Posted

This is the Burtch Works’ Monthly News Digest, the place to visit for a quick high-level breakdown of the month’s top stories from the world of data science & analytics.

This post was developed with input from Bill Franks, internationally recognized thought leader, speaker, and author focused on data science & analytics.

 

What’s Deepfake Bruce Willis Doing in My Metaverse?

There has been a lot of talk about the risk of AI deepfakes. This article comes from a different and very interesting angle. What if a deepfake is officially approved by the person being imitated? Does this take the output from a “deepfake”, which is a term with a lot of negative baggage, to a “digital twin”, which is suggested by the owner of a company creating officially approved deepfakes? These are questions that we’ll need to consider.

At first glance, it would seem fine if someone consents to a deepfake AI version of themselves being used in a movie, advertisement, or even a corporate training video (yes, that is one true life example in the article!). However, there are still issues to think through. For example, what type of notice does the viewer need to be given? In the metaverse, where everyone knows everything is fake to begin with, do the same rules apply as in real life? We can expect to see a lot more of what this article discusses in the future.

Read the full article here.

White House Unveils ‘AI Bill Of Rights’ As ‘Call To Action’ To Rein In Tool

As with any new technology, lawmakers are playing catchup with AI. The capabilities of AI today far outpace what has been considered in the USA’s rules and regulations. Earlier this month, the White House announced an initial draft of an “AI Bill of Rights”. As of today, the document’s recommendations are only suggestions. Of course, over time, Congress could act to give some or all of the White House draft full legal standing.

While on the surface, the key tenets summarized in the article sound reasonable, the real challenge will be figuring out how to implement them in practice. Companies utilizing AI should study these recommendations and consider them to be a blueprint for future regulations. Then, work to help shape whatever final regulations are enacted to ensure that they are both effective and practical. Leaving it purely to the politicians to try and dictate something as complex as AI is probably not the best path. The effort and expense for companies to get involved early could be far less than what it would cost to have bad regulations finalized later.

Read the full article here.

The Multi-Billion-Dollar Potential of Synthetic Data

While the idea of using synthetic data to build mission-critical analytical processes might at first sound crazy, it is an important and increasingly common reality. This is especially true when it comes to generating data to train AI models. This article outlines some of the many uses for synthetic data, as well as the relative ease in generating it as compared to collecting “real” data.

Whether training autonomous vehicle algorithms using video game footage or creating a myriad of fake faces that meet specific criteria, synthetic data is in wide use today. Further, there are companies that have started to focus exclusively on generating synthetic data of one type or another. As companies struggle to adapt to increasingly stringent privacy laws or to train AI for high-risk activities like autonomous flight, synthetic data is likely a part of the solution.

Read the full article here.

Medtronic Leverages Data Science to Launch Personalized Insights Program for People Living with Diabetes

Medtronic plc, a global leader in medical technology, is introducing a new program called My Insights exclusively made for individuals using a popular Medtronic diabetes management tool. Using an individual’s data, My Insights relies on the power of data science to provide personalized tips, trends, and reminders that patients can use to help with their diabetes management goals. The My Insights personalized recommendations are shared via a monthly email with educational content that is most relevant based on what an individual may currently be experiencing.

The MiniMed 770G pump system, which is available to people living with diabetes who are ages two and above, uses self-adjusting technology to help prevent highs and lows in blood glucose levels by automatically adjusting insulin delivery every 5 minutes to adapt to an individual’s unique needs. It also uses smart phone connectivity to view, share, and automatically upload insulin and CGM data, and provides access to future innovations via software upgrades.

Read the full article here.

Trick or Treat! – Bizarre Halloween Candy Courtesy of AI Tool ‘Dall-E’

AI researcher Janelle Shane has made a hobby of prompting machine learning systems to engage in the 2022 equivalent of some very weird improv comedy. Her latest endeavor includes asking Dall-E (an AI art creation tool) to paint a picture of the most popular candies in each U.S. state. The result? – Dall-E associates Halloween treats very strongly with candy corn across the board, which is understandably given its abundance during this time of year.

To keep from being buried under a mountain of neural net candy corn, Shane had to mix up the phrasing of the prompts she presented to Dall-E. She tried moving the name of the state to the front of the sentence or asking for product photos of candy bars or bags of candy. It’s not exactly clear what there is to be learned from this exercise, other than AI seems to think humanity’s over-supply of candy corn means it must be popular. Unfortunately, Dall-E doesn’t present us with many other classy confections to choose from should we find ourselves trapped at an AI-generated Halloween party in the metaverse this year, but this is definitely a great start!

Read the full article here.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.