Marco Fuxa

*Mise à jour il y a 17w* · L'auteur dispose de réponses 1k et de vues de réponses 1.3m

Accounting is a perfect *binomial no trust system*. It is not in need of improvement. Blockchain will simply extend the double balance verification and no trust paradigm outside the corporate balance sheet. In other words it will allow multiple balance sheets (id est corporate entities) to be linked and reconciled for mutual no trust verification.

In accounting no “artificial intelligence” is needed, its a congruent arithmetic system. *Accounting is not a probabilistic non mathematically congruent and consistent system*.

You could argue that AI is not logical. Well, because its not.

You could also argue that AI is not rational. Well because its not. It has neither of the classical greek aristotelian logos, or the rational roman *rapport *(in arithmetically congruent proportion).

Where irrational, transcendental and infinite mathematics fail, AI offers some level of empirical feedback. At some level it smooths out the irrational by empirically limiting the infinite. But since its based on a fundamentally flawed system to begin with, you are only improving on the improbable.

Finance, as opposed to accounting, is not a backwards looking statement end-fixed at a specific calendar date. But even finance needs to terminate the progression (limit) at some future time. Since this is a speculative forward progression, artificial intelligence can help determine the limit of the progression.

Algebra cannot go beyond four variables. This was proven more than 200 years ago. That means that the limit of our mathematical logic system is incapable of addressing all real life possible eventualities. The real world is not limited to four variables.

Accounting can account for transacted data points perfectly within the logical limits of the binomial number system, but finance cannot assess an open ended system absent a risk limit.

Real world dynamics means that risk is entirely open-ended. AI, being mathematically incongruent in whatever base it presents its mathemathematical findings, on the other hand, *can empirically limit the probabilities*.

AI is a *probability limiting system*.

What the corporation does, is fixed in scope. Its activity portfolio is constrained by its organizational reach. Its activities are finite, and consequently its accounting is finite.

A bank on the other hand has the opposite problem. It is the sum total of all its risk-limits however conceived. If you price risk on an individual unitary basis, you can somewhat decide what the limits are, or price the out of contract, out of progression limit cost. All buy orders are programmed with a-priori sell orders.

What the bank cannot do is truly price a progression of multiple separate investments. **You can sum the progression from individual unit limits, but its the law of large numbers, and those are probabilistic. **Anything probabilistic can be improved via so called artificial intelligence. Large data sets can reduce the risk associated with a unitary event, but that is only by virtue of diluting its potential reach with the progression of *sommes *(emphasis on plural). What it cannot do is eliminate the black swan events, or eliminate systemic risk.

Accounting does not require ** analyse**. Either both posts balance or they do not. You cannot approximate anything. A one cent discrepancy wrecks its counter party entry. Thus you have two evident errors. You have the original entry error, and a secondary proof of error. Its method

*et la preuve.*

Problem is of course that you cannot do logically mathematically congruent arithmetics in binary. You certainly cannot run binomials that way. Binaries have weird properties, and the two entries will never match.

I even have to use the written word *seize* because 16 in binary means something completely different. Regardless:

16 1’s = 65535

but using two’s complement binary addition negative 16 1’s can only work out to:

Neg 16 1’s = 32,76**8**

And by extension the positive signed reciprocal:

Pos 16 1’s = 32,76**7**

Which is just about right if you take the original 65,535 and divide it by two = 32,767.**50**

So if you are going to make two entries in a balance sheet to offset each other, how pray tell are you going to do that in a binary system? You cannot.

Binaries do addition with a known error rate, subtractions with a con and an estimated error adjustment, they cannot do multiplication other than through addition and are completely incapable of doing division.

Doing a division by multiplying the reciprocal is substantially impaired by the logical flaw in the binary numbering system.

If you want to give your self a headache knock yourself out:

Jones on reciprocal multiplication

Using “AI” to do accounting is a method by which you can average out the error rate inherent in the binary arithmetic base system. But why average out an error when the original system does not present the error in the first place?

You can extend the binary system to 64 bits or even 128. But that is the limit of the system. After that logical limit you loose the decimals. They in fact go bat shit crazy irrational and no longer make sense. Or expressed mathematically the error rate goes from exponential to infinite.

