Many have a tendency to look at things like the Turning Test, various contests such as the X-prize for an AI to deliver a TED Talk, some even saw Deep Blue winning at chess as a sign AGI was approaching. While these things are important pieces on the road to an AGI one will never be developed from them. These are set static goals thus the humans developing the software will build machines to accomplish them not AI that also happens to be capable of completing these tasks. This is similar to the major issue with standardized tests. You aren’t teaching something to be smart, you are teaching it to beat the test even if it doesn’t understand the material.
Instead I’d look at the consumer level market. In chaotic real world where dynamic and versatile thinking is needed. A niche where it is required to make nuanced choices and solve a variety of problems. It would also need to fill this niche without being a full AGI. There are many tasks that require, at a bare minimum, human level thinking. Such as assembly jobs still performed by humans. An AGI is not going to manifest fully formed. It will be an evolution from a simpler program into one of genuine intelligence. This means a simplistic computer application that is functional at a current level but will be of greater use as it becomes more sophisticated. One that has to develop learning software, adjust to its environment, and make value judgements. What technology do I feel will evolve into AGI?
Intelligent Personal Assistants.
The modern day programs that will be AGI’s ancestors are the likes of Siri and Cortana. Both are developed and funded by the wealthiest and oldest software developers. That is just reason to believe they are capable, the reason I think this will evolve into AGI is simple need. As technology becomes more ubiquitous and information more readily available we will see a growing need to manage it. Intelligent Personal Assistants are primarily a management tool tasks simply to make use of other programs to meet your needs.
What separates these programs from others is that dynamic and predictive reasoning directly enhances their effectiveness. A user wants these programs to understand the intent of what they are saying not the literal meaning. It will also be far more effective if it can anticipate what a specific user prefers. This requires not only the ability to predict what a user will likely want but the capacity to do so for a variety of different people. It will have to learn the users personal tendencies rather than rely on scripted parameters. In fact some of the core designers of Siri are working on a new program dubbed Viv. Which will be design to learn user preferences and better understand our nuanced requests. I tend to be very conservative with my estimates on technology so I suspect that initially Viv still wont quite reach the level its developers hope for but it does show we are moving in that direction. Over time we will see these programs grow and evolve and I predict these proto-AGI will become so common place nearly everyone in developed nations has one. I know I personally want someone that can screen my calls, keep track of appointments, and even hold a nice conversation from time to time.