AI World is a yearly gathering of professionals applying advanced AI solutions to solve enterprise challenges. They gather to exchange lessons learned and best practices as well as to track the latest technology which can aid them in their quest to improve the application of thinking machines to real world problems. Attending and helping shape several sessions at the event has been a good way for us to maintain our Special Series on Artificial Intelligence, which is maintained for member reference. This post provides some insights worth highlighting to our members.
OODA participated in AI world this year by organizing a full track of sessions aimed to bring insights into security, ethics, transparency and bias issues with AI. We hosted discussions and panels featuring real experts who were able to shed light on what is working and what is not in this domain of cybersecurity for AI. We capped our session off with a panel of experts on the topic of Due Diligence for AI centric companies, since increasingly due diligence is required to really understand the strengths and weaknesses of a company with AI technologies. This due diligence may be required on a startup before investment, on a VC funded firm prior to another round of funding, or a firm about to be purchased by a major PE firm or strategic. Knowing the value of an AI capability requires knowing if that firm is secure or not. We also participated on a panel which focused on the role of AI in misinformation and disinformation.
Some Take-Aways From AIworld:
Business leaders have been seeking advantage in AI for years and there are plenty of proof points of businesses making optimal use of AI over data to succeed. There are also plenty of lessons learned from failed approaches. One thing most successful AI projects have in common is the involvement of leadership. Something many failures have in common is an attempt to bite off too much and do it all at once. Another issue is an attempt to just automate everything and hope that AI or ML driven advanced analytics will flow from that. Companies can derive value from their data using AI and ML, but it takes vision, strategy and work. Businesses in most sectors are also finding that they are limited in AI progress because of the regulatory environment they work in. AI solutions must make things easier on firms to comply with rules and regulations, not harder.
AI solutions that take the above into account require the selection of the right architecture and models, an ability to protect algorithms and data, an ability to explain results and a focus on fairness.
The many use cases we lay out in our Executive’s Guide to AI are still relevant, but we are adding a few new ones to that reference based on the event. Here is an update on use cases in key domains:
- Manufacturing AI in manufacturing includes both optimization of current production paths and also design of future methods of manufacturing. One topic of high interest in the community today is the use of “digital twins.” First examined as a way of modeling design of real world equipment, it then became a working way of predicting failure and troubleshooting. It is now being used to optimize manufacturing by enabling AI over data generated by digital twins.
- Healthcare is a topic of interest to most all of humanity and since data is involved throughout most all aspects it stands to reason this is a hot area for AI. Personalized medicine evokes concepts of tailored treatments that will improve patient outcomes based on the needs of the individual. Research into personalized medicine is very data intensive and involves a wide range of sources which may include data on the human genome and at times the specific genetic code of the patient in question. AI in healthcare also includes improved administration, billing, insurance and services. Other solutions include rapid diagnosis of ailments and rapid assessment of imagery including xrays and MRIs. These solutions are already saving lives and in the coming years will result in even earlier detection of problems.
- Pharma AI in Pharma is closing related to concepts of personalized medicine but is a discipline of its own, including optimization of research into new medicines and integration of AI into the pharmaceutical business model.
- Energy AI in energy is helping optimize the full lifecycle of energy exploration, production, distribution and use. AI helps target most likely areas of oil discovery to target exploration. Metrics on oil wells are tracking potential issues including equipment failure and maintaining optimal production. AI use cases we saw at AI world include optimization of building energy use based on occupancy and dynamics inside buildings
- Retail and eCommerce AI for retail and eCommerce is helping businesses better understand customer buying behaviors and patterns. There are certainly issues of privacy and security here, but customers are voting with their pocketbook and companies are using AI over data to keep improving what is done for customers.
- Telecom and Mobile. Much of the AI we have seen in this domain is around customer satisfaction (or lack thereof). But AI is also being used to optimize network design for optimal throughput. Fraud detection and AI for cybersecurity are also key topics in this domain.
- Financial Services AI in Finance, including banking and insurance, is delivering capabilities right now including counter fraud, improved cybersecurity and optimized customer service. Perhaps more than any other sector, AI here must be carefully built to enable transparency and security and traceability of decisions to ensure compliance with regulatory environment.