Similarly if you price a Bitcoin above $1000 you logically loose the ability to capture the decimal cents. I wrote a brief post on this: Marco Fuxa's answer to What are the odds that Bitcoin is undervalued and 99.9% of all other cryptocurrencies are overvalued?

In order to do proper accounting and logical mathematics you can call on your trusted IBM suited representative and have them deliver a proper accounting system designed around real decimal numbers in base ten and a real world clock based on base sixty.

Even the original IBM PC running at 4.77 Mhz had a BIOS chip (6818) running Binary Coded Decimals running real base sixty. Even the MOS 6502 running at 1 Mhz in 1975 had a better arithmetic unit than that found in a modern processor.

You could also chase down the BCD (Binary Coded Decimal) supported functions in the Intel 8086x MMX extensions. Those assembly language routines (mostly undocumented) are used for common vector operations in pixel graphics manipulation.

Or if you are hell bent on reversing prime number multiplication you can buy your self a proper video card and use their numbers. At least they are congruent with a *three variable* algebra.

Anything that is “probable” can be refined using AI.

Anything that must be a double proofed given, cannot benefit from AI.

Blockchain will allow separate entity balance sheets to communicate “no trust” and reconcile.

This reconciliation will allow for automatic property transfer of title, which is the fundamental basis for Ethereum.

AI will allow for efficient arbitrage in credit markets, thus reducing the price of risk, and increasing liquidity.

Every step in the supply chain can affect a corresponding change in the title holder *with the addition of the previous holder’s value add*. Thus you can eliminate trade financing with the block chained movements of goods for components, to sub assemblies to merchandisable product. With every transfer you transfer the financing risk to the next chain link. Every title change entry gets double posted on both balance sheets, and from there down the chain.

The same can be done with purely digital goods, where all purchases are simultaneously sales with the addition of some other digital good/improvement/innovation.

A digital good is a transformable good. Digital rights management is not about title passing per se, its mostly about the ability to innovate by transforming and adding value to an ecosystem of innovations.

Every entry must have its corresponding entry, and every entry must be reflected in somebody else’s digital balance sheet.

Every step must be a ledger to ledger verifiable transfer, or transformation in series.

The accounting will not change. What will change is the communications protocol and methodology for balance sheet to balance sheet reconciliation.

Two Prime multiplication and reversal (original concept for Bitcoin), like accounting, cannot be done in binary. What goes for “settlement” is just an *approximation of the reversal*.

The future is not known. Anything that allows you to place a limit on the range and reach of possibilities will fall under the rubric of AI.

Accounting is a known system. It does not need and cannot use AI. It will benefit from blockchain if it will allow arm’s length balance sheet to balance sheet reconciliation within an extensible no trust paradigm based on congruent mathematical logic.

Based on use case that I have seen, AI does not even reduce risk. Its just a *rule based system* that eliminates common sense and enforces its own rules at the expense of individual choice. Its a choice limiting instrument.

A little bit like calling in management consultants to have them recommend what you already knew had to be done.

AI is nothing more than rule based consensus governance system with empirical feedback based on an averaged mean.

Step outside of the mean, and you will be treated with the same contempt as I have for binary numbers…

Merci pour le demander.

**Richard Leddy**, Worked in AI research groups and share a patent for an AI system

*Répondu il y a 18w* · L'auteur dispose de réponses 528 et de vues de réponses 494.6k

My impulse response is blockchain only because it is beginning to change it. But, that is only the record.

True, accounting is really about reporting the progress of a business. So, blockchain is really changing bookkeeping.

But, if something like blockchain is pervasive, then there may be a record that could be used to feed AI for accounting purposes. But, of course, there would have to be clear text describing transactions.

If AI gets good enough, then it might be able to figure the health of a business and make suggestions about new directions. But, a deeper thinking may be required than is found in popular suggestion engines. So, depending on what you know of the subject, you might be able to guess about what some accounting AI expert systems might be able to do.

**Amanvijay Jindal**, Engineer,Actor,Lawyer(2/3),Manager,Inventor,Entrepreneur

*Répondu il y a 18w* · L'auteur dispose de réponses 516 et de vues de réponses 516.7k

artificial intelligence will change it