Many companies believe they are adopting artificial intelligence, but the reality is only about five percent are really doing it.
Last week at MassTLC’s Enterprise AI Workshop, Indico Data Solutions’ CTO, Slater Victoroff, provided insights into how companies can begin using AI and reap the benefits of this emerging technology.
As a start, Slater defined the fundamental terms:
- AI – any computer program that automates a process assumed to require human intelligence. This can be done through a number of tools including – but not limited to – machine learning and deep learning. In other words, AI flips the method, from putting the burden on the human to program in the machine’s language, the machine must learn to adapt to the human.
- Data Science – a generic set of skills that includes machine learning, deep learning, and transfer learning used to produce enterprise value from data through understanding, automation and optimization. This is not just the algorithmic piece, it is anything that is using data to deliver value.
- Machine Learning – a field of computer science in which the computer is not explicitly programmed to do a function, but is able to fill in the holes on its own to fulfill a desired outcome.
- Deep Learning – a set of machine learning algorithms that work well on unstructured data. This is really the first time we’ve been able to get this far using neural networks, which have been around since the 1950s. These can really achieve near human or, in some cases, beyond human performance.
- Transfer Learning (Indico-specific) – a deep learning method where a model developed for a task is reused as the starting point for a model on a second task.
Slater went on to provide the four “building blocks” that he sees as “must haves” for a successful AI adoption.
- Data – Data prep is single most important aspect of machine learning, and to obtain the desired outcome, the data must be labeled.
- Expertise – You must have subject matter experts who understand the business problem and as well as at least some level of data science understanding to interpret and refine the approach. Note, there will be a lot of refinement.
- Compute – For machine learning systems, you are can use CPU. However Deep Learning systems require GPU. GPU was originally created for video games, they are only a couple of years old and most large cloud providers have them, however, most data centers don’t support them yet.
- Definition of Success and ROI Hypothesis – You must have a defined outcome that is connected to tangible business benefit. This is a very expensive process in terms requiring investments in hard and opportunity costs.
Slater closed the evening with Indico’s “Prime Elements” of Enterprise AI:
- Classification & Regression – classification of data, this is basic but so important.
- Unsupervised Discovery – a tool that requires you to give it a bunch of text to help frame your problem.
- Comparison – instead of taking one piece of text or documentation, you take two pieces of text and compare them to assess if they fit together somehow
- Search & Extraction – when you have a large set of documents and you need to pull from there.
Interested in learning more about AI and machine learning? Check out our upcoming MassIntelligence Conference on November 9th.