Ben Goertzel

About
is an American author and researcher in the field of artificial intelligence. He is currently Chief Science Officer of Hong Kong financial prediction firm Aidyia Holdings, Chief Executive

Hosted AMAs


Highest Rated Comments


bengoertzel18 karma

I have been told that drinking a mixture of Robitussin PM, vodka and DMT allows one to create one's own private Singularity !!!

bengoertzel15 karma

I think they are wrong, at the level where it matters....

If you look at things abstractly enough, sure, there's a single algorithm like "Search the space of all procedures and find the one that, based on your experience, has the property that its execution is most likely to achieve your goals; then execute that procedure, while also continuing the aforementioned search."

But if you look at things at the level needed for practical, scalable, real-time implementation -- THEN the one-algorithm perspective is not really the most useful one.

The brain is a huge complex mess with many different subsystems doing many different things. It does not have the elegance and simplicity of a computer science algorithm.

bengoertzel15 karma

There is plenty of evidence that AI can work well for financial prediction, yeah....

But my particular thoughts and discoveries about that are not for public consumption!

With some more hard work and just a little luck, within a few years, the AI-based hedge fund Aidyia Limited that I've co-founded will have a huge trove of money, and then funding AGI will be something I can do personally..... But that's still early-stage, we're still testing and haven't started trading yet....

bengoertzel15 karma

Regarding AI boxing --- the question is: If Eliezer Yudkowsky were locked in a box, only able to interact with the world via text interface, would I ever feel like letting him out?? ;D

bengoertzel14 karma

As an example, a year ago I spent a while outlining a design for making OpenCog fully exploit massively distributed computing infrastructure: http://wiki.opencog.org/w/DistributedAtomspace

How much progress has been made on implementing that? Zero. Why? It's too big for a volunteer, and I don't have access to extra $$ to pay someone to do it....

bengoertzel13 karma

With a very large amount of funds (let's set aside "unlimited" since potential infinity leads to various strange things!), I would pursue the same research program as now, since it's the only one I solidly know how to make work.... But we would certainly proceed a lot faster. (And also, of course, if I had a truly massive amount of $$ I could fund other AGI projects besides my own, and we could see which ones worked best, and maybe create many kinds of AGIS...)

Yes, there is the talent and infrastructure to build an AGI much faster than is currently being done. However, people have this frustrating desire to pay rent and their electric bill and buy food and put their kids through college, etc. --- and with very rare exceptions, society has little desire to pay them to work on AGI R&D .....

So as fast as progress may seem in historical terms, when you're involved in the actual R&D, you see progress is way slower and more awkward than it could be, for purely boring practical/financial reasons

bengoertzel12 karma

it would be a tie, since they are both dead

bengoertzel11 karma

If you wanna learn about AGI and have a little technical background, the links here are a place to start:

http://www.agi-society.org/resources/

bengoertzel11 karma

However, if you mean SIAI/MIRI's notion of "provably Friendly AI", I think that is a fanciful notion without much foundation in reality.

We don't understand the universe in which we and our AGIs are embedded that well, so even if someone miraculously came up with a proof that some AGI design would always play nicely even as it improved its own sourcecode and became superintelligent --- the premises of their proof might get invalidated by new discoveries about real life.

While those guys are struggling to grasp with the math of phantom nearly infinitely powerful self-modifying AIs, others are going to build real thinking machines and obsolete the SIAI/MIRI daydreaming ;)

Not that there's anything wrong with daydreaming...

bengoertzel11 karma

For myself ... well, I'm already happily married and my wife seems not that open to polygamy.... So I may have to say not to the robot wife myself...

But for others, sure, why not?

Actually, marriage in its current form is an artifact of legacy human society and psychology, which will seem pretty archaic in the future to those of us who become superhuman superminds....

BUT some kind of partnership/ coupling / partial fusion between minds may still exist, in forms we can't yet understand or foresee...