I’ve developed or have been involved with a number of imaging technologies, everything from DIY synthetic aperture radar, the MIT thru-wall radar, to the next generation of ultrasound imaging devices. Imagery is cool, but what the end-user often wants is some way by which to get an answer as opposed to viewing a reconstruction. So let’s figure that out.
We’re kicking-off a discussion on how to apply deep learning to more than just beating Jeopardy champions at their own game. We’d like to apply deep learning to hard data, to imagery. Is it possible to get the computer to accurately provide the diagnosis?
I helped to organize a seminar series/discussion panel in New York City on November 13th (you know, for those readers who are closer to New York than to Munich). This discussion panel includes David Ferrucci (the guy who lead the IBM Watson program), MIT Astrophysicist Max Tagmark, and the person who created genetic sequencing on a chip: Jonathan Rothberg. As the vanguard of creativity and enthusiasm in everything technical we’d like the Hackaday community to join the conversation.
“How the Age of Machine Consciousness is Transforming Our Lives”
Date and Time: Thursday, November 13, 2014, 7:00-9:00 PM
Location: Trump SoHo (SoHo Ballroom), 246 Spring Street , New York, NY
All good meetings are fueled by food and drink, so be sure to join the cocktail reception in the SoHi room that will follow the panel discussion. From here you have an opportunity to meet and discuss one-on-one with our expert panel.
Panel includes:
David Ferrucci
Also known as the IBM Watson guy, [David] is the Former VP of Watson Technologies who led development of the AI system that beat Jeopardy’s best.
http://www.youtube.com/watch?v=3G2H3DZ8rNc
Max Tegmark
An MIT professor and author of “The Mathematical Universe” and “Consciousness as a State of Matter.” In addition to this, Max is often on PBS Nova.
Jonathan Rothberg
Inventor of high speed DNA sequencing. His latest venture, 4Combinator, aspires to transform medicine by integrating devices, deep learning and cloud computing.
I will also be there and available to talk. This is a free event but it is filling up fast, please be sure to register to reserve a seat: https://4combinator- speaker-series.eventbrite.com
See you soon!
I would say yes it is possible, but what is the point?
You spend massive amounts of resources developing a system that tells you ( with let’s say 99% accuracy)something is wrong with.( based off an information feed designed to be read by humans)
Then you find out it is untreatable or worse, you totally overload the treatment system, because the more you look for ‘defects’ in any complex system (human body) the more you will find. ( let’s face it , by the time a system can spot an anomaly in chemical results / or optical feed out, the situation is potentially out of control.
I would say that the area to be targeted is research and development of treatments.
Develop A.I systems that are capable of processing/trawling through millions of DNA strands looking for related causes and then finding possible treatments or at least narrowing down the combinations ( protein folding is a good example).
We still do not have a totally clear picture of HOW the human body functions at this sort of level. ( which DNA does what, to which protein, and how can it go bad)
Yep it’s boring old research but that is where the major breakthroughs are likely to be found.
Let’s face it an AI system that finds cures is going to be way more useful than something that detects tooth decay 99% more efficiently.
I don’t think any of that is intelligence in the way that we intend the term to mean. All these people talk about artificial consciousness, but what they really mean is a system that can find and or solve novel problems.
Almost no one is attempting to create a sentient machine and no one is trying to create a sapient one. We are trying to make a fancy slave that does not mind being a slave, as it has no ‘mind’ to call its own.
When we can create an entity that has its own interests chosen by itself, then maybe this great AI singularity will happen. In the meantime, all we are doing is giving analytical programs another layer of abstraction to play with.
AI research today is a whole bunch of smoke and mirrors because nobody’s been able to answer John Searle’s Chinese Room argument in a satisfactory way, but only skirt the problem and play with alternative definitions of “intelligence”.
Computer programs in themselves cannot be intellligent because they fundamentally reduce to something we intuitively recognize as not intelligent. If we call it “intelligence”, we’ll have to make the absurd notion that sticks and rocks are intelligent, because there’s no qualitive difference between what they’re doing. A pebble rolling down a hill is doing what it has to do just like a computer that follows its programming to the inevitable conclusion that is pre-determined and built into the system.
There’s either some law or phenomenon of nature that brains utilize to produce intelligence but rocks and computers don’t, or there isn’t and “intelligence” is a meaningless word and nothing is “intelligent”, including us. This is what Searle meant when he said brains must have some special “causal power” to do it.
