How to notice if machine learning or AI is real

Gradually it seems that all application and cloud service has been equipped with artificial intelligence and machine learning. Presto! They can now perform different magic.

Most of the marketing tactics around AI and machine learning is deceptive, making false promises and listing the terms they don’t apply. In other words, many BS are being marketed. Make sure that you do not fall for those snow jobs.

Before I discuss how you can check if the software or service uses AI or machine learning, let me explain the meaning of those terms:
Artificial intelligence is a broad category of cognitive technologies that allow planning, perception, communication, learning, situational reasoning and the capability to manipulate objects for planned intention. These technologies in different combinations help to build software units or machines that act as the natural intelligence which other animal species and human possess. It is similar to natural life's knowledge that varies significantly across species, so as is the intelligence of AIs.

AI intelligence has been a popular pattern in science fiction for over many years. It is a strong concept among technologies. For instance, MIT, the U.S. Defense Department, IBM, and Carnegie-Mellon University have been working on the AI for decades; these showcase similar examples over and over again for a long period. The prospects are many, but lots of incremental development has brought us a little closer to making the promise a reality.

Machine learning is a division of AI. It refers to special software designed to observe results and recognize patterns, then make use of the analysis to change its behavior or direct people to better results. Machine learning does not require the type of cognition and perception that we relate to intelligence. All it needs is the right, fast pattern matching and the capability to apply those patterns to its recommendations and behavior. Human being and other animals learn in the same way: you recognize the design that works and do that every time while preventing what you detect that does not work out well. In contrast, a machine followed the instructions or programming that it is instructed to do. 

Snow job 1: Perplexing logic with learning

In recent years, there are many changes in machine learning, as a result of this, not all the machine learning declaration are snow jobs. The best way to detect is to ask the salesperson if the robot or software can adjust or learn on its own without any software update. Also, find out how you can train it. Training help to change your environment and provide desired results.

However, the things those marketers identify as machine learning merely is logic. Programmers have been using logic software for a long time to inform robots and programs what they have to do. A logic that is sophisticated can provide many paths for the robot or software to take, depending on the parameters the logic is designed to process.

Presently, hardware can easily run sophisticated logic, so devices and applications can appear to be intelligent and easily adjust on their own. But most do not work out well of their developers didn’t predict a situation, they cannot adjust on their own to manage the pattern-analysis-based trial and error as a real machine learning system.

If the real machine learning is available, a machine learning system is incorporated with all the necessary parameters its logic has put in place to "know." Unlike a real AI, it cannot detect new facts outside its programmed environment; it can only learn to interact and understand the programmed world on its own.

Snow job 2: The use of cloud technology or IoT makes it simple

Marketers prefer to follow new technology terms and make use of them on whatever they possess. Most of them do not understand the term, or they don’t care. All they want is your attention. You can recognize a snow job faster by checking buzzword-to-detail ration. If you only detect the buzzwords and technology “how” details are lacking, you will understand that it is the same old technology but new marketing tactics.

Presently, the internet and cloud computing are competitive, so they are usually at the heart of much new marketing. Both can perform a role in AI systems (AI precursor systems) and machine learning. It is the smart use and not the use of the terms that is a red flag.

IoT depends on traditional, networked sensors including a combination of local and server (cloud) logic. Also, both actuators and analytics perform some functions in the analysis. Altogether, these enable devices to look smart because they are programmed to adjust to different events they encounter automatically. For machine learning, they have vast inputs for the learning parts including outputs for adjusted events.

In the past, it seems impossible for cloud computing to open up processing and data storage capability. Devices don’t necessarily need to carry all components; instead, they can download to the cloud and hardware to support it. Presently, this is how Google, Siri, Microsoft’s, Cortana and Apple now work. They transfer your speech to the cloud that helps to translate it and recognize a response, and then sends it back to your phone. In this way, you do not have to go along with a mainframe or datacenter or keep it on your desk.

Of course, it is possible to do that before the cloud transfer server/client, but the cloud offers some magnitude with more potential than your typical data center. Presently, you can now store and process at any scale that the whole populations can benefit from it.

Snow job 3: Machine learning process is exceptional 

It is usually impressive the services that Google, Siri, and Cortana provide now. And the incredible things that developers use with tools like Microsoft’s Bot support with Cortana. But we can quickly recognize their disappointment in areas without programming; their noticeable flaws restored a simple web search that wasn't programmed to learn.  Microsoft, Google, and Apple are making use of machine learning at the back end to make the process look smarter.

If anybody thinks that an application, a machine or service is smart, you might undoubtedly get snowed. Also, people can use the word “smart” to replace mean “more capable logic,” a phrase that won’t change anything. If they don’t explain the meaning of “smart” on their offers, you should know that they think you’re dumb.

Many of the technologies tagged “smart” are not smart; they are just savvy. The difference is that smart requires cognition and intelligence; while savvy requires only data and the capability to take advantage of it (it is not by mistake that “savvy” is a French word for “to know”). A robot or savvy app is good, but it is not smart. We have not gotten there yet.

IBM's vaulted that Watson is not smart. It is savvy, fast and can learn quickly. It has been in existence in many forms at IBM since the early 1980s. However, if Watson is not smart, IBM should be controlling the world business by now. Watson won’t create new tax breaks, cure disease, solve world hunger or make peace in the Mideast, but it can help people to manage all sorts of actions if it comes with the right price.

If you keep that target in mind and you are getting AI precursors and machine learning in your business, you will be contented. But don’t wait for a sci-fi desire version such as Data from Star Trek, Philip K. Dick’s androids in Do Androids Dream of Electric Sheep? Or A Space Odyssey (inspired by IBM’s 1960s AI research!). Finally, don’t depend on sellers that sell their technology under such appearance.