trivial

2. the machines are surpassing humans (as often as humans are surpassing other humans) in the fields of literature, art, humanities, mathematics, and physics - the whole formation and discover of new concepts as well, not just reiterating ones written by humans

3. at which point i'll probably definitely have just about enough energy to go run a coffee shop ;)

---

I don't think I have much to offer besides time for money - once you're working for someone else's research project, you're stuck with their mandates. Which is why I treat this as a hobby, and am happy to find a living elsewhere...

by the time I've caught up on my C++, Haskell, Erlang, stochastic calculus, and linear algebra... I hope to be able to give employers a run for their money, working in the field besides fresh graduates ;P

---

You can, with sufficiently "deep" learning, come to a model of emotions and human mores. In fact, people who navigate through life without actually feeling the emotions are called sociopaths. :P

Yes, I'd be happy with sociopathic computers.... it's better than these silly boxes we have around today ;) It's progress.

Another Twitter Account I Will Not Register

Selamat pagi Malaysia. Dengan segala hormatnya and diingati supaya mengingatkan langkah ini: sesuaikan diri anda, sebelum anda disesuaikan. Sekian. Yang menjalankan tugas. Ketua Jabatan Penyesuaian Malaysia.

And that's all I could come with on the jam to work today.

Job Data Protocol

5 years on; I conclude that we need an open standard for communicating data between "job seekers", "employerers", and "brokers".

Looks like nothing's changed.

Big Med, Little Med

Medical strategist asks for an app idea. Aiyo. Do lah:
Do a mobile/web app for finding a doctor who is free to see you right now.

Revenue is commissions from doctors.

Great for small clinic owners running under capacity.

Growth potential is simply... at some point, there'll be enough information liquidity to eat into the margins of higher end providers. Which makes this a diversification play for anyone who already owns big med stock

Long game is still the "tricorder problem."
Not that hard what :P

2014-05-25 at

Plebeia

Doing the laundry. Wiping down my computing gear. Sitting on my bed and taking into account all the wonderful plebeian activities I have engaged in for the last quarter, and in the last year, in an effort to make this the decade of "trying to fit in".

:)

If I believe I can, I'm sure I can...