Other Context from AIWorld:
- An interesting comment from Sam Ransbotham of MIT regarding AI in business. He indicated his surveys show 84% of company executives think that AI will give them competitive advantage. Obviously this is not mathematically possible that all 84% will have advantage, so this leads to the business disappointment, even when there are gains.
- Gartner projects that by next year the vast majority of enterprise customer interactions will be managed without a human. Doing this with success requires comprehensive strategies for ensuring a good customer experience including a well thought out data strategy to ensure the customer gets the results they want. This presents challenges for data architects, who must also ensure compliance with regulations and protection of customer privacy. ML is providing useful solutions here.
- An Accenture survey discussed at the event indicated more than two thirds of organizations foresee trouble ensuring data integrity of AI solutions. Even more cited challenges in adapting AI logic and reasoning into their use cases. Most all said they see big challenges in integrating AI into their back office functions.
- Innovations in AI from academia are expected to continue. We see it as a positive trend that major universities are not just focusing on AI, but on ethics around AI. The many problems that are with us today are being addressed in academia and that gives us great hope that methods of resolving them will be coming.
- MIT’s famed professor Alex Pentland (Director of MIT Connection Science and Human Dynamics Labs) underscored for attendees that in AI, “the winner is the person who has the most data.” He added that this is probably not you! But if you collaborate, together with others, more data can produce more results. We are not sure if Alex’s observation is relevant to real world AI solutions, but it sure makes a good sound bite. This is the trouble with listening to academics. You really need to filter what they say based on what your business model is. We can give example after example of AI working on smaller data sets, some of it making life saving decisions (see tech review below, for example). Another example of Pentland’s guidance: “If you create a data lake, you should be fired,” he said. “You’ve just told the bad guys where to find the data!” Sorry, but we do not believe anyone should be making hiring or firing decisions based on Pentland’s perspectives from his MIT perch. We did love his endorsement of building in interoperability and API’s for data access, but this is something we endorse because we have seen its benefits to real world business.
Interesting tech at AIworld
We spent time with almost every firm on the expo floor at AI world. There are many businesses in this space including products, services and consulting firms. We track 100’s in our online reference at CTOvision and maintain more detailed references to many in support of our due diligence services. Please ask us if we can be of services to your market research activities or if you have any other questions. Here we want to highlight just a few of the most interesting firms we saw:
- AI.Reverie: This is a simulation platform that generates advanced text data that has known features your AI should detect. Using this synthetic data can improve your AI because you will see how your system is being trained and you can also evaluate its performance in production.
- AInfinity: They produce advanced solutions that combine AI and IT Ops/DevOps capabilities. This enables developers to field solutions faster and in more reliable ways.
- CampTek: Providing RPA as a Service. RPA is clearly NOT AI, but the concept of RPA is related and is frequently a first step in a more comprehensive AI program.
- CapeStart: Providing outsourced data prep services and software development.
- Capice: Provides machine learning tools as a services.
- Firefly.ai: Gives AI power to any business with data via an automated machine learning SaaS solution.
- Jaxon.ai: Dramatically increase the accuracy of machine learning by increasing labeled data.
- Kyndi: Very intelligent process automation platform which provides explainable AI as a feature built in vice tacked on. Works very well with RPA.
- Lazarus: Using patient health data to improve early cancer detection and save lives. Improve accuracy of diagnosis. Doing this with pictures taken by smartphone or at times review of text based test results.
- VMware: We were amazed at how complete their vision was for support to the future of IT and since the future has a big focus on AI, VMware has rightly built that into their entire framework of offerings.
This is just a sample of some of the more interesting firms at AI world. Please get in touch if you would like more insights.
What do we recommend based on the above?
- Prudent planners using AI will stay focused on use cases that are not reinventing the wheel. Look for use cases where there are already lessons from others. Contact us with info on what your business needs are and we can more than likely connect you with someone in your sector who is working AI.
- It is ok to aim high in seeking results, but remember, aiming high might mean making a 10% improvement on a process that repeats year after year. You don’t have to go for 1000% gains as soon as you switch on your project. Showing incremental improvement is good!
- Enterprises with AI needs will have to decide on how much work will be done in house and how much will be done by commercial services or commercial product. There are trades that will be different for each enterprise but we recommend your bias be towards leveraging commercial services and product. Use your staff to manage and lead requirements and ensure projects are well managed and stay on track.
Review our other special reports on AI at:
