Is The Path To General AI Clear?

A lot of progress has been made in AI lately. Any application that could benefit from automation or benefits from increasing the amount of data you can process has had some kind of AI attached to it. Because automation requires intelligence to work discussion about robots involves AI as well. Automotive, finance, internet searches, music channels, advertising and product suggestions, news sites, lots of business apps that require decision making, etc.

Algorithms have taken the spotlight from software (which years ago took the spotlight from hardware) and the word is used in advertisements to showcase their product is superior. Which is funny, because logic and decision making algorithms have been around forever but they were very often not taken seriously or implemented well or just very simple. But either way it isn't AI. Not the way that Spielberg presents it. Their all just algorithms. Take a huge amount of data, process it, make some decisions, then repeat every 10 milliseconds. That's AI right now.

But as awesome as "AI" is, is the path to general AI clear? Let's break down what we have and see if there are steps to what we want (i.e. Lt. Cmdr. Data). I like the idea of using a search engine as the "state of the art artificial intelligence", emphasis on intelligence.

1. Current search engine algorithms
2. First answer efficiency: Given an immense repository of information (i.e. a cached copy of the internet that has been marked up), the best test of a search engine is not the number of search results or how valuable they are, but how good it can understand your request and then responding with the correct information. Almost all queries are simple command and response but you can build from there.
3. Algos: Search engines that are knowledgeable and understanding are one thing but they are not by themselves making decisions about what actions to take. But algos do. And they are imbued with the power to make those decisions. My favorite example are "robot advisors" now offered by financial firms. They manage your portfolio by optimizing your mix, minimizing tax losses, given you alerts when changes are made or things go wrong.
4. Basic Life Skills: if you can teach a robo advisor to make decisions about your financial plan you can teach it to make decisions about the most basic of life skills. Nutrition, budgeting, transportation times, shopping, etc. This is a larger, but very manageable scope, but it highlights the concept that at leaset some of what being human is not exestential. And any general AI or non-human conscious being would have to somehow manage these issues. Interestingly enough, we are seeing services become available (with AI of course) that will manage these tasks for the already existing humans.
5. Human-Like Automaton: If an algo can be taught those skills it can be taught to control motors and perform basic tasks in a factory. If you could wrap a set of job skills and life skills into one set of algos, with an algo to manage to algos, and put that inside an android shell (humanoid or otherwise) you have yourself a basic human-like automaton. It doesn't have any feelings, it would be crude, it couldn't hand a wide variety of circumstances and it certainly wouldn't be human. But could you tell? Interestingly it's "brain" would probably to cloud-based due to the limitations of local computation power and battery energy density. And if you buy that then basically this is close (but not quite) to any smartphone that's on the market. They have large amounts of local and cloud storage and processing power and a huge number of apps many of which use algos to automate tasks.
6. Advanced Human-Like Automaton: I haven't discussed the ability to learn or adapt. Algos don't learn or adapt. There is a human to study and rewrite when needed. Any worthy automaton or android would need the ability to study it's environment and adapt it's own algos. This is possible and within reach. And with much development and sophistication any algorithm can be smart to "lookup" other algos when a problem is encountered. For example if a train is late don't just shut down,
   call up the mapping sw and program a new route. If you noticed this kind of adaptability in a robot I think the average person would be surprised. But really I don't think it is that much of a stretch of the imagination that with a few years of effort and a mature, sophisticated selection of algos one could write a static schema that would be pretty darn robust. I also haven't mentioned purpose. That's very existential. But an automaton could be preloaded with a static "todo list" or sorts. If your smartphone had a schema like this, with all it's algos and apps as well as a todo list then we're pretty much there. Stick that in a shell you'd have a soulless and admittedly pretty rigid "person".
7. Pre General AI: What if an automaton were able to write it's own algos or update it's own schema? This is a large step forward. Tuning an algo is one thing, but replacing the logic or perhaps using new information to update and write it's own algos is another league altogether. Or rewiring its internal epistemology. Or rewiring/rewriting the management algo. These are all huge tasks. Where is research on efforts to accomplish this? I don't know. This single step is probably more effort than all the other efforts combined.
8. General AI: ie algos with purpose. Pre General AI androids would be flexible, capable and altogether very human like but they wouldn't seem human. They wouldn't choose their own destiny or their purpose. I think the path to general AI is complete once an entity takes control of it's own life and chooses it's own destiny, or in other words decides what goes on its "todo list" and the priority of such.
9. Concious Beings: will general AIs be beings? Will they have emotion? Is emotion anything but a set of priorities or just animal like compulsions? And what about consciousness? Is it a switch that can be flipped on with the right hardware? Or is it simply a function of the size and scale or hardware?

The path is more revealing than I would have thought. But the path to general AI needs some work yet.