Wiki Contributions


EURISKO resurfaced 

"Doug Lenat's source code for AM and EURISKO (+Traveller?) found in public archives

In the 1970s to early 80s, these two AI programs by Douglas Lenat pulled off quite the feat of autonomously making interesting discoveries in conceptual spaces. AM rediscovered mathematical concepts like prime numbers from only first principles of set theory. EURISKO expanded AM's generality beyond fixed mathematical heuristics, made leaps in the new field of VLSI design, and famously was used to create wild strategies for the Traveller space combat RPG, winning national competitions two years in a row, even across rule changes to stymie it, before semi-voluntarily retiring. His magnum opus Cyc was originally intended to be a knowledge assistant to EURISKO's discovery engine.

These first two programs have intrigued the symbolic AI scene for 40+ years, with their grand claims but few eyewitnesses. While AM was technically available to fellow Stanfordians at the time, Lenat kept the source code to EURISKO close to his chest. Papers written about them carefully avoided technical implementation details. Lenat said he didn't retain any copy of the programs, when asked in recent decades, nor have any copies of AM carried publicly into the present."



I still think that space attack is taking longer time - than missile attack during a nuclear war on Earth, and thus anti-ballistic missile systems of the future will be even more effective in its stopping; including space lasers. Incoming spaceship will be visible for days or even months.

Therefore, the only winning strategy is to make the attack spaceship either invisible or moving with very high speed. It is contradictory requirements as getting the spaceship to the high speed requires a lot of energy (or time if gravitational maneuvers are used). 

One more problem: no help from grandmothers. 

Younger parents can expect being helped by their own parents in babysitting, cleaning, money etc. Older parents are on their own as their parents are already either dead or too old to help. 

This observation is based on personal experience. 

ChatGPT can't report is in conscious or not. Because it also thinks it is a goat.

May be also erosion of human values, so they will be more alienable with AI?

My approach to the personal stupidity issue is the "additive nature of intelligence" theory. It suggests that by spending more time on a problem or by seeking assistance from others (like AI, search engines, or people) and using tools (writing, drawing, etc.), you can achieve the same outcomes as a smarter individual.

This concept also posits that certain ways of thinking, such as Descartes' method, can yield notably good results. 

There are a few other escape strategies which also can be inverted:

- runaway at a speed of light from unaligned AI

- become extremely small, for example, upload humans into nanorobots who live on a meteorite

- become transparent (see also internal immigration)

- active boxing: AI is boxed but also its thoughts are observed and controlled

- black toroid:  AI is built around human mind in its center as augmentation.


There are different symmetries here: becoming small is equal to becoming far. 

Active boxing is similar to the governmental control over AI development. 

Black toroid is equal to multi-peer mutual control in the net of AI users.

  1. Agree with all that, but also think that AI will takeover not AI labs, but governments. 
  2. A weak point here is that such global AI doesn't have overwhelming motive to kill humans. 
    Even in current world humans can't change much about how things are going in the world.  Terrorists and few rogue states are trying but failing. Obviously, after human disempowerment, individual humans will not be able to perform significant resistance a la Sarah Connor. Some human population will remain for experiments or work in special conditions like radioactive mines. But bad things and population decline is likely.

Models that are too good in deceptive alignment can deceptively look dumb during testing.

Load More