Behind every “overnight success” is years of work in the shadows—and the same is true for artificial intelligence. AI chat bots may have exploded and captured public attention but scientists, mathematicians, and philosophers have been working on this since before the 1950s according to Harvard. For a long time we’ve used AI/ML for limited applications and now we see the opportunity to create efficiencies, make smarter decisions, supercharge employee capability, create new products, and have even more cat videos.
What’s changed is that these tools are out of the lab and available world wide to regular people ready or not.
Even CTOs are having a hard time keeping up with all the rapid changes and advancements in AI while trying to decide how to use AI to create and grow businesses. To help, here are six things the experts want you to know.
You may be able to do it, but no one may be able to afford it
AI may provide the best solution to your problem but the price tag to go from an idea to something working in production may be hundreds of thousands of dollars according to CTO and technology advisors I’ve interviewed . While this is likely to go down in price in the near future it can be challenging to justify the expense since most organizations are focused on keeping costs down.
Most of the work is cleanup
Experts estimated about 80% of the job is cleaning data and 20% is training and testing. It is a significant effort to do this cleaning and if you discover you don’t have the right data or enough data you’ll have to start over again. Sometimes teams will use data from a proxy group instead of their desired group and find out that they are off in their projections and have to rework everything.
Speed and accuracy really matter
How will data be served to the model? How can we fast serve it? A Fortune 500 company requires a response time to under 100 milliseconds and will keep shaving the time down to get under the bar. Speed isn’t the only important metric, the same company requires 95% accuracy before an AI-powered feature can be released.
Understanding your audience is important in setting your success metrics.,While an audience of professionals may expect perfection where a more casual audience is fine with getting a general idea of what’s happening. Nobel winning psychologist Daniel Kahemen found that while we forgive humans for making mistakes we expect very high performance from machines or AI.
People are always the hard part
Technology makes it easier and faster and AI is no different, but companies are run by people. Leadership teams will need to decide together what their AI strategy is, what investments to make, what rules or governance to put in place, and how to lead their teams. These are going to be challenging conversations that will require vision, strategy and people skills to navigate.
Explainability is a work in progress
Some B2B companies are using traditional forecasting and modeling instead of AI because an audience of experienced professionals is unlikely to put their trust in something they don’t understand or explain when pressed by leadership or clients. Additionally there is a risks models are not complying with regulations like fair housing or fair credit.
Double check, always
An LLM is the loud person in the bar who is always confident, not always right, and shouldn’t be trusted with the keys. They may be able to pass the bar but make up cases for a legal filing. So just double check.
Image: This is an old mixed media piece I did by cutting up the strips and putting them over a watercolor painting.
Kommentarer