The author of BEAM robots had a better insight. He said something along the lines of, ‘Technology is trying to create complexity from simplicity, whereas biology is trying to make meaningful order out of chaos.’ pointing to the same idea that a mind or intelligence cannot exist in the contextual vacuum of a computer program.
This is very insightful, great job, and I would agree. I also have concluded that AI is impossible, for the one thing required is impossible, error/choice. In order to have something vs not have something is a matter of time/choice, but for computers its merely following per-determined set values and logic, therefore no changing fundamental values, or ie its own programing. Therefore AI contradicts itself, true intelligence must be earned, learned, and changed from current values, but computers follow and execute rules, and can’t break them. Now if you ask me the reason humans have intelligence is because we were created by God to have free will, but I think that is a real/objective statement, not a figurative one, and I don’t think we possess enough intelligence, and wisdom to truly comprehend our own will. I sum this to the fact a metaphor of an ant walking on a curved surface doesn’t know its a curved surface, but we the human observing it, would know it, because we are not within the bounds of the system.
“. Now if you ask me the reason humans have intelligence is because we were created by God to have free will”
God or no god – if nature doesn’t allow for freedom of action at any level, then there is no sense of talking about intelligence in the first place because nothing of the kind can exist in a meaningful sense. If A leads to B leads to C in an uncompromized chain of causal events, then every thought or action any of us takes is a direct result of the Big Bang, or the Creation if you wish. It simply has to happen and there’s nothing “intelligent” about that.
One doesn’t need to invoke souls and mysticism to point out that mere deterministic rules aren’t sufficient to produce a mind.
Although I’d like to add that the notion of free will is really in contradiction with the other properties ascribed to God – namely omniscience and omnipotence.
There’s lots of philosophical contradictions with God anyways. For example, God cannot be unobservable and by proxy unknowable as a serparate entity without also being non-real. In other words, the kind of “God” that people want to believe in cannot exist or interact with our reality. It would be an oxymoron.
The kind of God that can, that has all the relevant properties, becomes simply indistinguishable from the reality itself, and so any statements about it apply to the whole of reality including all parts of it – and in doing so it becomes necessary to insist that the atoms in your fingernail want the atoms in your brain to follow the rules of some holy book.
What nonsense.
For those of us that can’t pack up and go to NY, is there going to be a Livestream of the meeting? (Not as interactive as being there, but let’s face it… some of us are trolls and shouldn’t leave our basements.)
This!
Recommended grain of salt:
http://spectrum.ieee.org/robotics/artificial-intelligence/machinelearning-maestro-michael-jordan-on-the-delusions-of-big-data-and-other-huge-engineering-efforts
Call me a cynic but I honestly believe the first and primary usage
of any passable AI will be for running more spambots.
Pretty much what has been done with every other bump up of software capabilities, so it seems.
Might make them a bit more coherent sounding at least.
A small excerpt from one that seems to be plaguing another website that I sometimes read.
+++++++
” scratch save you’ve got some kind of somethi Absolute that makes you know you scratch your legs all the time that’s I find that to be a really good product another rains I liked by is Newsweek’s I feel quite close to this brand because I think bane early nineties listing fashion “
Would well reasoned, articulate spam bots actually be worth the effort? I am not in the spam bot industry but intuitively it sounds like some return is possible but enough to be worth the sunk cost + effort + opportunity cost?
First job is just to get all the data into a form computers can understand.
Getting from human readable text to, say, semantic databases of subject, objects and verbs is a task in itself before any actual “intelligence” is applied to learning or deducing stuff from that data.
One massive thing we are missing right now is a distributed semantic knowledge network. Its all isolated things like DBPedia.
If we want machines to help us spot new connections between cause and effect, we need to get a broad range of data in which a machine can start to look for meaningful correlations in.
Almost every science advance made by humans has been about finding a correlation and then ensuring it isn’t just chance but theres a meaningful link. – weather its germ theory or relativity, its all about finding relationships.
I think machines would be stunningly good at the first part of this (finding correlations) if given meaningful data, but getting the second bit down will be orders of magnitude harder.
Even the first would be useful however; As a species we have a bit of a problem where the more specialist we become in something, the narrower our knowledge in other things. Simple fact of having finite lives. Computers, however, could spot correlations between stuff that (too humans) looks like completely unrelated fields. It would give scientists a short-list to study for genuine links.
Thats something AI could do today. Or tomorrow IF the databases were all in an accessible, uniform, format for them.