Powerset’s First Public Demonstration
July 2nd, 2007 –
I was invited last Thursday to Powerset for the first demonstration of their NLP search engine. They are a startup from Lorenzo Thione, Steve Newcomb, and SIAI Advisor Barney Pell. The invite promised that the singularity will happen at Powerset; while I didn’t see any signs to believe that were true, it was clear they have achieved some impressive, early results. They have a sharp, experienced team, a sound strategy for scaling by riding the growth curves of computing capacity, data storage, and bandwidth; and the best technology-base from which to try to solve the complex problems of robust natural language processing. The search engine will be available in September through their Powerlabs community, which you can sign up for here. I also recommend the following write-ups about their demonstration:
Powerset: The natural language search mashup platform
Powerset: No pretender to the throne but the real deal
Powerset: Open source approach to beat Google
Powerset is not a Google killer!
I’m skeptical that this is “Open Source” It does not appear to meet the definition of “Free Software”, either.
“The patents and technology are locked down,”
“We could have 50,000 people QAing **our** product,” [empahsis mine]
Having a relatively open API that users and developers can access royalty-free is certainly admirable, but it bears noting that the freedoms granted by publicly licensed software are as missing from Powerset’s operations as they are from Google’s.
It is trendy for businesses that thrive on withholding freedoms to move from proprietary shrink wrap software models to web services models, where they run their software, unreleased, on their own hardware, where they can control it. But calling such a model “open source” is deceptive.
Just a clarification here. Powerset is contributing heavily to the open source community in terms of distributed computing infrastructure. But we have no plans to make our core NLP technology open source.
“But we have no plans to make our core NLP technology open source.”
And there’s nothing at all wrong with that, in my opinion. At it’s core, there is nothing immoral about seeking a profit; buisness “morality” should be judged by other metrics. Without an economy and money serving as motivator, a Singularity would surely never happen.
“search engine based on natural language processing”
A lot of people say that in order to understand natural language you have to actually have a fully-fledged human level AI. I’ve certainly heard it referred to as an “AI complete” problem. Perhaps this sentiment is wrong?
Anyway, I’ll be watching this closely. Good luck Powerset
Yeah, if Powerset is not exaggerating then this is crazy exciting. It should really streamline the world, and make the big S all the more tangible. Talking with machines is fun.
Go Powerset!!!
“The invite promised that the singularity will happen at Powerset”.
Did Powerset actually say this in the invitation to the demo, or was this just implied?
This was just a joke in the invite I received; it wasn’t part of the demo.
“A lot of people say that in order to understand natural language you have to actually have a fully-fledged human level AI. I’ve certainly heard it referred to as an “AI complete” problem. Perhaps this sentiment is wrong?”
I’m *far* from an expert, but I wonder if NLP could be conceptualized as something similar to a subsystem or submodule of an AI mind. Similar to say the visual cortex, where very valuable processing occurs - but a component that in isolation doesn’t embody general intelligence or “consciousness”. …?
I wonder what effect, if any, robust NLP will have on future program writing. Could some form of NLP serve as the next paradigm in high-level programming languages?
Just wanted to harp-on a bit more (commenting is fun)… Perhaps a mature, powerful NLP capability would rival (or even surpass) the discrete complexity of a submodule of the human brain (eg. the visual cortex), yet still not have the broad applicability or general-purpose utility of a complete human brain - and so wouldn’t have general intelligence. But, I’m just spit-balling here. Dr. Pell very briefly discusses this ambiguity in his SIAI interview video segment. It’s a bit difficult to glean his take on this issue, but my impression is that he suspects that a fully realized NLP capability will equate to general intelligence. My impression could be wrong, of course.
Jeffrey: Here is my view.
I think it is important to distinguish the ability to communicate using language from the ability to have knowledge and experiences that can be communicated and shared.
If I am blind, then I can’t have the same kind of conversation about visual aspects of the world, but I can still use language effectively.
Similarly, we can have computers with powerful NLP capabilities without this requiring that they have the same breadth and depth of understanding as people. Language is a very powerful window into a person’s mind, but the rest of the mind still has to be there too.
Thanks for the clarification, Dr. Pell. Robust NLP will definitely revolutionize the world - no doubt about it
Do you believe that a fully mature NLP (with a huge knowledge base) capability will embody a general intelligence at all (at any level)? Some people seem to conceive of “understanding” in terms of complex concept associations/linkages. Or do you consider an AGI to be a different animal altogether?
Can’t wait to try out Powerset.