For the first time I came across a book where the author
tries to coerce us to read the 'prelude', which, of course, I did and later
found that it was only a cold-war-era type introduction to a clandestine group that
forms the dark side of the internet- and that controls the artificial
'statistical' intelligence. This, the author asks while I was fast
reading the second chapter. The Omega and Prometheus are probably the fictional
entities that rule the internet and thus have scores of information and data that
could be used constructively or otherwise.
Written in an accessible and
engaging style- and at times tedious- the book is aimed at the general public
and offers a political and philosophical map of the promises and perils of the
AI revolution. Instead of pushing any one agenda or prediction, Tegmark seeks
to cover as much ground as possible - as you would read here, reviewing a wide
variety of scenarios concerning the impact of AI on the job market, warfare, and
political systems.
However, Life 3.0 does a good job
of clarifying basic terms and key debates (with enough references), and in
dispelling common myths. While science fiction has caused many people to worry
about evil robots, for instance, the author rightly emphasizes that the real
problem is the unforeseen consequences of developing highly competent AI.
Artificial intelligence need not be evil and need not be encased in a robotic
frame in order to wreak havoc. In Tegmark’s words, “The real risk with
artificial general intelligence isn’t malice but competence. A super-intelligent AI will be extremely good at accomplishing its goals, and if those
goals aren’t aligned with ours, we’re in trouble.”
On the
easier side, instead of diving deep into ethical issues and valid concerns
around AI, he speculates on different scenarios in great detail. So much so that
at times we would think that this is a purely fictional work. He goes into endless speculations of how AI
might go wrong in multiple ways and takes the obvious joy of making those as
detailed and long. And then he would
suddenly talk about the existing computers and their accomplishments with the
aid of Physics and Biology.
To give an example of how preposterous his scenarios are he
claims the advanced AI will need a lot of energy that would be hard to fulfill
(about half an mg of matter to empower the 13W brain) and also that once
we are able to build mega architectural structures we could tap into the energy
of a black hole to provide the necessary power to get along Life 3.0 (AI
revolution). Of course, Life 1.0 is pure bacterial and animal evolution while
Life 2.0 is cultural evolution as a result of human beings. The Pandemic was
secretly targeted to ensure that nobody who knew anything about Science and Technology
survived.
Here are some highlights in the order of the book:
Tiruchirappalli gets a mention as a place for fake suppliers
of 'orders' based on AI by the ghost Omega.
That the DNA can store 1GB of energy and not beyond is a
topic to ponder upon.
Intelligence is the ability to accomplish goals. John McCarthy
coined the word Artificial Intelligence. Natural language translation using AI
would be very easy in the future. There is a lot of detail into how memory works and
the description of AND and NAND gates take an entire chapter.
There is a description of the Ariane 5 rocket of the ESA which exploded 37 seconds after its launch due mainly to intelligence failure and was mainly a buggy software manipulating a number too large to fit into 16-bit allocation. The same kind of thing occurred with NASA's Mars climate orbiter - the difference in units caused a 445% error. Another of its Mariner spacecraft malfunctioned due to an incorrect punctuating mark and a Hyphen caused end of shutting down the mission to Phobos.
Robotic surgery accidents were linked to 144 death and 1391
injuries in the US during 2000-2013 and AI drones can kill accurately. Was that catching?
A chapter endorsing fears talks about AI generating fake
realistic videos which could help the government and pose dangers of misuse by miscreants. An instance quoted was the failure of a guided missile system that killed 290 people in a Gulf war. Just
as chemists and biologists have no interest in chemical and biological weapons
AI researchers also have no interest in building AI weapons.
Work keeps at bay
three evils; Boredom, Vice, and need - Voltaire
An intelligence explosion is expected just like an explosion reaction
that can occur due to chain reaction as depicted in Nuclear Science. Game
Theory and Nash equilibrium find mention.
Freeman Dyson gets almost a chapter. His sphere the size of
Earth's orbit would give us about 500 million times more surface area to live on!
The efficiency values given in Table 6.1 is nice and a few are
mentioned below:
- 1. Digesting a candy bar is 0.00000001%
- 2. Burning Coal is 0.00000003%
- 3. Burning gasoline is 0.00000005&
- 4. Fission of Uranium is 0.08%
- 5. Black hole evaporation is 90%
Galaxies, ET life, and Inorganic intelligence are cited in an
elaborate manner in a chapter. Chandrasekar limit which says any mass 1.4 times greater than the sun
would explode supports all the explosions observed by present-day telescopes using AI. Big Crunch is also obvious. For the Big Crunch to occur a rubber
band is equated that can burst with over expansion.
Goals: If we cede
control to machines that don’t share our goals then we are like to get what we
don’t want. Pursue goals instead of following rule of the thumb. That
reality is relative and more so because visual experience can’t reside in the retina.
Brain rebels against its genes to
become a monk or nun. To gain weight we carve to have more sweets. It is a kind
of hacking. The ultimate goals of the system can be independent of intelligence and
Consciousness is subjective experience. Water is wet but ice and steam are not
- arrangement governs the emergent phenomenon
If it is falsifiable
it is not scientific - karl Pepper
The last chapter about AI ethics and principles would make
an interesting reading if only the
reader is truly connected with AI and the photos pertaining to these meetings
and conferences of all the scholars involved make a banging end.
