AI News Roundup — July 2020

by Gabriella Runnels and Macon McLean

Opex Analytics
The Opex Analytics Blog
5 min readJul 31, 2020

--

The AI News Roundup provides you with our take on the coolest and most interesting Artificial Intelligence (AI) news and developments each month. Stay tuned and feel free to comment with any stories you think we missed!

_________________________________________________________________

Acidic Language

Photo by National Cancer Institute on Unsplash

What do amino acids and typical language have in common? An unlikely question, but one nonetheless asked by researchers at Salesforce who recently discovered that natural language models can be applied to protein structures.

The transformer, a neural network type that was introduced back in 2017, has come into vogue in its short existence as the model architecture of choice for many natural language tasks, supplanting long short-term memory (LSTM), gated recurrent unit (GRU), and other recurrent neural network types. Here, the transformer is used to infer representations of proteins’ overall structure and binding sites, which researchers can then compare with the best available traditional (i.e., biochemical) understanding of the protein to improve understanding and fuel further inquiry.

Accelerating Artificial Organ Growth

Photo by Mathew Schwartz on Unsplash

The process of creating artificial organs begins with stem cells, which transform over time into cells belonging to a specific organ in a process called differentiation. However, when cultivating retinal cells in particular, the process is fairly random: cells in the same batch often differentiate at different rates, making any would-be standardized retina growth process more complicated and less reliable. Current methods of detecting differentiation in cells rely on fluorescent proteins, but this methodology is unsuitable for organs cultivated for transplant.

However, researchers have recently discovered that neural networks are capable of identifying retinal cell differentiation with 84% accuracy — seventeen percent above the human rate. This incremental advance paves the way for future advances in treatment of retinal disorders, and can also serve as the basis for similar networks trained on other organs’ cells.

Fax and Figures

Photo by Markus Spiske on Unsplash

Every data scientist is all-too-familiar with the problem of messy and incomplete data. But most of us rarely work on issues as pressing or time-sensitive as tracking the spread of a deadly virus. Furthermore, we don’t typically have to deal with the greatest obstacle to Houston Health Department (HHD) data collection: the fax machine.

The New York Times reports that the HHD, which is currently trying to deal with one of the largest outbreaks of COVID-19 in the U.S., recently encountered problems when their fax machine began printing hundreds and hundreds of sheets of paper — lab results from coronavirus tests. But this unwieldy data transfer method isn’t the only roadblock facing health officials who are trying to track the disease: inconsistent data quality and disjointed digital health care systems are making it difficult for officials to control the spread at every turn.

AI Predictions

Photo by Muhd Asyraaf on Unsplash

Forbes’s “AI 50” list of companies to watch recognizes U.S. business that use AI in an exciting or innovative way. This month, Forbes talked to the founders and CEOs of some of these companies about how they think COVID-19 will affect the future of AI.

One CEO thinks the pandemic will have a positive impact on people’s perception of AI, as “AI will become associated with safety while human contact will become associated with danger.” Other AI leaders believe that the drastic societal changes due to COVID-19 will open the door to AI applications in healthcare, transportation, and the workplace, potentially making our lives and businesses safer and more efficient.

What R They Talking About?

Photo by Brian McGowan on Unsplash

Before the coronavirus pandemic, most of us non-epidemiology folks hadn’t heard much about the reproduction number R. But now, politicians around the world are seemingly fixated on this one little metric.

So what exactly is R? Put simply, it’s a measure of how infectious a disease is. In technical terms, it’s the average number of people to whom an infected person will spread the disease. If R is greater than 1, then the average infected person will go on to infect more than one other person, so the outbreak will continue to grow; smaller than 1, and the disease will eventually peter out. Although it can be addicting to refresh websites like rt.live to see what the R value currently is in your area, experts warn that this metric is not always completely reliable, and it should not necessarily be taken at face value. To understand some of the assumptions, nuances, and concerns of putting too much emphasis on R as a guiding metric in public policy, check out this useful guide to R from Nature.com.

That’s it for this month! In case you missed it, here’s last month’s roundup with even more cool AI news. Check back in August for more of the most interesting developments in the AI community (from our point of view, of course).

_________________________________________________________________

If you liked this blog post, check out more of our work, follow us on social media (Twitter, LinkedIn, and Facebook), or join us for our free monthly Academy webinars.

--

--