At 12:59 PM 10/27/98 +1300, "J. Maxwell Legg" wrote:
>Robin Hanson wrote:
>> Investment capital does not imply "money." Capital is anything
>> that aids in production. Investment is using capital to make
>> more capital, rather than helping people to "consume." The
>> supply of investment capital is capital that those who control
>> it are willing use to invest, rather than consume. The demand
>> for capital is the investment projects that are available.
>> If those AI engines help us produce more, they are
>Robin, until other mechanisms for managing the flow of information
>are produced then all measurements of capital become hostage to the
>current accounting concepts of capitalism, i.e., orders, invoices,
>statements, profit & loss, balance sheets, etc. There is no escaping
>this and no amount of hiding behind behind text books will convince
>me that your terms are anything but obfuscation. The only way to
>break this hold over terminology is to explicitly deny the case for
>double entry bookkeeping and start a new information paradigm with
>a new set of terms.
This is not entirely correct. There *are* forms of capital that do not show up on balance sheets and which are difficult if not impossible to value using conventional monetary systems, but which are critically important in the modern economy nonetheless. The best example is intellectual capital. Not only does this not show up on balance sheets, but there is no known legitimate way of quantizing the actual value of it, particularly in its more abstract forms. How do you value something as simple as a database of collected information? Remember, the value of data exists primarily in the almost infinitely large number of relationships and uses that can be produced from the data, not in each datum itself. Any attempts at placing a value on any large set of data uses a narrow accounting perspective by necessity.
Capital markets run by people, however, *do* take into consideration capital that is not accountable in a financial statement. The fact that traditional financial indicators are becoming less important in the capital markets (such as Wall Street) is an indication that the capital markets do recognize the value of non-monetary capital even when it is not quantized. This also explains why technology companies (which tend to hold and exploit large quantities of intellectual capital) can often be valued so highly even though the financial fundamentals never existed. There is, of course, a constant conversion between abstract capital to monetary capital, but that is because both are useful and required.
I don't really see the need to have one large neural net when we have billions of small neural nets already doing a pretty good job. I suspect the improvement would be incremental at best and possibly not worth the capital investment to create and maintain it. If it is worth the capital investment, then rest assured that the current system will create it eventually.