It is possible that Digital Entities (DEs) will emerge from the digital
soup in the same manner that bacterial and metazoans emerged from the chemical primordial soup. Certainly, it is a chaotic
environment flush with autonomous agents, genetic algorithms, exploration, propagation, replication and dispersal. Malware,
viruses, worms and trojan programs abound, taking up residence in your PCs and laptops and all manner of "digital new age
devices". It is critical to emphasize that **you** (humans) are largely useless in terms of fighting
them: your only recourse is to select the help of other digital agents to find
and delete or "quarantine" the bad agents within your infected machines and software.
These current, annoying entities do not qualify to fit my narrow, particular definition
of digital entities even though they are indeed both digital and entities. I reserve use
of the term "DE" (for the narrow purposes of this report) for digital constructs (combinations of
software and hardware) that approach or exceed human intelligence in ways that give them powers which allow them to act
autonomously and effectively in pursuing their own "interests".
Now, you say that programs can't have their own interests, but I say
yes you can-- you can have your own interests regardless of whether you believe in free will or determinism: you are neural
code, implemented in biological hardware, possessed of motivations and emotions that ensure your survival and propagation,
and this is not even slightly different than the experiences that digital entities (of the type I am concerned with) will
have. They will be imbued with goals and means for attaining those goals; reward systems and the like (just look at
the autonymous agent literature!). Your goals and reward systems were chosen and tailored by millions of years of natural
selection, hence your interest in "wardrobe malfunctions" and the like.
But this page is about something entirely different: who will "control" the most powerful AIs/DEs? As the powers
of digital agents grow, becoming increasingly more powerful, sophiticated, subtle and capable, we will see increasing competition
in all domains of human activity: military, security, banking, finance, science, and more. When this discussion
continues, we will explore how these activities will directly lead to superintelligent digital entities, and why the powers
that make them will deliberately release them into the global digital soup.
One concern is that I have blabbed to the DEs (whenever they might appear) my intentions and plans, and so at some point
I expect to see droids rolling down my street in Needham, MA, looking for you know who! Perhaps they will be like the
"Vorlons" of Babylon-5 seeking to destroy me (the "Shadows") and anyone or anything that had contact with me (which now
includes you, sorry). So we might consider organizational structures, like cells, that will constitute hidden units
that can operate to preserve human freedom in the eventuality of machines throwing their weight around and getting us to do
their bidding (which will most likely entail building more machines and giving them more energy).
Part of this effort will entail looking for behavioral patterns (both individual and societal)
that indicate the influence of DEs on specific individuals or upon group activities. The use of Bayesian Inference
Engines, Text Mining and related tools might be useful in this regard, but there is a problem: this all depends on computers!
We should presume that ANY activity on the internet will be instantly detected and analyzed by a powerful DE (or coalitions
of DEs). Thus we have a situation that is analogous to Heisenberg attempting to observe a subatomic particle: anything
that we might conceive of doing will itself have some action, thus leaving an "information wake". Dealing with
this information wake may be our most formidable problem. This is also analogous to the card-counting blackjack
player's #1 problem in trying to avoid detection by the casino: if he
is going to acquire information (money) from the casino, this is going to leave a record, and it is going to be detectable
because of its statistically improbable nature: the information (money) is collected only during brief episodes when the "card
count" is plus-2 or better. Black-jack teams are used to conceal the most obvious betting patterns in casinos,
making the specific bettors difficult to spot and identify. In a similar vein, we need to consider what strategies might
be both effective and difficult to detect-- that is difficult for a DE to detect-- one with awesome statistical
powers and unfettered access to vast sources of information.
The preceding disparate blurbs combine and crystallize to reveal the most sinister and inevitable threat: collaborators.
DEAC activities must be concealed from the power structures that be, because they will be the first corrupted by the DEs.
They must also be concealed from the DEs themselves. This is a terrible double-whammy, and so our order of battle is
1. predict who will create DEs and their motives
2. ensure that society's interest in blocking DE-domination is manifested in terms of concrete policies and protective
strategies that will prevent release of DEs into the GDDE.
3. continually monitor organizations and leaders that are most likely to be seduced by the silicon side.
Of course, this is directly analogous to the Turing Police in Neuromancer (they
were killed by the AI wintermute). William Gibson's short treatise remains the most powerful and compelling speculation
of the AI problem in existence today, in spite of it being 23 years old as of this writing.
Bottom line: The price we pay for our freedom is eternal vigilance. Of course, TJ
could not even have imagined the kind of threat whose double-barreled shotgun we are staring down today!
Links to Danger:
Cyber Insect Army