3 ms·

Ex-Googlers aim for 'general intelligence' to make software do what you tell it

8 points by samizdis 5y ago

egfx 5y ago
Actually a pretty cool approach for general intelligence. Aggregate platforms like mashape could be set to go this way too. I could see programs of the future contributing to an open knowledge map on how to interact with their interfaces rather then just the data. Smart stuff.
togaen 5y ago
So… it’s Siri.
- ethanwillis 5y ago
  I mean, what else would you expect from Грантожери? In my opinion the Silicon Valley culture is very adept at identifying/inventing and tackling problems in the best way for maximizing investment. This is sometimes different from identifying and tacking problems in the best way.
ethanwillis 5y ago
I had recent thoughts on this exact topic coming from a different direction and my experiences with foreign language learning[1]. Here is the fundamental problem as I see it. When you, as a programmer, are trying to build an integration between two programs, what are you doing? Generally today the two programs will each define an API interface. You as the programmer grok your understanding of these two interfaces and in one or both programs you will create what is essentially a creole language to allow one-way or two-way communication. In order to create this language, you are using your human brain to "acquire" the language of each program first to then be able to build this creole bridge. What if instead there was simply a third program that was an analog to the human language acquisition device(LAD) that could carry out this process automatically by inspecting two other programs? There are a lot of open possibilities here. My original idea is to allow automatic communication between these programs as a way of knowledge transfer where each program could in a way "absorb" parts of the other one. To then let each program be able to implement new functionality that makes use of routines, data, etc. from both programs. They go from being silos where they each have a specific view of the world, to a combined world view. You can see this in differences between human languages[2]. A new, second idea, that is a more nuanced version of the first.. is that this 3rd program only is concerned with creating this creole language automatically. From there either an AI or human only has to interface with that language in order to develop new functionality . I'm just going to cut this comment short because I could go on and on exploring a million different angles on this. [1] https://github.com/ethanwillis/thoughts/blob/master/Communication.md https://github.com/ethanwillis/thoughts/blob/master/Communic... - Not super well thought out and I've thought about it a lot more since then. [2] Interestingly there are no direct words for "bias" or "hypocrite" in Ukrainian in the way an English speaker would mean them. There is also a great word Грантожер(Hrantozher), "Grant Devourer", that roughly means "A person who doesn’t care about what they have to do." An example would be someone who is working on some scientific project, just because it is funded, not because they care or believe in the problem being solved.

8 points by samizdis 5y ago

egfx 5y ago

Actually a pretty cool approach for general intelligence. Aggregate platforms like mashape could be set to go this way too. I could see programs of the future contributing to an open knowledge map on how to interact with their interfaces rather then just the data. Smart stuff.

togaen 5y ago

So… it’s Siri.

ethanwillis 5y ago
I mean, what else would you expect from Грантожери? In my opinion the Silicon Valley culture is very adept at identifying/inventing and tackling problems in the best way for maximizing investment. This is sometimes different from identifying and tacking problems in the best way.

ethanwillis 5y ago

I had recent thoughts on this exact topic coming from a different direction and my experiences with foreign language learning[1]. Here is the fundamental problem as I see it. When you, as a programmer, are trying to build an integration between two programs, what are you doing? Generally today the two programs will each define an API interface. You as the programmer grok your understanding of these two interfaces and in one or both programs you will create what is essentially a creole language to allow one-way or two-way communication. In order to create this language, you are using your human brain to "acquire" the language of each program first to then be able to build this creole bridge. What if instead there was simply a third program that was an analog to the human language acquisition device(LAD) that could carry out this process automatically by inspecting two other programs? There are a lot of open possibilities here. My original idea is to allow automatic communication between these programs as a way of knowledge transfer where each program could in a way "absorb" parts of the other one. To then let each program be able to implement new functionality that makes use of routines, data, etc. from both programs. They go from being silos where they each have a specific view of the world, to a combined world view. You can see this in differences between human languages[2]. A new, second idea, that is a more nuanced version of the first.. is that this 3rd program only is concerned with creating this creole language automatically. From there either an AI or human only has to interface with that language in order to develop new functionality . I'm just going to cut this comment short because I could go on and on exploring a million different angles on this. [1] https://github.com/ethanwillis/thoughts/blob/master/Communication.md https://github.com/ethanwillis/thoughts/blob/master/Communic... - Not super well thought out and I've thought about it a lot more since then. [2] Interestingly there are no direct words for "bias" or "hypocrite" in Ukrainian in the way an English speaker would mean them. There is also a great word Грантожер(Hrantozher), "Grant Devourer", that roughly means "A person who doesn’t care about what they have to do." An example would be someone who is working on some scientific project, just because it is funded, not because they care or believe in the problem being solved.