Artificial Intelligence: How Olenick AI Could Save Your Company
What is Artificial Intelligence?
At its core, Artificial Intelligence (AI) is intelligence demonstrated by machines. Pretty simple right? Not so much.
The origins of artificial intelligence were born in the in the 1940’s and legitimatized in the 1950’s when Dartmouth College created a field for it. Fast forward 40 years, and you may not consider some intelligent devices… artificially intelligent.
Between technological advances and Hollywood’s interpretation of self-healing robots and Matrix-like worlds, AI has gotten much more complex! Honesty, the concept of artificial intelligence is very misunderstood; a huge part of the problem is the lack of a consistent definition. Does AI need to understand its environment and react? Does AI need to learn from past mistakes and self-correct future actions?
Regardless of how artificial intelligence was born, today it is generally referred to as “machines that respond to stimulation consistent with traditional responses from humans, given the human capacity for contemplation, judgment and intention.” While the AI definition may evolve as innovative technology is developed and new James Cameron films are released, you can sleep well at night knowing Olenick’s AI technology wants to help you!
How Will Olenick AI Save My Company?
You are the IT security manager at your company. It is your responsibility to keep up to date with new viruses, phishing techniques, software updates, bugs, etc. You spend your day trying to stay ahead of the information curve, while fixing Patty’s computer because she clicked on that Nigerian prince email for the second time this month.
Time. You simply do not have enough time. While you were fixing Patty’s laptop you missed a zero-day code exploit that was just announced that affects most of your web servers. Your server was targeted, and now you have a much bigger problem on your hands.
The Olenick Advisory. The Advisory is a custom Olenick solution that continuously monitors the Internet for threats, issues, attacks, updates, bugs, etc. After the Advisory’s proprietary AI seek engine finds a news article based on your company’s criteria, an automated series of notifications begins to occur. This early notification system not only warns you ahead of the attack, you also don’t have to monitor the 10 IT news sites you read every morning. In fact, the Olenick Advisory monitors hundreds of sites every hour of the day.
How does Olenick use Artificial Intelligence?
The Olenick Seek Engine is built upon several layers of Artificial Intelligence. After your company profile has been loaded into the system (software commonly installed, computer model information, network infrastructure, etc.), the seek engine will optimize two custom modules to find articles that are useful to your organization.
For example, say during your advisory profile creation process we learn your company installs Google Chrome and Symantec Endpoint Protection. Those two applications along with others are loaded into your profile as separate terms. As the Olenick Advisory system continuously monitors the Internet, we search and rank articles based on these terms using artificial intelligence.
So how does this magic work? Glad you asked – let me introduce our Connotation and Fuzzy modules!
The Connotation Module evaluates strings by meaning and not just the exact terms of the executed search. When this module is used, matches are found by implementing a few procedures:
- Word-breaking: The process of separating your string into individual words based on word boundaries. Word-breaking for languages like English is somewhat straightforward because spaces separate words but the process becomes very important for languages that don’t, like Japanese or Chinese. In addition, words like “database” may be split into “data” and “base” to help broaden match results.
- Word-Stemming: The process of reducing inflected words to their word stem or root form. For example, any blue variations of the word “Affect___” are reduced to the base word “Affect.”
- Thesaurus-Extending: The process of identifying similar words based on synonyms, an almost reverse word-stemming process identifies relatable root words.
The Connotation Module is very useful for finding organic match results if your search term(s) do not match the search strings exactly.
For example, a Connotation search term of “Error AND Affect” would flag this article:
“Bugs” is stemmed to “Bug” which is a synonym of “Error.” “Affecting” is stemmed to “Affect.” Even though some translation steps were necessary this article would receive a decent score from the Connotation Module using the search term “Error AND Affect.”
The Fuzzy Module evaluates its strings for fuzzy or non-precise matches. Fuzzy matching is a technique used to match words that may be less than 100% perfect. However, we do include exact matches if found; that would just be rude if we didn’t.
A common need for fuzzy searches is unfortunately due to misspelled words. Most Fuzzy Logic is based on some form of the Damerau-Levenshtein algorithm which returns a value for measuring the distance between two sequences (aka Edit Distance). For example, if the word we’re looking for is “Start” and we find “Strat” the Edit Distance would be 1 (we only need one switch operation for the correct spelling). The farther the distance the less likely the match will yield a useful result.
The Fuzzy Module also allows us to search for terms within a certain distance of another word. This operation is extremely useful for everyday terms that may frequently appear in articles, but when not grouped closely together mean less to us.
Take for example a fictional application called “Start” which is developed by Olenick. Searching for articles with the word “Start” would return too many false positives. Searching for “Olenick Start” may fail to find articles labeled as “Start: an Olenick application” or “Olenick’s brand new Start application.” Creating a search term which looks for “Start” and “Olenick” within 5 words helps promote articles that are most likely within our searching domain.
Fuzzy logic’s main pitfall is relevance. You could imagine searching for the word “Power” only to have article results returned because they contain the word “Lower” (1 Edit Distance value). To combat this issue, we leverage scores from both the Fuzzy and Connotations Modules and determine an overall ranking for our potential article match.
Putting it Together
As mentioned previously, we combine the Connotation and Fuzzy Module scores to determine an overall ranking for the article. If the overall ranking for the article meets a certain threshold then an automated alert is sent.
The system will also begin to build an internal weight matrix to know how much importance it should place on either of the modules depending on how useful previous results were.
Take for example a search term that yields a 100 Connotation score and a 0 Fuzzy score against an article. The system will review previous articles that the term returned and based on those results make a human-like decision if it values one module score over another.
Artificial Intelligence is simply not going away anytime soon. People will always find new applications and implementations to introduce machine learning and self-correcting. The Olenick Advisory is a solution to help make better use of your time and help call out issues that are important to you.
To stay up to date on future Olenick webinars and blogs, make sure to sign up for the Olenick subscription list here.