busySome observations ...
* Exuberant claims are almost always vaporware. If they don't come with at least demo software or hardware, you may ignore them.
* Most people working in the AI field aren't trying to create digital life. They're trying to create marketable projects. While it is entirely possible to knock up your computer by accident, it is not probable for that to happen while trying to create a simplistic product. And compared to life, all the products are simplistic.
* To create life, you need:
-- a growth medium capable of supporting great complexity
-- a learning system of some sort
-- an environment full of experiences to learn from
(Step 0: If you are not already a capable parent, take a comprehensive parenting class before attempt to create life. Do not engender life irresponsibly and be prepared to take care of all life you create. Congratulations, you just saved the world and prevented a science fiction novel.)
If you want to create digital life, you need a totally different route than product development. You create a seed program and equip it with the ability to learn, grow, and adapt. Then you turn it loose in an environment to have lots of different experiences. Provide new and more challenging experiences as it grows. And then wait, because developing significant complexity takes time, even at quicklife speeds.
Some of the best work in digital life actually involves virtual ants. Another deals with the evolution of locomotion.
* If you wish to develop generalized intelligence, or at least mimic that, then you need to stop overspecializing. Generalists are idiosyncratic. Generalized systems that work look like garbage unless you know what you're doing. The "junk" DNA? That's the Kitchen Junk Drawer. Every time a calamity smacks the biosphere down to almost nothing, evolution reaches in and pulls out something. Even! More! Awesome! than before. That's a generalized system in prime working order. The human brain contains strong connections in logical ways, but also a lot of others that look like noise. Those extra connections create plasticity, resilience, and intuition. If you're trying to build tidy code, you're doing it wrong.
* So if you are interested in AI, look at what is happening now, think concretely about what could happen next, and leave the elaborate predictions as an interesting thought experiment. Confusing steps 2 and 3 will kill your funding -- and everyone else's -- for a decade or so.
* And in all cases, make sure you behave ethically. Robots may not be people yet, but it is possible for that to happen, and missing that transition can make you a monster. Proceed with caution.
