Matching Inflated Expectations with AI
Supplying advanced technology is undoubtedly an exciting job. However, there are downsides to working with Artificial Intelligence (AI) and providing solutions to problems which were, until now, a human prerogative.
One inevitable annoyance is what we can call the Curse of Inflated Expectations. Being on the cutting edge of the hype leads to dealing with companies having often unrealistic opinions about “the art of the possible”.
This is especially true for the world of Automatic Learning, the most common realization of AI these days. Given the technology breakthroughs we’ve seen in recent years, it’s easy to believe that Deep Learning can solve any problem and, being an automatic data-based approach, you just shovel in the data and press a button. Flawless and effortless. Except it is not.
The Silver Bullet
Many companies are perfectly willing to invest months in well-known tasks like starting an e-commerce site, and they would be more than happy to have a 1-year ERP adoption plan. However, they may not be willing to dedicate a significant amount of time for automating an important document analysis process. Everyone has seen a demo where you just upload a bunch of annotated documents (i.e. of which you already know the expected result), train the system and, after some minutes, magic happens.
But many real-world Text Analytics problems are much more complex, requiring deep human expertise and skills and, quite often, an enormous amount of training data, which may not exist at all.
At Expert System, we often encounter these problems. We know there are some cases that can be solved by just correctly implementing the right Machine Learning Model, but these are not the majority of Knowledge Management problems that Enterprises face.
The Curse of Inflated Expectations leads companies to the idea that someone has already invented the silver bullet; now all you have to do is load and fire it.
A recipe for exceeding expectations
The Expert System technology, Cogito, is a large platform that includes several tools and components that serve as a toolkit for Knowledge Engineers, Data Scientists and Integrators to solve all types of Text Analytics problems. When it comes to being effective, using the right tool for the job is the key factor.
This is why Cogito exploits Automatic Learning models and techniques in several components, but it is far from being just a mere AI toolkit that integrates libraries and frameworks. The core technology exploits the Cogito AI engine and a large Knowledge Graph for each language for real Word Sense Disambiguation, putting Cogito many steps ahead of other technologies currently on the market. A powerful mix of cutting-edge technology tools is the only way to tackle complex enterprise use cases.
In the end, there are many ways to exploit the power of Automatic Learning for Knowledge Management of documents. We commonly leverage multiple Learning techniques:
- To extend or customize the Knowledge Graph
- To classify documents onto custom taxonomies
- To extract custom concepts, entities and relations from documents
We will describe each of the above techniques in upcoming posts.
Moreover, the Expert System platform is open to plugging in new AI models and techniques to support, reinforce or replace existing ones. Tomorrow, we will have better tools than today, and we want to make them available to the market in the most powerful and effective technology suite available.
Chief Scientist, Expert System
Want to learn more?