HNHacker News
TopNewBestAskShowJobs

jeffmcmahan

388 karma · joined August 1, 2016

[ my public key: https://keybase.io/jeffmcmahan; my proof: https://keybase.io/jeffmcmahan/sigs/LkRxkjFGiURCcGAZhebTYaCjrrZ6HbydIvnH9FJA2HE ]
submissionscomments
jeffmcmahan··on Training an AI to convert design mockups into HTML and CSS
Yes, this is typical AI weenie optimism. "We can handle many of the easiest cases, so in X years we will be able to do all the really hard things that we haven't even thought of yet!" No.
jeffmcmahan··on The Good Place: how a sitcom made philosophy seem cool
I like the show. With that said, I have a bachelor's degree in philosophy, and the show (i.e., the character Chidi) mentions philosophers and concisely recounts their views in broad strokes, but I can't recall an instance where it invites the audience to understand an argument as such; more like, "Kant thought X, so maybe we should X."
jeffmcmahan··on Show HN: Jazzle – An Innovative, Fast Transpiler for ECMAScript 2015 and later
How does it compare to buble in terms of speed? (And yes, I realize buble does not support a variety of es2016 features).
jeffmcmahan··on Joplin – A note-taking and to-do app with builds for desktop, mobile, terminal
Good God - it's a note taking app that installs sqlite, sharp, and more than 300 other things... well into the millions of lines of code.
jeffmcmahan··on You Will Lose Your Job to a Robot and Sooner Than You Think
This is a tech-ignorant political blogger with a serious case of AI weenie-ism.
jeffmcmahan··on JSLinux – Run Linux or other Operating Systems in your browser
It hung at the boot screen for me, which really took me back.
jeffmcmahan··on The merging of machine capability and human consciousness is already happening
And all God's people said: amen.
jeffmcmahan··on The merging of machine capability and human consciousness is already happening
"The distinction between us (humans) and them (the machines) will become almost imperceptible."

... Absolutely nothing in the text substantiates any of the very strong claims.

jeffmcmahan··on Montreal Is Leading the AI World Takeover
"AI is now the technology shaping our world more than ever before."

This, the first sentence of the article, does not express a thought. Read it carefully. It's on a par with "More people have been to Russia than I have."

jeffmcmahan··on Ask HN: What are some good podcasts?
Radiotopia would be the place to begin. Just give each show a try, and you've sampled the best available.
jeffmcmahan··on The Disadvantages of an Elite Education (2008)
I have a bachelor's degree in philosophy and linguistics from an elite university. My parents owned and auto parts store in a small town, which was next door to my grandparents' bakery. Based on my experience, it seems this guys' problem is a boring backstory; no university can or should make up for the fact that you don't regularly mix with non-university types.
jeffmcmahan··on “BRAAAM”: The Sound That Invaded the Hollywood Soundtrack
BRAAAM is first heard in Star Wars Episode II: Attack of the Clones (8 years before Inception): https://youtu.be/3ME5jhsgmB4?t=53 (two in a row spot-on BRAAMs)
jeffmcmahan··on Maciej Ceglowski – Superintelligence: The Idea That Eats Smart People
His arguments are not tight, but the talk is cathartic for those of us who see "AI weenies" for what they are.
jeffmcmahan··on Error Handling in Node.js
If you're going to check every argument's type and throw on failure, either use a statically typed language or adopt a concise way of type checking. Many of the examples have big groups of assert() calls at the top. Gross.
jeffmcmahan··on AI is always less impressive than is claimed
>> I guess "the" is like "distinct" in sql

That's original Russellian theory, in a nutshell (http://www.uvm.edu/~lderosse/courses/lang/Russell(1905).pdf). None of its many variants cover the data.

>> It was a joke

Fair enough!

jeffmcmahan··on AI is always less impressive than is claimed
>> it's amazing they can do language at all. But research on that is just starting. Give it time.

Not so. Academic work in natural language processing has been very active for decades. Check out NL SOAR, which is an awesome attempt from the early 90s by top CS/AI scholars: http://www.cs.cmu.edu/afs/cs/project/soar/utc/nl/doc/nl-soar...

In general, language processing has been an important motivator and funding source for AI work. Remember that US Navy Signals Intelligence funded Noam Chomsky's early work on syntax. When early advances in basic linguistic research alongside early advances in computer hardware made machine translation seem plausible, money poured in. And it's pouring in again (now from private industry too).

jeffmcmahan··on AI is always less impressive than is claimed
There's a clear difference between "bring me an apple" and "bring me the apple". The latter requires that there is either only one apple available, or requires some special contextual knowledge to comply with.
jeffmcmahan··on AI is always less impressive than is claimed
It's back now!
jeffmcmahan··on AI is always less impressive than is claimed
Here's the kind of claim I mean to attack in the polemic:

"The algorithm's understanding of language 'has moved from a 2-year-old infant to something close to an 8 or 10-year-old child,' said Amit Singhal, a Google Fellow, an honorific reserved for the company's top engineers."

Here's what shapes my thinking about medicine, moonshots, and Google: https://www.statnews.com/2016/06/06/google-star-trek-fiction...

About PHP - I am aware that no one would actually try to implement this kind of thing in PHP. I was going for snark. I probably shouldn't have.

jeffmcmahan··on AI is always less impressive than is claimed
I hear you. I would say that we are not omniscient, yet we know which footstool is intended in the context, and without any conscious effort. Human-like general intelligence is marvelous. And it's well out of reach, and will remain so.
jeffmcmahan··on AI is always less impressive than is claimed
Yes, the trouble is that the work on performative acts (speech acts) only goes so deep, and then everything turns to confusion and ad hoc rationalizations presented like theories (I have a couple in mind). There'll be no means of implementing this body of knowledge in software -- or even taking it as a guide. So, if it were going to work out, you'd first have to solve the immensely challenging scientific problems, and then see about using the resulting knowledge to build something. As a person with some experience in the area (admittedly nothing published), I don't expect the former to occur. Ever.
jeffmcmahan··on AI is always less impressive than is claimed
The term refers to work on syntax (or phonology) that does not extend into semantics. So, basically anything Chomsky writes is on narrow syntax, whereas anything (say) Irene Heim or Barbara Partee writes is concerned with a broader notion of syntax (which includes semantics).
← PreviousPage 4 of 4