Researchers in Maryland claim to have made a significant step forward in machine learning, and it has potentially game-changing consequences to enterprise in general.
Machine learning which relies on neural networks is referred to as specific AI. While it mimics human thought patterns, it can only operate within the boundaries of the data sets its been programmed with.
This means the ML can only follow certain predetermined steps. However intricate the actions are it performs, from recommendation engines to autonomous vehicles, a specific algorithm needs to be put in place to account for it.
This is why when something completely unforeseen event happens an ML systems fail. No matter how intricate it is, it's only capable of so much adjusting. For a much more malleable experience, you need to enter the realm of general artificial intelligence.
This is the hurdle the researchers at Z Advanced Computing inc., (ZAC) claim they have surmounted. Their website explains that they are 'developing a revolutionary technology leap in Artificial Intelligence, analytics, recognition, and search, based on General AI.' This general AI platform they are developing will enable 'fast, accurate, and reliable recognition and search, e.g., for people, objects, patterns, and actions, within images and videos.'
It might seem insignificant, but if made functional, the potential applications of this advancement are boundless. The team use the example of a shoe. Despite its mundanity, a shoe has a number of complex variations which are hard for a machine learning system to reconcile such as size, color, height, patterns, style etc. As the team explains, '[A neural network] can only generically recognize shoes, e.g., ‘brown boot,’ and it is incapable of detailed recognition, especially for small features (e.g., bands, rings, and buckles)'.
Image courtesy Z Advanced Computing Inc.
However, with the new software they have developed, the ZAC Cloud, they are now capable of using general AI to recognize a plethora of distinguishing features by simply exposing the system to the image. More impressive yet, there is no need to input the shoe's dimensions and they can enter images taken from a variety of angles and it will still function.
The other significant point of progress they have made is, unlike specific AI where the addition of new parameters or information can take significant retraining, this system does not require as deep an overhaul. With ZAC Cloud, one can just input new directives over old ones. This adds a layer of speed and flexibility to the process which has significant practical benefits. ZAC executive, Bijan Tadayon expands in an email released, 'With our demo, we have demonstrated general-AI technology for the first time in history, as the tip of the iceberg, developed by us to revolutionize AI & Machine Learning.'
This would open the door to medical professionals receiving near-instant diagnoses from uploaded 3D images, e-commerce sites would be able to make alterations and additions to products and add a host of new descriptors with relative ease. Even law enforcement can use it to assist with speedy facial recognition in times of a crisis in say, a manhunt.
This is why the Tadayon pair, Bijan who is the CEO and Saied the company's CTO, are referring to this advancement as the 'holy grail of AI and Machine Learning'. And coming from a pair who graduated top of the class at Cornell, hold dozens of patents, and have worked with a number of startups and technological ventures, it isn't a claim which can be completely dismissed out of hand.