The AI of the movies and books is here, right?
The Artificial Intelligence AI sci-fi world paints amazing, and sometimes scary worlds with AI.
- Matrix and its robot run dream world
- Her and the evolution of a relationship with an AI assistant
- Terminator and its DoD created neural network Skynet which invokes ‘bulky’ time traveling ‘assassins’
- Eagle Eye takes out the time travel part, ARIIA taps into IoT, communication systems, and uses its AI to take over executive branch decision making skills to run the government ‘more in the spirit’ of the founding documents.
- Blade Runner and its replicants which cross over and pass the concept of singularity
- 2001 and good ol’ HAL 9000 which has sentience gone wrong while guiding interstellar travel
We actually call this type of AI ‘Super AI.’ This, for good and bad, is the expectation of True AI and how good and terrible it will be within the timeframe of its implementation. This is what computer scientists, between the hardware advancements, componentization of software, improved interconnectedness and speed, and finally creation of 90% of data every year will eventually move to. When there are many models out there, and none have converged in our lifetime, the next or the next. Some brave soul or body will be hinting us soon enough though.
So If Not Super AI, What AI Do We Currently See?
We currently have and use Narrow AI. It is the precursor to the general AI which we are getting closer to. We’ll get back to Narrow AI (or ANI) in a moment. In the case of General AI (or AGI), while it has the required computing, it lacks the data. It is basically equivalent to a specific persona of a human (yet not all human types or capabilities). AGI asks us “What should we do?” type questions on a broader scale. It’ll handle those WISDOM products. Yet the ANI of today is not sensitive, predictive, or has sentimental inclinations that humans use to adapt rules and applications on a broad basis, yet can do so on very specific or narrow applications.
Meaning, AI has clearly entered the phase of tactical applications. This is the first phase of AI or ANI – Artificial Narrow Intelligence. It is narrow. Also it can answer Jeopardy questions. Furthermore it can drive cars. Or it can be an entry-level assistant. It can even solve these narrow tasks. With that all said, if we are in Narrow or Niche AI, using the industries we focus in Emergency/Hazards, Science, Energy, Health, CIO Services, Natural Resources/Land Management, and Transportation/Travel, we’ll provide some examples of actual ANI.
Is The “Narrow AI” In My House A Smart Device? Kinda, Sorta?
On a broad scale, it’s not really in your house yet. “Smart” devices most of the time are devices with pre-programmed rules, additional sensors and internet connected feeds to then apply those pre-programmed rules. These environment sensed changes then adjust the logic of how the smart device will operate and interact, or not. These devices we think are AI, as they are labeled smart, aren’t really AI, rather they adjust to environmental events – time change, light sensors, additional data feeds, and then apply rules to a pre-programmed range of said rules. The intelligence of the rule is still in what indices, logic, and ranges which are programmed in by humans. AI, by contrast, is more about reading the sensor, source data or content, and then learning new rules.
Don’t get comfortable yet. Those assistants are saving your every phrase you prompt it with (in theory, not the unprompted as that would be illegal, yet we digress). Other IoT sensors are sending its environmental, motion, event, and action triggers back to HQs and making those obfuscated, anonymonized databases to inform THEIR AI and then sell those patterns off or use to inform their products. So the AI itself is not in your home as a device, yet the service behind it can be AI powered in some cases.
So Data Is Advancing Narrow AI? …Almost
Yes, that sensor data from smart devices is informing how we advance in the board space. At Xentity, we’ve discussed for many years that AI is on the forefront – moving into WISDOM for AI will require a lot of data and computing:
- 2019 – The Importance of Discovery in Metadata – Now That We Know What Metadata Is, What Is The Current State?
- 2019 – Data Patterns – Changes, New, Old – The data science tools can be on workbenches to get your workforce trained, playing, plugging and checking out the AI, ML, DL tools.
- 2016 – Two Geospatial Architecture Issues With Cognitive Processing: Veracity Is Still Struggling To Support Integrated Information Products – We Need To Learn From These Challenges To Move To Knowledge Products
- 2013 – Data Science Research Areas Punch List hits many rules-based and AI concepts
- 2012 – How Do We Stay On Top Of The Evolving World Of Data Science – How Can Integrating Data Enhance Products, Management, Applications, Remote Sensing, Knowledge Building, And Culture Impacts (Positive And Negative)?
How AI Learns
AI learns from its inputs,. Then it generally refines those initial rules in the hopes that most AI then creates rules which it should keep. Most AI today still comes up with rules that aren’t better than what is pre-programmed. Chatbot fails are abundant mostly in the training data coming from people messing with it. Meaning AI will fail if the data used to train does not give it the statistical averaging needed to create the confidence internals to use those rules in x situations. That is the amazing thing about human development.
Those who adjust their rules constantly and learn to keep, throw out, modify, split rules, and merge rules have better chances of success against their baseline situation. Some AI is advancing quite well with training data. We do have self-driving cars that can handle most human rules while also handling changing weather conditions, constructions, and things darting out in front of vehicles.
The trend in these blogs hit home the point – AI or ANI requires good data to make it happen. Enough volume to help the machine learn, enough veracity to make good rules; enough variety to consider multi-dimensional complex challenges; and at the right velocity to remain temporally relevant. Yes the 4 V’s of data.
In Summary: no need to be afraid of AI… yet
We know, we know, it’s popular to be wary of AI and what it’s capable of. And like any other new or emerging technology, there is always reason to be cautious. The AI future of Smart AI and Robots are coming:
That’s how it is when you’re stepping into unfamiliar territory. It is unfair to write off something that could have a positive impact on industries because you’re only considering the possible negative impact. This is especially true when you look back at all the aforementioned examples of how industries are using AI right now. Also, how much it already has helped those very industries.
If any AI that is being used or will be used in the future starts singing about a daisy girl or starts talking about how ‘imperfect’ humans are like they do in the movies, you might want to panic and we’ll gladly eat crow on this one. You know, if we haven’t been enslaved already by AI overlords.